Datasets:

DL02

/

TM

like 0

AIML Generative AI.pptx

A brief introduction to Generative AI AIML Synthetic Data Generation Team 922023 01 02 03 04 Generative AI and its applications Overview of important algorithms Neural Networks LSTM Transformer GANs Fractals Data Generator architecture References Agenda What is generative AI Generative AI refers to a class of artificial intelligence algorithms that focus on generating new previously unseen data that conforms to certain patterns or characteristics learnt from actual real world data Resolving data scarcity unavailability Enabling data privacy Improve accuracy of AIML solutions Image Generation on Text prompt Generating HRES versions of images Enabling Humanlike conversation Semantic Image to photo translation Video prediction Music generation Applications Neural networks are a type of artificial intelligence modelled after the structure and function of the human brain. Neural Networks are the foundational block for Generative AI algorithms A neuron in a neural network is a basic unit of computation that receives inputs processes them and outputs a result. Neural Network Neuron Recurrent Neural Networks processes sequential data by retaining memory of previous inputs. RNNs were the initial choice for processing sequences LSTMs Long ShortTerm Memory Networks solves the problems of traditional RNNs by retaining memory over long sequences. Recurrent Neural Network LSTM LSTMs are further enhancements on RNNs capable of better handling long sequences LSTMs Long ShortTerm Memory Networks are a type of Recurrent Neural Network that solves the problems of traditional RNNs by retaining memory over long sequences. Transformers in deep learning are a type of neural network architecture designed to process sequential data efficiently using parallel computation. Transformers have revolutionized processing sequential data by introducing the concept of attention Discriminator Loss Noise input Generated data Deep Neural Network Output Score Deep Neural Network Real Data Generated Data 1 0 Generator Discriminator Tries to replicate the real data distribution so that it can trick the Discriminator. Is trained to distinguish between real and fake distributions. The two models compete with each other in a zero sum game with opposing goals to optimize. Generator Loss MLP in VGANs MHAN in TTS GANs CNNs in DC GANs MLP in VGANs MHAN in TTS GANs CNNs in DC GANs GANs introduced a unique architecture for data generation Summary report Scoring Model Selection Model Elimination Load In our utility we have combined the best bits of multiple algorithms with an aim to generate most accurate synthetic data PSI score based Data connectors to load from various sources RAW DATA PREPROCESSING MODULE Clean Clipping Imputation Transform Encode Transform Scale DATA GENERATION MODULE SYNTHETIC DATA EVALUATION MODULE SETUP GRID Creates a grid of GANs VGAN WGAN CTGAN TTSGAN Network single MLP vs Diverging MLP Hyperparams Learning rate activation epochs TRAIN Training the Generator and Discriminator on every combination in the grid. GENERATE Generating data on every trained model STORE Store the generated data for further evaluation MAE MAPE MI metrics Cross correlation Comparative analysis Summary of NN for advanced users only References Research Papers Ian J. Goodfellow Jean Pouget Abadie Mehdi Mirza Bing Xu David Warde Farley Sherjil Ozair Aaron Courville Yoshua Bengio 2014 June 10. Generative Adversarial Networks . Retrieved from arXiv1406.2661   stat.ML Mehdi Mirza Simon Osindero 2014 November 6. Conditional Generative Adversarial Nets . Retrieved from arXiv1411.1784   cs.LG Martin Arjovsky Soumith Chintala Lon Bottou 2017 January 26. Wasserstein GAN . Retrieved from arXiv1701.07875  stat.ML Ishaan Gulrajani Faruk Ahmed Martin Arjovsky Vincent Dumoulin Aaron Courville 2017 March 31. Improved Training of Wasserstein GANs . Retrieved from arXiv1704.00028   cs.LG Lei Xu Maria Skoularidou Alfredo CuestaInfante Kalyan Veeramachaneni 2019 July 1. Modelling Tabular data using Conditional GAN . Retrieved from arXiv1907.00503   cs.LG Code Repositories SDV CTGAN httpsgithub.comsdvdevCTGAN Generation and Evaluation of Synthetic Tabular Data using GANs OntheGenerationandEvaluationofSyntheticTabularDatausingGANs Generating Tabular Synthetic Data using GANs httpswww.maskaravivek.compostgansyntheticdatageneration Deep Convolutional Generative Adversarial Networks using Gradient Tape httpswww.tensorflow.orgtutorialsgenerativedcgan Improved WGAN Implementation httpsgithub.comkerasteamkerascontribblobmasterexamplesimproved_wgan.py Suggested Selflearning topics before GANs A neuron in a neural network is a basic unit of computation that receives inputs processes them and outputs a result. Neural Network Architecture Transformer encoders are responsible for encoding input sequences into hidden representations. Transformer Encoders MultiHead Attention is a mechanism that allows the model to attend to different aspects of the input simultaneously. MultiHead Attention Inner workings One Forward and Backward pass Fake Data Gz Label 1 Label 0 Discriminator Gx Gz W D DISCRIMINATOR TRAINING GENERATOR TRAINING Maximize log Dx log1 DGz H yperparameters and Evaluation metrics Scaling Scaling can be done to improve training time and can be used to apply limits on the range of the input data. MODEL TUNING Epochs and Learning Rate Training faster or training the model more can be achieved by tweaking epochs and learning rate. Number of Layers and Neurons For the model to learn the relationship between the columns better the layers and neurons can be tweaked. Activations The activation layer can be adjusted according to the scaling to limit values to ranges according to the desired output. EVALUATION METRICS KL Divergence It measures how one probability distribution diverges from a second probability distribution JS Distance Is a measure of similarity between two probability distributions bounded by 01. It is symmetric Cosine Similarity Quantifies the similarity between two or more vectors. Its the cosine of angle between the vectors. PCA Reduce the dimensionality of data to visually compare two datasets.

Automotive_Customer_360_POV (1).pptx

Fractal POV on Customer 360 in Automotive Industry Complexity and relevance of todays customer experience demand a clear understanding of all relevant touchpoints and all possible journeys The concept of fragmented phases in a traditional funnel logic to be replaced by an ongoing infinity loop through which manufacturers dealers and customers jointly drive and shape their experiences in a connected automotive ecosystem T he buying journey in the automotive industry is an iterative dynamic circular process ideally binding the customer to the brand Lack of guidance and expertise due to limitless unstructured information tend to overwhelm customers. Lack of personalization customers seek for optimal offerings for themselves given their unique circumstances Awareness Consideration Consultation Purchase Handover After Sales Two major pain points emerge throughout automotive customer journey lacking personalization and lacking guidance or expertise Lack of personalization Personal interactions between the salesperson and the customer strongly influence the choice of model and can even alter brand preference. Dealership visits in the contact phase are perceived as stressful customer experience may be impacted in a pushy environment Customers wish to be in a state of complete control and have an open discussion with dealers prior to signing the contract. Lack of personalization can lead to generic products being sold to customers which are eventually of no value to them Customers look for a clear and proactive communication from the dealer especially since a car purchase represents a major expense. Massive decrease in interaction and communication down to nearly zero contact can lead to severe customer dissatisfaction Poor follow up on the customers satisfaction with the purchase by displaying a lack of interest in identifying new needs or by failing to interact with the customer at further touchpoints lead to poor customer experience Lack of personalization tend to overwhelm customers with untargeted sales letters and invitations Build a comprehensive platform that potential customers can visit in order to obtain all necessary information in a onestop shop manner Considering existing sources on mobility behavior e.g. from other apps devices or sources and expose potential customers to the information content that is most relevant to them Complete customer profile by interlocking online and offline touchpoints can help lead customers seamlessly from online contacts to offline prospects Awareness Consideration Consultation Purchase Handover After Sales Product seamless chat support by immediately forwarding quotes and offers to customer devices Improve customers perceived efficacy and control by making contracts and legal documents available in an easily understandable manner in customers devices Drive personalization by offering needbased products or having settings adapted to personal information Target and provide customers with the relevant information at the right time via the right channel drive customers from seeing the brand on Instagram to the manufacturers website to confirming a local dealer appointment in minutes Build an app through creation of a singleaccess login that grants customers access to all brand channels and sites and provide them with immediate answers Enable transparent timely and proactive communications by helping customers track production status occasionally providing photos and a realtime estimate of the handover date Keep customers engaged by providing digital concierge on adjacent requirement like renting a garage spot or registering the license plate Align the handover event to customers personal preferences Create customer database and build a unified customer profile to enable sales associate greet them with their full profile and history at hand and provide a personalized and appreciative service experience. Help dealers get to know their customers better and help them proactively suggest matching products that aligns with the customers new lifestyle changes Product Type Passenger Car Truck Van Region Europe UK North America China APAC Customer Group Private Customer B2C Business Customer B2B Distribution Level Retail Wholesale Business Division New Cars Used Cars Financial and Mobility Services Service Parts Accessories Customer Journey Phase Prepurchase Purchase PostPurchase Potential Scope Introducing Genomics to build the customer 360 and drive personalization scale Consumer Insights Hub Consumer DNA Sales patterns Demographics profiling Consumer Next Best Action Product Offer Opportunity Browsing Drivers of behavior Decision explainability EMOTION LAYER Define enquiry Seek insights Validate insights Gamified Research Emotional appraisal ENTERPRISE CONSUMPTION Marketing platforms Sales platforms CRM platforms Dashboards and Apps Selflearning AI models  Explainability of drivers EMOTIONAL ASPECTS KNOWN UNKNOWNS Consumer or customer goals motivations preference sets SENSORIZE Consumer response to communication as a feedback loop AUTOMOTIVE ORG DATA KNOWN Demographics Transactions Ridesharing apps Word of mouth Magazines Advertising Automobile shows Contract signing Financing Insurance Waiting period handover status information Service Repair Requests newsletters mailings INTELLIGENCE LAYER Leverage nonconscious design principles INTERVENTION LAYER PERSONALIZED ConsumerLEVEL RECOMMENDATION DESIGN EXECUTION LAYER Automated ETL Data integrity and quality checks Curated statistical features HARMONIZATION LAYER SOLUTION ARCHITECTURE Solution Highlights 2021 Fractal Analytics Inc. All rights reserved Confidential Alwayson automated and explainable client decisioning LEARN Responsible AI toolkit Customer Genomics accelerator modules can help accelerate business value across the data to decision spectrum Data Decision How a Dealer of an auto showroom would consume personalization insights through Customer Genomics Priority Ranking and Next Best Action Recommendations Deepdive insights Deepdive insights behavioral markers Illustrative Example AWS Step Functions workflow Storage Ingest BATCH Processing Governance AWS Identity and Access Management IAM We are live on AWS marketplace and can deploy using AWS native components Appendix Channels of contact DRIVING SUCCESSFUL CROSSSELL THROUGH CG Drive sense of urgency NCD recommendation Customer 360 When a service request has had positive resolution would be the best moment to spontaneously recommend. z Highly likely to buy Pet insurance Inapp popup Channel When exploring Moment Intelligence Layer Sarah Joe An email with a comparison summary or checklist and key policy highlights would be the best way to get her onboard. Creating a sense of urgency with a timebound discounted offer for the policy would help him make the decision. Felix Intervention Layer Prediction Drive sense of economic utility NCD recommendation Customer 360 Highly likely to buy Auto insurance Email Channel Late evening Moment Intelligence Layer Intervention Layer Prediction Drive sense of control NCD recommendation Customer 360 Likely to buy HomeRenters insurance Call Channel Servicing session Moment Intelligence Layer Intervention Layer Age 45 CLV Low Recently updated home address Prediction Emotion Segment Distrustful High of call center operational queries no product addons Employed Browsing peak 8pm10pm Recently applied for an auto loan Emotion Segment Pragmatic Initiates enrollment into new productservice but seldom completes the journey Location Rural Digital engagement High Recent transactions in pet stores vet clinics Emotion Segment Ambiguous Long frequent selfinitiated inapp info browsing no action 2021 Fractal Analytics Inc. All rights reserved Confidential

AWS Manufacturing Fractal Capabilities_20230124.pptx

AWS Fractal Assetheavy industry workshop Jan 2023 Data to Decisions scale Building Connected Data Computing on the edge running enterprise ops insights on the go self service BI capability Building Manufacturing nerve center Overview of solutions and offerings across the value chain Manufacturing nerve centre Cloudbased cognitive control tower to assist in strategic and operational decisions Remote monitoring and providing feedback for actionability Parts management Visibility across parts availability to manage daily operations Predicting parts requirements  shortages to manage production maintenance schedules Asset utilization Determining long term capacity requirements to assist in Capex decisions Maximizing asset utilization for tactical operations by optimizing the schedules ESG consulting analytics ESG process consulting for assessing clients technology and process maturity Holistic ESG data model with prototypes for visibility across all the ESG KPIs Top focus solutions for todays discussion in the assetheavy industry 01 Manufacturing nerve center Fractals vision of manufacturing nerve centre Diagnose Predict Digital enablement Control towers visibility to transform offline siloed decision making to online connected decision making Cross functional Insights and Early warning predictive alerts recommendations Self serve diagnostics with automated RCA Early warning nudges to proactively manage deviations Automate Optimize Diagnostic capability to deep dive into the production metrics maintenance quality issues machine metrics etc. Prediction assessment to drive value for shop floor operators and plant managers Identify the bottlenecks for future n days Alternative option evaluation with financial service lead times Scenario planning Intelligent decision support system supported by robust cloud platform Smart automation for translating decisions to actions on underlying systems Automating work flows for actions on the shop floor Optimization of processes considering the real time plant constraints Manufacturing nerve center Enabling cognitive and autonomous decision making for synchronized operations ITOT Integration Manufacturing control tower case study CS1 We delivered 30 dashboards across multiple workstreams at various granularity levels Region BU country site levels Monthly quarterly yearly Category Brand Brand segments Executive insights across supply chain operations cost KPIs through supply chain control tower Manufacturing control tower Illustrative 1 Manufacturing control tower Illustrative 2 Real time insights for operators Shift summary for plant managers Manufacturing control tower Illustrative 3 Warehouse control tower CS2 Mission Control in a 3 phase approach Outcome Optimized action recommendation ML based learning system to plan operational activities basis real time delays AIML Interventions Create models to predict Target end times Pick Pack Stage OT Delivery risks Automated recording and rule creation Outcome Integrated realtime insights Decision cockpits for all Warehouse functions Enable MisCon pilot to undertake all operational decisions AIML Interventions Data ingestion harmonization Standardized KPIs Design intuitive and actionable UI 1 Outcome Automated alerts to MisCon Pilot through notification center Trigger actions from MisCon Auto execution of WMS MHE transactions through XMLs and screen automation AIML Interventions Smart automation through XML and screen automation all the operational interventions Automated triggers or alerts 2 3 Mission Control Mission Control Mission Control Pilot Semi auto Auto pilot 02 Parts management Parts Forecasting Parts segmentation to determine the critical vs regular parts Accurate prediction for uninterrupted production using ML algos Leveraging machine parameters to sense exact requirements Parts Optimization Determining minmax quantities of parts across nodes Recommending the safety cycle stocks based on the demandsupply volatility Visibility to planners on availability and shortages Predict shortages Part consolidation based on demand aggregation Predicting the parts availability based on lead time contracts volatility POs etc. Alerts warnings to better plan production Optimized production plan Providing timely signals to planners for Clear to build scenarios Optimizing the production plan and using flexibility to rejig schedules Determining impact on downstream systems Ensuring the right parts at the right moment for uninterrupted production Parts forecasting inventory optimization CS1 Automated e2e forecasting inventory optimization for spares Forecast the consumption of Spare parts required to run operations Optimize Safety Stock for the parts without compromising on Service levels criticality e2e automated integrated solution AIML modelling to dynamically understand temporal behavior of demand and consumption Insights and recommendations enabling actions on various inventory and stocking parameters Automated planning process 2Mn savings across 30 of spare parts accessories globally Huge potential savings with scale Client ask Solution Result AIML driven methodology eventually leading to a fully automated procurement system Differential evolution optimization Performing various statistical analysis and hypothesis Differential evolution optimization for SKU Positioning into quadrants to understand Past behaviors and global behaviors Decide modelling technique for Forecasting demand Replenishment parameters at SKU level Min Max ROP safety stock EOQ Dynamic recommendations on minmax inventory norms Estimating and highlighting emerging trends and disconnects. Simulation of inventory norms w.r.t service levels supply demand parameters DL CNN LSTM ML Random forest Statistical A SRIMAX GARCH Parts shortage clear to build CTB CS2 Improved OTIF through predicting Product Shortage and integrated insights on Clear to Build Improve customer service level OTIF across network Proactively identify material shortages and enable early warning alerts by predicting material supply volatility Project material shortages to arrive at a potential Clear To Build status of their finished goods Estimating Receipt variance confirmed vs receipt at different days lag Delays in confirmed orders date by suppliermaterial Probability of variance at MaterialSupplier fulfillment pattern other Supply disruptions Project material shortages to arrive at potential Clear To Build at Finished goods Model to estimate upcoming receipt quantities for all confirmed purchase orders Simulator for Supply chain planners to capture Clear to build status mitigate shortages by supplier connect additional stock or stock transfer Client ask Solution Result High level solution approach for resilient material supply Integrate with CTB Project the product shortages to arrive at potential CTB at Finished goods Recommend best potential fulfilment date Recommendations on optimizing the components usage to maximize CTB of finished goods Predict Shortages Predict upcoming receipts quantity and ETA based on volatility in receipt quantity PO delays demand surge etc. Predict future shortages of parts Determine automated root causes for shortages Potential Fill rates Predict potential fill rates through Supply reliability Demand Mitigate the fill rate risks through optimal part allocation to CTB Capacity utilization planning 03 Long Term Capacity Optimization Determine the unconstrained demandbased capacity requirements across years Optimize longterm capacity plan to determine constrained demand which can be sufficed Identify the CAPEX requirement considering the CD ratio Scenario Planning Visualize opportunities and problems in network for overutilizedunderutilized capacities over years for simulation Integrate the optimization model to iterate based on userdefined constraints and inputs Decision Enablers Recommendations to mitigate capacity gaps for respective years Decision enablers to determine magnitude and timing of the excess capacity including CAPEX requirement in each year Capacity Utilization Impact in capacity production and demand gap w.r.t constraints input parameters Process reliability and schedule utilization achieved and RCA for improvement opportunity Capacity utilization planning approach Capacity Planning Capacity utilization 5Mn per line savings by recommending right changeover time based on demand Ideal changeover between presses to accommodate new demand O ptimize utilization based on mould to press compatibility Maintain the right CD ratio considering the outlook utilization metrics Identifying ideal CD ratio looking at CFR investment and asset utilization metrics Optimizing the ideal resources required with number of change overs Real time run time identification of next system based on Idle time and availability Long term simulated outcomes with optimum CD with lowest total cost There exist an optimal combination of Capacity and Inventory given a certain CFR CD ratio of 1.25 came out to be optimal ratio given all constraints Client ask Solution Result Manufacturing sub system Molds unit subsystem Molds availability Molds compatibility Press unit subsystem 6 cavity presses vs 12 cavity presses Our approach towards solving through optimization Maximum utilization of resources by risk pooling across pooling demand Reducing machine changeover times and demand supply gap to optimize schedules Generate optimized packaging schedules considering available capacity constraints Reduce the demandsupply gap while minimizing packsize change overs make time Provide offshelf site specific solution for validation and assess the change Optimal packaging schedule through linear programming Automated userdriven optimization with inputs Risk highlights for potential over underutilization of machines Optimized PPDS with following results Achieved avg demand supply gap 2.5 vs Current state 8 gap Reduced average changeover time by 5 Efficient short term plan with visibility into rolling 2 weeks Client ask Solution Result Oven L5 Machine 1 Machine 2 Machine 3 Output 4107 kg hr for all flavors One Flavor at a time Demand at SKU and Week level Machine 1 97g116g Machine 2 97g116g 194g233g Machine 3 194g233g 388g466g Machine 4 Portion Bulk i.e. 3kg Each machine Portion different capacity of production per hrs. Flavor change timecost is varying from 1 to 2 hrs depends Pack size change over is 1 hrs change over Whenever there is flavor and package change over comes together addup the both time Stop production for module cleaning 30 min. every 48 hours i.e. every alternate day of flavor 9 40 SKU Flavor Pack size Brand Production runs in 3 shift 24 7 Always should be on Weekly Demand of SKU vs Actual Production should be in 10 range Portion always being used incase when it m1 m2 m3 are not able to compensate op pf oven to fullfill the demand of SKUs is related to Pack size Bulk OP Production Plan for rolling 2 week at SKU level with details about SKU Machine Start time end time flavor change over Pack size setup time etc. details Diversion of Flow to portion incase Machine capacity incase not able to compensate Only one Portion can be active at time Machine 4 Solution schematics objective inputs constraints output Building capacity asset utilization Assist production managers analyze the asset utilization of machines across the plant Help organizations unleash capacity opportunities in their manufacturing lines 04 ESG consulting analytics For manufacturers to embark on ESG journey they often come across a multitude of challenges Energy Waste Water Management Energy Transition Emissions Management Data Security and Risk Management Occupational Heath Safety Lack of a centralization Lack of a uniform structure of info across cotton farms No track record of labor rights irregular pays . No opportunity to create crossfunctional insights and learnings High Manual Intervention High Manual intervention for KPI reporting High probability of errors and inaccuracies Due to high manual intervention more time and effort required Lack of endtoend visibility Absence of frameworks that can co relate different ESG pillars to each other No sense of status at desired time and frequency Data Collection Data collection for ESG from multiple departments and vendor. For traceability we use verified sources of data e.g. SAPPM FACTSET CSRHUB OAG supply chain systems POS Data Quality Data available at different granularities and frequencies across different stages of merchandise procurement. E.g. Quality of data from 3PL logistics not consistent thus we perform data cleaning and harmonization. Lack of standardization Lack of standard definitions of KPIs across different regions categories Inconsistent terminology Issues with measurement units of measure Data Challenges Priorities Accelerate ESG journey through a centralized data platform across enterprise Integrated data model Leading sustainability indicators Algorithmic decisions Decision cockpit Manufacturing SAPPM MES Enablon Human Rights Sedex SMETA Audit files OEM sourcing Logistics TMS SAPBW Mapping files Sourcing Supplier files Packaging SAP ECC PLM Mapping files Corporate Giving and DI inclusivity flat files Data harmonization KPI definition Source to target mapping Data model development Enterprise level automated audits Governance Supply chain vendor transparency Data security risk management Environment Energy Waste Intensity Water stress Manufacturing emissions. Fugitive Leakage Emissions Social Sustainable Sourcing Ethical supply chain transparency Supplier compliance Occupational health safety Waste Driver analysis Carbon Footprint prediction Simulation sandbox Safety incidents prediction ESG decree compliance Sustainability target glidepath Sustainable procurement Invest in right green projects Built an e2e data platform for ESG reporting and democratizing data for CPG manufacturing majors Integrated data platform to Automate Standardize Collaborate and Harmonize data across all functions for e2e ESG visibility on SAP BW platform 40 ESG KPIs across functions for periodic tracking and auditing Reduced TAT by 88 from 68 months to 1 month through automated standardized reporting Client ask Solution Result To build e2e data platform for ESG reporting and Democratizing data Automate data collection and minimize manual intervention Ability to align with key external frameworks on ESG This data model represents the logical structure of how ESG data is captured for different verticals within an organization and how the relationships are mapped between them. Example Below data model represents Water and Energy Verticals Holistic ESG d ata model to drive vertical startup on analytics journey ESG solution snapshot CS3 Inventory Optimization Monitor inventory across network Visualize current inventory across SKUs brands categories inventory segments regions DCs Monitor KPIs Current Minmax Days forward coverage stock OH overstock outofstock inventory turns etc. Identify opportunity areas for improvement across SKUs DCs segments Predict inventory risks OOS risk identification at SKUDC level and impact assessment on cost and CFR Prediction of Inventory waste due to excess stock assessment on cost Codify business rules thresholds for alertsrisks Highlight inventory risks due to product obsolescence SLOB inventory Determine inventory minmax norms Generate optimal recommendations for minmax stock at each granularity Generate stock projections into 12 weeks horizon Trigger automated alerts for current week future 2 weeks horizon MEIO optimization simulation Simulation of inventory requirements basis CFR demand supply lead time etc. Incorporating constraints like MOQ Customer Service level DC Capacity Holding cost to optimize MEI Recommendations on inventory allocations based on Case fill Promotion demand factors AI enabled approach for e2e inventory optimization 1Bn savings through e2e inventory optimization Reduce NonPerforming Inventory NPI across 13 categories for both Finished Goods and Raw Materials Establish business review process to track actions Optimize SKU portfolio and inventory Consultantled monthly inventory review process Key insights on NPI trends root causes recommendations on depleting inventory Action monitoring system to track Inventory reduction 1Bn savings across 21000 Finished product and 5300 raw packing product code over 4 years Client ask Solution Result Optimized e2e inventory enabling business decisions through AED approach Root causes and recommendations Automated root causes for NPI at Category Plant and SKU level Recommendations to enable MRP Category planners on MOQs Production orders Sourcing plan Purchase orders Inventory classification Blocked Discontinued Obsolete and Excess classification Enabled business to address what where why and how of Nonperforming inventory Business review sessions to tackle top contributors Potential Phase out SKUs Predict patterns of declining SKUs for upcoming quarters their product life cycle transitions Guidance on revising production parameters sourcing decisions for potential phase out SKU Optimized Inventory norms across the network nodes through AIML interventions Improve customer service levels Reduce high on hand inventory levels Static Inventory targets thresholds Optimize Inventory norms safety stocks at every Supply network node Predict Safety stocks based on demand and supply volatilities Optimize for Overage and Shortage at SKU Supply network node level Dynamic determination of Inventory norms process automation 40 reduction in Inventory norms and safety stocks with improved customer service levels Client ask Solution Result Dynamic Safety Stock through predictive algorithms Safety stock prediction Outcome Recommended safety stock Dynamic calculation to enable early interventions Scenario planning Driver analysis for all components of supply and demand Supply disruptions Demand fluctuation Supplier reliability Fulfilment rate Quality checks Leas time variance Supplier capacity Actual sales Demand forecast Promo events Returns exchange Similar products Calculate historical supply related KPI at required time granularity For demand factors calculate risk profiles to capture variance Leverage modelling techniques like Random forest SVM NN to predict safety stock

AWS Manufacturing Fractal Capabilities_Consolidated.pptx

AWS Fractal Manufacturing use cases Oct 2022 Index Predictive maintenance 01 Predictive maintenance for varied time to failures Unsupervised anomaly detection failure prediction for Industrial printers Unscheduled printer component failures resulting loss of printing time ink revenue high operational expenditure Predict anomalies failures for proactive scheduled maintenance Identified features impacting printer performance Unsupervised anomaly detection model using adaptive distribution at dailycomponentprinter level and survival models to predict time to failure 662 anomalies for 26 printers for the entire year Accuracy measures 67 on Ink consumption as of print area 72 on aborted jobtotal Jobs Client ask Solution Result Unsupervised anomaly detection to predict the failures Key Features indicating printer performance Enabled packaging line with real time anomaly detection Production line with multiple machines robots. OEMs equipped to provide static thresholds alerts for individual critical parameters Better understanding of machines swift decision making Integrated shop floor with data streaming Developed various analytical insights to enable smarter machines Adaptive algorithms on data from 21 sensors Real time anomaly detection for course correction at minute time granularity Client ask Solution Result Real time anomaly detection through adaptive algorithms from 21 sensors Reduced 57 of unplanned downtimes by continuous monitoring of process parameters Develop predictive maintenance algorithms based on the historical manufacturing data Failure prediction with 37 days in advance Power BI based UI with machine level predictions and early warnings Unsupervised anomaly detection through Gaussian classification failure prediction model Classification failures by creating lag variables Heuristics rule to reduce false positives RegressionTime series forecasting failures Machine level downtime risk prediction Reduced unplanned downtimes by 57 compared to previous qtr. Client ask Solution Result Increasing reliability of assets with Prescriptive analytics Early Anomaly Detection Actionable Insights The model Unsupervised anomaly detection for gearbox using sensor data Develop algorithmic workflow to enable automatic detection of anomalies utilizing gearbox sensor data Data streaming from various sensors for identifying anomalies in real time and predicting failures Anomaly detection model using adaptive machine learning technique capable of predicting from live streaming sensor data Support Vector Regression technique for model development Identified 32 days in 9 months with spectral data was observed to contain an anomaly Maximum anomalies observed in September for consistently longer period 17 days indicating high likelihood of a breakdown post this period Client ask Solution Result Reduced batch cycle time through discrete event simulation To simulate for most optimal batch cycle times with achievable combinations of parameter metrics schedules Simulator with statistically viable and robust solution to ensure that historical anomalies leading to longer Batch Cycle Times can be avoided in future Reduced number of Change overs across manufacturing lines Outcomes Multiscenario matrix of CPPs using Monte Carlo Simulation Genetic algorithm Simulation to find best possible combination Simulate CPPs and assess the impact on BCT Comparison with BIC batches Optimize BCT Monte Carlo simulations of risks for the optimal BCT relevant CPPs Genetic algorithm for integrated Manufacturing Scheduling Identify best possible combinations of CPPs to attain optimum BCT Simulate CPPs using conditional probabilities from BNN modeling Shop floor management Product Quality 02 3Mn reduction by avoiding product off spec Identify the root cause leading to issues with firmness of product Provide recommendation on how to control the various controllable factors to produce products as per firmness specification Critical process parameters for Golden batches Developed supervised learning drivers models to identify the significance of critical tags impacting product firmness Enabled SOP enhancements alerts for course correction 3Mn reduction by avoiding product off spec reducing rework product waste Improved Customer satisfaction Client ask Solution Result Business problem Nearly 50 of quality test samples of a dairy product whipping agent had lower firmness value than specification threshold Resulting in rejection of production batches Annual loss of 22MN Euros Customer complaints High amount of rework Samples having firmness values below threshold 70 Objective Identify the root cause leading to issues with firmness of product Provide recommendation on how to control the various controllable factors to produce products as per firmness specification A European dairy cooperative wanted to identify root causes for a product firmness issue monitor the quality in real time Sensor tags used in the dairy process manufacturing not exhaustive Collection consolidation of data Data Consolidation of all batches with corresponding quality samples Mapping batch sample output with Sensor tag values Feature identification Trend analysis of firmness of samples within a batchbetween batches Univariate bivariate analysis between 63 Tags Outlier identification Driver analysis Identified important Tags which influence firmness Built and evaluated competing supervised learning models LIKE decision trees for finding drivers Real time monitoring Directional relationship between important tags and Firmness Optimal ranges for controllable tags to achieve firmness level 70 Sensors like Temperature nozzle pressure valve pressure etc. Control charts trend charts Predicting product firmness based on real time monitoring of critical process parameters Clustering Decision Tree models Recommended threshold values of sensor tags Identified critical manufacturing process parameters leading to a Golden Batch Identified important tags that drive the firmness of the product during process manufacturing. E.g. Plenum Temperature Nozzle pressure HD Flow Pump Inlet Temperature etc. Line operators can focus on the controllable tags in the recommended operational ranges for better firmness of product Revised the SOP ranges of various sensory tags through the decision tree approach Useful for business to get sense of which valuescombination of parameters will lead to a Golden Batch Would help line operators to identify thresholds Workload improvement for operator and analysis Recommendations for NPD process Decision Tree to arrive at tag thresholds Solution Capabilities Key results Clustering for identification of important tags Variable importance from model output Created saving opportunities of 400K Euros per product SCADA MES Gather EnrichStream Sensors Actuators IOT Gateway HTTP WiFi serial ports MQTT Event processing Storage Datalakes Hadoop clusters Realtime message brokering kafka ActiveMQ Complex event processing event pattern detection filtering Device registry AIML Models CNN LSTM Decision trees App backend HTTP Consumption Visualization Portal HTML5 Leveraging IIOT on a large scale for real time monitoring Process control chart for Nozzle Pressure Revised SOP range for Golden batch Nozzle Pressure 231 Consumption layer Identification of CPP Plenum Temperature 0 30 C Temperature 40 Flow HD Pump 5500 Flow 5000 Nozzle pressure 260 Pressure 231 Current status of sensor tags Predicted value of Product Firmness 68.5 Plenum Temperature 3 110 C Temperature 90 DP Plenum 1 70 C Temperature 40 Plenum Temperature 1 100 C Temperature 92 Plenum Temperature 2 105 C Temperature 90 DP Plenum 2 49 C Temperature 50 DP Plenum 3 59 C Temperature 60 Consumption layer Real time monitoring of sensor tags course correction MRO Spares 02 Automated e2e forecasting inventory optimization for spare parts Forecast the consumption of Spare parts required to run operations Optimize Safety Stock for the parts without compromising on Service levels criticality e2e automated integrated solution AIML modelling to dynamically understand temporal behavior of demand and consumption Insights and recommendations enabling actions on various inventory and stocking parameters Automated planning process 2Mn savings across 30 of spare parts accessories globally Huge potential savings with scale Client ask Solution Result ESG Capability 02 Lack of standardization in data collection Limited data available for Private companies Low correlation in ESG scores by leading agencies Weightage given to ESG metrics differ by region Lack of transparency in underlying factors driving the ESG score The lack of a single regulatory framework The lack of consensus on terminology and definitions The lack of auditing which limits the reliability of disclosure reports. There is no universal system to verify reported data and public reports tend to highlight positive contributions to ESG and underplay or omit less flattering reporting this is called greenwashing . The fact that only larger public companies are required to report ESG data which is published in their annual reports. This limits the data pool to what these companies selfdisclose. There is no regulation governing the assessment of sustainability requirements and performance and every ESG rating agency has its own methodologywhich evolves over time. Copyright 2023 fractal. All rights reserved. Challenges commonly faced by organizations during data gathering for ESG metrics Need to update as per heavy asset Analyze data relating to traffic patterns realtime demand and networkresource availability It allows for quick automated decisions on the parts of the system that can be put into sleep mode or shut down Shutting down frequency carriers or shutting down a site momentarily in areas where there is overlapping coverage Smart shutdowns IOT sensors can gauge gridpower input fuel levels the number of hours the generation set has been running Battery voltages and consumption by different types of equipment Analyzing the data operators could uncover potentially costly anomalies Real time alerts. IOT enabled optimization In cement plants the AI realtime optimizer RTO adapts to continually improve optimization outcomes. AI can help optimize energy consumption and throughput of cement kilns and mills by improving equipment productivity in view of rising environmental concerns and CO2 costs. AI powered optimization Identifying Leak detection and repair systems at compression stations. Using analytics to predict the failure of the equipment at compression stations and pipelines. Predictive analytics Purchase or generate green energy Direct procurement and competitive sourcing. Sustainable sourcing Energy savings levers for Heavy Industries Heavy Industries like Cement Oil Gas Utilities metals mining can leverage upon multiple areas like AI infra site optimization sustainable procurement etc. Need to update as per heavy asset Embedding Analytics Capabilities into Sustainability building blocks Need to update as per heavy asset VISIBILITY OF SCOPE 1 2 AND 3 EMISSIONS ACROSS THE VALUE CHAIN IS KEY TO JUMPSTART THE JOURNEY Scope 1 emissions are direct emissions from owned or controlled sources. Scope 2 emissions are indirect emissions from the generation of purchased energy. Scope 3 emissions are all indirect emissions not included in scope 2 that occur in the value chain of the reporting company including both upstream and downstream emissions. 1 2 3 EP Emissions from drilling well testing Operational spills Methane leaks and gas flaring at the wellhead 1 Shipping Transportation Emissions from fuel consumed by company owned vehicles such as trucks ships helicopters Processing Combustion of natural gas and other fuels in refineries plants or other facilities Chemical emissions from plants 1 1 Operating IT emissions on premise data centre 1 Purchase of electricity and heat in every company owned facility Plants mining sites Operational facilities Data centre operations 2 Employee Commuting Daily commute to the office Employee air and rail travel car rentals business travel Offshore site visits 3 Embedded emissions In buildings equipment In leased buildings vehicles assets 3 Transportation and distribution In the entire value chain from maritime shipping to the retail dealers if utilizing vendors rather than company owned vehicles 3 Other operations Franchises Paper consumption Endoflife disposal of unsold products Scrap from equipment or waste generated in the production facilities 3 Usage of products from end consumers 3 Copyright 2023 fractal. All rights reserved. Objectives Challenges Solution Approach Objectives To build e2e data platform for ESG reporting and Democratizing data Automate data collection and minimize manual intervention Ability to align with key external frameworks on ESG Business challenges Multiple data sources Regular Manual intervention Nonstandardized process and data flow Irregular data granularity Lack of Data Harmonization Integrated data platform to Automate Standardize Collaborate and Harmonize data across all functions for e2e ESG visibility on SAP BW platform 40 ESG KPIs across functions for periodic tracking and auditing KPIs and metrics to track Energy and Water consumption CO2 Emissions Benchmark emission across SC nodes Logistics Impact on emissions and longterm roadmap and projections Packaging Recyclability index Building a e2e data platform for ESG reporting and democratizing data A global database providing Standard Reports Benchmarking Baselines Single point of truth for nonfinancial performances Data validation and rectification at the source Outcomes impact Capacity utilization planning 03 Long Term Capacity Optimization Determine the unconstrained demandbased capacity requirements at SKUline across years Optimize longterm capacity plan at SKUplantline level to determine constrained demand which can be sufficed Scenario Planning Visualize opportunities and problems in network for overutilizedunderutilized capacities over years for simulation Integrate the optimization model to iterate based on userdefined constraints and inputs Decision Enablers Recommendations to mitigate capacity gaps for respective years Decision enablers to determine magnitude and timing of the excess capacity including CAPEX requirement in each year Capacity Utilization Impact in capacity production and demand gap w.r.t constraints input parameters Process reliability and schedule utilization achieved and RCA for improvement opportunity Capacity Utilization Planning approach Capacity Planning Capacity utilization 5Mn per line savings by recommending right changeover time based on demand for CPG giant Ideal changeover between presses to accommodate new demand O ptimize utilization based on mould to press compatibility Maintain the right CD ratio considering the outlook utilization metrics Identifying ideal CD ratio looking at CFR investment and asset utilization metrics Optimizing the ideal resources required with number of change overs Real time run time identification of next system based on Idle time and availability Long term simulated outcomes with optimum CD with lowest total cost There exist an optimal combination of Capacity and Inventory given a certain CFR CD ratio of 1.25 came out to be optimal ratio given all constraints Client ask Solution Result Manufacturing sub system Molds unit subsystem Molds availability Molds compatibility Press unit subsystem 6 cavity presses vs 12 cavity presses Our approach towards solving through optimization Maximum utilization of resources by risk pooling across pooling demand Improved packaging efficiency reducing both machine changeover times and demand supply gap Generate optimized packaging schedules considering available capacity constraints Reduce the demandsupply gap while minimizing packsize change overs make time Provide offshelf site specific solution for validation and assess the change Optimal packaging schedule through linear programming Automated userdriven optimization with inputs Risk highlights for potential over under utilization of machines Optimized PPDS with following results Achieved avg demand supply gap 2.5 vs Current state 8 gap Reduced average changeover time by 5 Efficient short term plan with visibility into rolling 2 weeks Client ask Solution Result 1 production line 5 Packaging lines 40 SKUs 9 flavors 7 Pack sizes 10 Constraints Oven L5 Machine 1 Machine 2 Machine 3 Output 4107 kg hr for all flavors One Flavor at a time Demand at SKU and Week level Machine 1 97g116g Machine 2 97g116g 194g233g Machine 3 194g233g 388g466g Machine 4 Portion Bulk i.e. 3kg Each machine Portion different capacity of production per hrs. Flavor change timecost is varying from 1 to 2 hrs depends Pack size change over is 1 hrs change over Whenever there is flavor and package change over comes together addup the both time Stop production for module cleaning 30 min. every 48 hours i.e. every alternate day of flavor 9 40 SKU Flavor Pack size Brand Production runs in 3 shift 24 7 Always should be on Weekly Demand of SKU vs Actual Production should be in 10 range Portion always being used incase when it m1 m2 m3 are not able to compensate op pf oven to fullfill the demand of SKUs is related to Pack size Bulk OP Production Plan for rolling 2 week at SKU level with details about SKU Machine Start time end time flavor change over Pack size setup time etc. details Diversion of Flow to portion incase Machine capacity incase not able to compensate Only one Portion can be active at time Machine 4 PPDS POC solution schematics objective inputs constraints output Region Business Unit Year Brand Saved scenarios Capacity Optimization Filters AMEA INDIA 2022 Brand 1 Create new scenario None Region Business Unit Year Brand Saved scenarios Capacity Optimization Filters AMEA INDIA 2022 Brand 1 Create new scenario None Optimized Capacity Plan Production quantity cases Total production cost Publish Capacity Plan Go back Total line capacity cases Building Capacity Asset utilization solution Assist production managers analyze the asset utilization of machines across the plant Help organizations unleash capacity opportunities in their manufacturing lines SOP 04 Digital enablement through integrated SOP visibility SOP Control Tower Nudges Early warning indicators Course correction using thresholds Visibility Single source of truth for track and trace Realtime monitoring control Collaboration Internal as well as External collaboration Instant feedback mechanism Selfserve Diagnostics Driver analysis for RCA Business scenario planning Security Governance Role and functions based access Secured data maintenance Deep Dives Deep dives into Siloed functions Performance and variance analysis Decision Support Cognitive based decision making Intelligent recommender systems Integrated Platform UpstreamDownstream integration Robust system with automated data pipelines Transformed SOP process with Supply Chain Planning Control tower Single source of truth with right data at right time across 11 Sourcing platforms 25 Plants 13 DCs Enable collaboration across functions to drive Operational Strategic decisions Forward looking capabilities Integrated Supply chain data platform Cross functional user stories Demand at risk Early warnings alerts on upcoming risks Optimized planning scheduling asset utilization Faster decision making across Sales Operations planning Client ask Solution Result Driving efficiencies across functions through Control tower Demand analysis Real time collaboration on forecast changes multifunctional stakeholders Raw Material Risk Evaluator Raw material supply risk demand at risk mitigation Manufacturing Driving efficiency through connected insights asset utilization and plant MRO Demand Supply Reconciliation Optimize Allocation priorities with Supply constraints Inventory visibility Capacity analysis Capacity constraints gaps between CoMan Inhouse Mfg alternative capacity plans change overs PR Financial Implications Reconcile financials against AOP IBP demand call significant COPS deviations Quality holds Working capital release through visibility and driver analysis of blocked inventory Customer Order Descriptive analysis to track customer orders allocations and processes Freight Costs and Drivers Insights on freight costs drivers load utilization and freight spot rates Supplier cockpit Speed to decision making through visibility in supplier management Integrated Demand analysis providing e2e Visibility of historical Forecast vs Actuals vs Forward projections Demand Supply reconciliation for informed decisions based on actual historical performance Financial Implications of demand supply financial CCH Mfg 03 Business needs Executive insights across manufacturing sustainability Challenges 16 data sources for gathering data 60 time was spent on data gathering 30 FTEs involved in creating executive reports Limited trust in data as everyone had their own interpretation Multiple tools were needed to get e2e picture Need for flexible views Comparison of data from different perspectives View historic vs. future trends View data as per specific needs for a user country vs. function vs. business level Need for next best actions Identify insights easily Identify and mitigate risks Identify next possible actions Create a single source of truth which would provide the right data at the right time to become more strategy focused Visibility across functional lines for 360 supply chain assessment Visibility across costs and enable benchmarking across business units and trend spotting Highlights and benchmarking key executive manufacturing KPIs Highlights and benchmarking key executive sustainability KPIs Plant performance visibility 04 Connected insights across manufacturing How to reduce manufacturing costs at the plantline level while overcoming the biggest challenge of global benchmarking and segmenting the plantslines for comparison on factors such as automation of the plant category produced in the plant and size of the plant. Improve Manufacturing KPIs by drawing insights and identifying opportunities for Productivity improvements Adopt analytical rigor in arriving at the KPIs across various lines within and across plants Dynamic Benchmarking capabilities across Plants Lines and Tracking against Designed and Operational Targets Process Reliability Manufacturing Yield Potential output PlannedUnplanned Losses PR opportunity losses Quality Compliance MTBFMTTF 2020 Fractal Analytics Inc. All rights reserved Confidential Measuring KPIs across reliability engineering for e2e visibility Scheduled Utilization Line utilization factor availability Designed Capacity Machine utilization rate Time loss Manufacturing Spend People Costs Indirect Costs Planned Unplanned Costs Manufacturing capacity planning and scheduling CU and SU visibility Business question answered What is the system availability at a shiftline level What are the root cause for machine failure and losses due to the failure What is the average and benchmark Batch Cycle Time of batches made during the selected time period for the chosen manufacturing site What raw materials impact system performance significantly The percentage of Process reliability Capacity utilized and schedule utilization achieved and opportunity for improvement Opportunity identification to reduce various manufacturing costs based on 1 improvement in process reliability as well as weighted volume produced Capacity Utilization Scheduled Utilization and Process Reliability facts and opportunity calculations Process reliability Based on the targets manually entered in the configuration the opportunity scope of improvement for the respective KPI is identified for the selected RegionPlantLine How is my plant doing and where are the problems Line 21 Analysis Unplanned Downtime is high for the line Inprocess measures impacting UPDT Line 21 is green on all the Inprocess measures except QA Completion Cost of production per MSU and opportunity is identified The opportunities identified are in par with the actual capacity for the selected RegionPlantLine and the Global Targets specified by the end user Volume and cost opportunity Planned and unplanned costs and number of production lines per day Based on the type of loss to be assessed the opportunities for losses along with changeover loss and the number of unplanned stops made at any plant per day is calculated. The targets are given in the configuration section Plant cost analysis and dynamic cost calculation to improve by a PR point The PR is calculated dynamically in proportion to the volume produced at the line and the respective opportunity is calculated. Scenario comparison between the fixed cost to improve a PR point versus dynamicactual cost to improve a PR point per line Cost per line and the PR loss at Line level Costs Supply cockpit Use case Supply chain cockpit for Global manufacturing users Final Supply KPI reports 60 KPIs categorized as Great People Great Plants and Great Pipelines at multiple levelsRegion Market Factory etc. and multiple periods Automated e2e calculations of KPIs at different levels as Quality People Supply Excellence Safety etc. Data Harmonizing in data bricks and tables setup in lake Flexibility to add new data or KPIs to the source Live data setup completion in Azure and automated pipeline to refresh data Business Problem and Objective Data Related Requirement Global Supply team wanted visibility across needs to develop a cockpit to get e2e visibility across supply chain network The cockpit will give visibility across assets people plants and their productivity Create a single source of truth as there were different reporting structures globally with different calculations logics and granularity across businesses and region Enable Supply Chain teams to use data to carry out self serve analytics Aligning KPIs across people pipeline plants and finance Fixed Asset Turnover Ratio Fixed Asset Addition TRS Adherence to Schedule Energy Water Intensity Total incident rate Loss time incident rate Right at First Time Logistics Conversion Cost Prod Conversion Cost Case Fill Critical Role Convergence Grand Mean Current Asset Turnover Ratio Great People Financials Great Plants Great Pipelines Fill rate ontime Factory Service Level Case Fill Rate Trade Attitude Survey Metric Engagement Score 20K 80K 80K 60K 40K Daily Plant Summary Trend Analysis Daily Network Summary Alerts Illustrative Snapshot 13 Non Periodic KPIs Production Line Periodic KPIs Global KPI Region Segment Customer Service Trends Snapshot 23 Loss BridgeMultiple Periods Market Performance Factory Snapshot 33 Mfg 360 We delivered 30 dashboards across multiple workstreams at various granularity levels Region BU country site levels Monthly quarterly yearly Category Brand Brand segments Executive insights across supply chain operations cost KPIs Vendor Supplier OTIF Total manufacturing cost Service Cost Effectiveness Severity total Health Safety environment Right at first time Quality CAPEX Cash Capital effectiveness Compliance to Schedule Customer OTIF Gross OTIF SFA Bias Gross Profit Net Profit Conversion PL by plants MR Labour parts services costs Waste costs Raw materials FG Fixed costs Total Accident Rate TAR Total Incident Rate TIR CO2 emissions Total Mfg waste Water incoming RFT sell composite Food safety total CTQ Consumer Complaints DIOH Inventory costs Opex Days Sales Outstanding Day Payables Outstanding DPO  Cash Conversion Cycle  Enabling 360 o visibility across multiple KPIs Days Between Next Run DBNR  Visibility to Compliance to Schedule metric across Regions BUs business categories and trends Visibility to Total manufacturing costs across Regions Categories plants and brands Visibility to Total Manufacturing Costs comparison w.r.t Plants SKUs Brands Regions BUs Sourcing Enabled buying decisions through improved price forecasting of Vanilla bean prices Essence manufacturer forecasted vanilla prices for Madagascar using traditional and intuitionbased approaches Current forecast methods was more time consuming and lacked the external variables which can potentially impact the future prices Automated data ingestion to harmonize disparate data sources to create a single source of truth for forecasting vanilla prices Leveraged advanced forecasting algorithm and automate execution to improvise on vanilla prices Self serve diagnostics Dashboards enabled to understand the trend seasonality patterns and of the vanilla prices Scenario planning Enabled users to simulate and work out the best possible scenarios in cases of business uncertainties Client ask Solution Result Inbound Supplier risks mitigation Improved visibility predictability of risk across Supply Network nodes with external internal data Impact assessment of risk across the nodes Prescriptive recommendations on alternatives and tradeoffs for risk mitigation strategies and Scenario planning Collaborative platform for visibility Real Time Visibility Early warning signals on risks Recommendations to mitigate risks Identified 150Mn opportunities through Inbound supplier risk prediction and traceability Client ask Solution Result Result Demand pooling for Optimized purchase volume frequency Optimal ordering quantity with schedule to reduce overall cost and prevent obsolescence Alerts and triggers on products obsolescence and expiry Classify products on the basis of shelf life and demand to prioritize buying Early warning signals to enable decision making at store level Recommendations on purchase frequency and volume at SKUstore level Annual savings of 1Mn per SKU group Scaled analytics product across categories Client ask Solution Result Total Landed costs for imports across regions with internal external data Explore ways to determine total landed costs from publicly available information Assess total cost to serve components Procurement Carrier SH and Duties with many subcomponents HTC codes across 3 Merchandising categories Driver analysis for total landed cost Modeled the Total Landed Costs through linear calculations Simulation to change routevendors to see probable cost impact 99 accuracy on category for Cotton Balls Absolute cost difference of 0.003 per unit was observed Client ask Solution Optimized Logistics Costs with e2e visibility cost drivers through AIML interventions Standardize definition of Logistics cost performance KPIs e2e visibility of cost components Standardize ways to attribute costs Forward looking view of costs rather retrospective view of costs Integrated Logistics data platform with reduced latency in data to decision making from 1QTR to 1 Week e2e cost performance management with descriptive predictive prescriptive analytics Drivers mix with customer level BBN models 25M reduction in Logistics costs through Transportation cost drivers improving routing guide improved dynamic spot price Client ask Solution Result Smart machines A manufacturing firm wanted to detect the anomalies occurring in the machines proactively and thus prevent the operational failures Business Problem Business Objective Analytics enabling smarter machines to help determine the anomalies real time Multi variate clusters on critical process parameters Determine thresholds dynamically Identify anomalies in real time Alert the machine operator Adaptive learning taking operators feedback Live Data Streaming Real time Analytics 1 2 3 Various data management and analytical components rendered to detect anomalies Multiple Signals Processing Real Time Processing Adaptive Algorithms Ability to evaluate multiple signals together Ability to process realtime Algorithms that adapt with live data Real time monitoring and clustering for anomaly detection Machine Learning Principal Component Analysis K means Clustering BigData technology Adaptive Algorithms Online Kmeans AutoEncoders Real time data from sensors were classified into clusters every 110 th of second to identify the potential machine with anomaly Data for 21 sensors of a machine measuring Temperature Pressure Oil Flow rate Water flow rate Etc... A reading is collected every 110 th of a second Real time streaming and clustering engine Output One output every 110 th of second 600 instances of data every minute 12000 data points every minutes UI fetc

es a instance of data every 110 th of second Real time streaming and plotting of clusters Plotting anomalies on bubble plots Taking user feedback for adaptive learning Solution helped business take appropriate corrective action against the machine and thus saving further operational time

CPG Future_ Fractal BCM Solution.pptx

BCM as a transformative solution A GoToMarket Shareout Business Continuity Planning for Supply Chain Resilience October 2022 Flow What is BCM in this context raw materials supplier risk 18 TH Oct22 Why is it so important today secondary research facts challenges post pandemic 18 TH Oct22 What kind of problems can we help you solve impact 19 TH Oct22 each case What kind Identification of raw material criticality View into supplier position risk qualification Monitoring and assessing Risk in BCP Recommending alternative materials suppliers Process eng e2e digitization DT of data systems business rules logic KPI Demo NO ACTION Pilot 6 to 8 weeks Shreya Who can we take this to PS BU COE Competition existing ERP solutions research Case studies Flow What is BCM in this context Why is it so important today  Disruptions not seen as a one off event Greater supply chain visibility efficiency and resilience are top of mind. changes in commodity prices supply chain bottlenecks continuing need to assess implications of choosing alternatives cost monitoring What kind of problems can we help you solve impact Identification of raw material criticality  View into supplier position risk qualification  Monitoring and assessing Risk in BCP   Recommending alternative materials suppliers  Process eng e2e digitization DT Demo Who can we take this to PS BU COE  Execution approach  Competition existing ERP solutions research Case studies  BCM Define criticality Assess supplier risk Assess BCP strength Identify subs materials Identify subs materials suppliers Business continuity management is critical for procurement and RD teams 01 02 05 03 04 A business users views and concerns pertaining to BCM planning execution Manufacturing RPM Material Planner Manufacturing RD leader Material quality Lead Procurement manager RegionalGlobal Procurement Head We dont adopt a projectionbased inventory monitoring Sourcing Spends are increasing post COVID Supplier OTIF within top 10 critical suppliers are dropping We focus on reacting but not anticipating There is no S2P data KPI visibility RM Quality notification trends have increased by 20 We bring bespoke services and accelerators to enable BCP risk identification mitigation Planning monitoring BCP strategy planning Supplier Material performance Design thinking approach Nireeha to add Enhanced E2E visibility to smoothen collaboration across nodes and post event analysis Risk prediction Alternative strategies Supplier material risks Criticality scoring clustering Single source vs multisource Alternative material suggestion optimization Alternate sourcing location supplier strategies Cognitive recommendations Cognitive automation Whatif simulations Smart automation with existing systems and landscape Cloud based planning capability to assess impact of various parameters and tradeoffs among scenarios How Might We help build a Business Continuity Plan and design a solution that _______ PRE DISCOVERY DISCOVERY SYNTHESIS IDEATION Understand current process of BCM Validate the current process Understand enduser painpoints gaps and needs through user focused research methods DESIGN FOR PILOT Define needs and opportunity areas and ideate on possible interventions Mapping painpoints and needs for ideation Design the solution as POC Pilot Give a sneak peak of the solutions Solution Flow Wireframes UI Screens Using Using Using Do we do some sort of discovery to understand the lay of the land Does the 68 weeks include development time as well Business and DataTech aspects to be added Walkthroughs and Sketches Identifying painpoints and needs DEVELOPMENT WIP What is BCM in this context BC for raw materials  what are my critical materials will I be able to source these materials from the right suppliers What substitutes do I have and supplier visibility who are my suppliers by critical materials What is their history with us Are they qualified to continue Therefore this business continuity plan is to be proactive rather than reactive to major disruptions in the inbound supply chain which are sourced by high impact supply risks KPI Secured Business Value Risk Exposure Rating x Business Impact Target is to Minimize risk exposure rating Process Steps prioritization of critical raw materials risk identification risk assessment and risk treatment. KPI to understand the completeness of your BCP Program secured business value Initial filtering Critical material Frame work Value at Risk Risk Exposure Rating x Business Impact Why is it so important today Disruptions not seen as a one off event Greater supply chain visibility efficiency and resilience are top of mind. changes in commodity prices supply chain bottlenecks continuing need to assess implications of choosing alternatives cost monitoring Supply disruptions may lead to severe costs for purchasing companies by causing supply chain delays which result in stockouts of raw materials and thus inability to meet customer demand and cost increases. Major disruptions may even threaten the very survival of a business therefore adopting a solid risk management approach is crucial Top executives at Global 1000 firms mention that supply chain disruptions and their associated operational and financial risks are their single most pressing concern Finch 2004. Supply disruptions result in severe costs for purchasing companies if handled poorly. They can cause supply chain delays which in turn trigger stockouts of raw materials and thus inability to meet customer demand and cost increases Blackhurst et al. 2005. Impact Radjou 2002 also lists a number of examples of quantifiable supplychain disruptions including General Motors experience of an 18 day labour strike at a brake supplier factory in 1996 which led to disruptions at 26 assembly plants with an estimated reduction in quarterly earnings of 900 million. Similarly Boeing experienced in 1997 supplier delivery failure of two critical parts with an estimated loss of 2.6 billion to the company. Likewise the lightning bolt that struck a Philips semiconductor plant in Albuquerque in March 2000 created a 10minute blaze that contaminated millions of chips and subsequently delayed deliveries to its two largest customers namely Nokia and Ericsson. As a result of this inbound disruption a 400 million loss was reported by Ericsson due to late chip deliveries from the Philips plant Latour 2001. How to identify critical materials impacted sales number of suppliers product grade count corresponding to supply chain complexity impacted customers corresponding to impacted reputation number of purchasing plants corresponding to supply chain complexity etc. Risk profile for supplier poor financial health of the supplier flexibility constraints of the supplier market constraints low number of qualified suppliers high lead time supplier being under no legal liability and geographical density of suppliers. One of the most effective ways of implementing risk assessment is to introduce risk registers Patterson 2002. Risk registers are tables that include information on the cause of a risk i.e. the risk source the risk event the outcome of the risk statement the risk consequence supply disruption and the impact and the probability of risk. Risk registers prove to be useful as a risk assessment tool as they provide an overview of all the active risks and their risk rankings. Enabled Recipe Mix Recommendation for the biggest US FMCG CASE STUDY Objective To minimize the complexity and diversity of Raw materials strengthening supply chain resilience and supporting the business c ontinuity p lans by recommending raw material in manufacturing for RD team to help them choose an alternate to prepare the final product Impact This will lead to a reduction of specs and higher share of market standard specs which will enhance buying power through volume A more competitive supply base therefore drive productivity which results in overall cost reduction Recommendations Recommendations for RD SMEs to replace a complex Spec with another better Spec based on similarity which helps in times of shortages and cost reduction Automated data workflow for seamless process integration output which results in elimination of manual clustering work Challenges Usage of similar raw materials with price variations which leads to increase in overall costs Shortage of certain raw materials in production which causes clear to build issues Solution Feature vectors Cleaning transformation and TFIDF to get feature vectors Clustering To a ssess the composition similarity between raw materials using hierarchical clusters Similarity Scores T o identify the most similar spec against each of the other specs within a cluster Model Explanability T o understand how why and with which features the model reaches a decision Shobhit to read Raw Material Specs Comparison based on the compositions VolumeSpend Distribution in selected Cluster Analyze CostSpend Variation to enable RD team to shortlistsubstitute Raw Materials in each region Hierarchical and KMedoids clustering approaches were followed to cluster the raw materials Spec distribution Clusterwise Raw Material comparison taking composition description spend and volume  Assess Similarity between individual Raw Materials using cosine similarity score Enables Raw Material substitution incase of shortage supply chain disruptions Spec comparison Supplier Uncertainties across value chain CASE STUDY Objective To improve endtoend experience of our users for strategic tactical supply chain resilience planning via proactive holistic decision making  to ensure business continuity which helps us to Identify Assess and Mitigate risks Impact 360 visibility into current and historic spend Identify suppliers that are most at  risk Mitigate risks associated with product availability and improve customer service Recommendations Generate multiple scenarios to visualize whatif state of the spend allocation and r ecommend multiple alternatives Suppliers to tackle the incoming risk Higher negotiation control to enable strategic partnership with lowrisk suppliers Risks Is there any opportunity to reduce overall yearly spend for product portfolio across suppliers Which factors to be considered to perform supplier risk evaluation and selection How to monitor and control supplier performance to track ontime infull delivery across value chain Solution Optimal spend allocation using forecasting optimization and scenario planning Determine supplier risk clusters based on risk profilesscoresfactors Predict risk severity occurrence its impact on OTIF Shobhit to read Data Requirement Plant level data SupplierVendor level data Qualified supplier table Critical Material Information Risk Tiering Supplier BCP Material composition Material Properties Raw materialProduct mapping Sales Product Data External Data Supplier Network Data Requirement VendorSupplier Manufacturing Consumption PropertiesDimension What is BCM in this context BC for raw materials  what are my critical materials will I be able to source these materials from the right suppliers What substitutes do I have and supplier visibility who are my suppliers by critical materials What is their history with us Are they qualified to continue Therefore this business continuity plan is to be proactive rather than reactive to major disruptions in the inbound supply chain which are sourced by high impact supply risks KPI Secured Business Value Risk Exposure Rating x Business Impact Target is to Minimize risk exposure rating Rethink the name Supply Resilience with Business Continuity Planning Nireeha Business continuity management is critical for procurement and RD teams 01 02 05 03 04 Shobhit Minimize text Check for sources articles Add some pointers from FLOW Business persona questions Manufacturing RPM Material Planner Manufacturing RD leader Material quality Lead Procurement manager RegionalGlobal Procurement Head We dont adopt a projectionbased inventory monitoring General Efficiency trends across lines are dropping Manufacturing plants are having low Compliance to Schedule We focus on reacting but not anticipating There is no S2P data KPI visibility We are unable to predict when MRO inventory norms S iddhartha to help explain the BQs Business persona questions BCM Define criticality Assess supplier risk Assess BCP strength Identify subs materials Identify subs materials suppliers Data Requirement VendorSupplier Manufacturing Consumption PropertiesDimension Demo Snippets from the Demo Nireeha Summary tiering Vendor qualification risk assessment Take substitute tool screen from Siddhartha Who to take this to What is the typical execution approach Competition Key Stakeholders and Org Functions Product Supply Team Business Units who would like to start their own BCPs Centre of Excellence Think of kind of problem and where they are at Full scale process engg digitization Recommend step 1 and 2 1 MVP Pilot 2 Scale What exists today BCP solutions can cater to multiple stakeholders and can be built in a phased manner Key Stakeholders Organisation Functions Centre of Excellence Product Supply Teams Business Units who would like to start their own BCPs 1 Execution Approach 2 3 Understand the problem at hand and where the group is at Phase 1 Pilot covering key part of the BCP process using a design infused approach Phase 2 Full Scale covering end to end BCP process with full scope engineering and digitization Typical solutions offer e2e risk management however analysis and plan creation is a niche Case Studies Transfer Pricing Map View Existing raw materials inventories by country location size of circle inventory amount Description of transfer pricing policies and standards by country Listing of available raw material suppliers by location and their payment terms Table View Planned supply volume and price for material from SOP incl. supplier payment terms consuming location Data entry simulation where changes could be made to source units or price Results screen to show outcomes of transfer pricing scenarios along with potential additional costs e.g. logistics and benefits incl. recommendations FINAL DO NOT EDIT THIS SECTION What is BCM Enable organizations to respond to events in such a manner that critical business functions can continue Target is to minimize risk exposure rating Proactive vs Reactive Focus Raw Materials Supplier visibility What are my critical materials Who are my suppliers Will I be able to source these materials from the right suppliers What is their history and associated risks What substitutes alternatives do I have What is BCM in this context BC for raw materials  what are my critical materials will I be able to source these materials from the right suppliers What substitutes do I have and supplier visibility who are my suppliers by critical materials What is their history with us Are they qualified to continue Therefore this business continuity plan is to be proactive rather than reactive to major disruptions in the inbound supply chain which are sourced by high impact supply risks KPI Secured Business Value Risk Exposure Rating x Business Impact Target is to Minimize risk exposure rating Rethink the name Supply Resilience with Business Continuity Planning Nireeha Can be deleted once previous slide is done Why BCM now 01 02 03 04 Business continuity management is critical for procurement and RD teams 01 02 05 03 04 Shobhit Minimize text Check for sources articles Add some pointers from FLOW Can be deleted once previous slide is done Steps prioritization of critical raw materials risk identification risk assessment and risk treatment. Wha t does Fractal Offer Risk assessment mitigation Identify material risk supplier risks Tier your materials and supliers Assess impact to business plan for Alternate sourcing location supplier strategies Alternative materials Planning and Monitoring Create BCPs Enable multi stakeholder collaboration Enable testing and monitoring E2E Digitization Process analysis and design standardization Digitize to gain e2e visibility for BC planning Shreya Can we deleted once next slide is done Risk assessment mitigation Identify material risk supplier risks Tier materials and suppliers by criticality Assess impact to business Identify alternate sourcing location supplier strategies alternative materials Fractals Solution Areas Planning and Monitoring Create BCPs Enable multi stakeholder cross functional collaboration Enable simulated testing E2E Digitization Process analysis and design standardization Digitization for e2e visibility and BC planning Shreya Data Requirement Shobhit to update Data Requirements Data Requirements Conceptual Dashboard LANDING PAGE to view critical materials by Type Criticality Product Region impact Status of the business units plan Detailed view of each critical material Conceptual Dashboard Tiering View historic history Tiering defined by all departments Define tier for each critical material Review conflicting tiers Collaborate live on one platform Conceptual Dashboard Supplier Risk History of supplier Status of qualification Conceptual Dashboard BCP Analyzer View factors determining risk Plan health by factor Recommendation on action BCP solutions can cater to multiple stakeholders and can be built in a phased manner Key Stakeholders Organisation Functions Centre of Excellence Product Supply Teams Business Units who would like to start their own BCPs 1 Execution Approach 2 3 Understand the problem at hand and where the group is at Phase 1 Pilot covering key part of the BCP process using a design infused approach Phase 2 Full Scale covering end to end BCP process with full scope engineering and digitization Fractal Solutions Delivered Logo Objective Logo Objective Logo Objective Shreya to add Nireeha to clean up Competitor What are the Top Supply Chain Management Software E2open SAP SCM Logility Perfect Commerce Oracle SCM Infor SCM JDA SCM Manhattan SCM Epicor SCM Dassault Systemes SCM Descartes SCM Highjump SCM IFS Watson Supply Chain BluJay SCM are some of the examples of best Supply Chain Management Software. Shobhit Keep only the top competitors Updated on next slide Competitor SAP SCM Oracle SCM Dassault Systems SCM Watson Supply Chain are some of the examples of Supply Chain Management Software. Provides 1 View into current ecosystem 2 Future order planning 3 Inventory levels Transportation planning Missing 1 Planning for what if scenario 2 AI material criticality benchmarking 3 AI Recommended alternate materials Custom Test plans Cost Disadvantage 1 License cost is by of users 2 Customization costs are very high Majority of the solutions today provides Supply chain analysis but provides zero outlook into robustness of your supply chain. Business Continuity Plan is not a priority of these solutions. Appendix Data Requirements Data Requirements Potential data landscape to consider PRODUCT MARKET FIT DESIGN EXECUTION MARKET EXECUTION AGILITY TO RESPOND DRIVERS OF LAUNCH EXTERNAL FACTORS REFERENCE Strategic Intent Product positioning Order of entry Product attributes Ingredients Flavor profile Packaging Benefits Concept testing scores  Raw material quality Packaging Labeling Health and wellness positioning Sustainability positioning Distribution Pricing Trade Spends Media Spend Competitive Launches GDP growth rate Per capita income Inflation Population and population density Regulatory compliance Ratings and Reviews Consumer sentiment Social chatter Feedback from D2C platforms snacks.com pantryshop.com

Direct Store Delivery & Inventory Allocation.pptx

Direct Store delivery Inventory allocation May 2022 01 Direct Store Delivery Benefits of enhanced fulfilment process Key steps to enhance fulfilment Product classification to identify suitable products for direct store delivery Order fulfilment through direct shipping of ordered DSD products from plant based on optimized DSD policy else through DC Guidance on inventory to hold at plant for direct store delivery and replenishment of products at DC to p rovide supply coverage to urgent normal orders of DSD products and for nonDSD products Order allocation optimization to provide consolidationwhat of items to be fulfilled from plants or DC based on shipment size Inventory levelsage at Plant DC fast turnover high consumer demand items   current inv. levels at stores Highlevel actions for an agile and costeffective fulfilment approach As fashion products are perishable in nature due to short demand cycle. Higher number of direct shipments leads to shorter lead time thus reduces overall time to market Direct shipments reduces to stocking of DSD inventory at DC leads to overall capital blockage and holding cost Guidance DSD fulfilment and optimized order allocation provides agile fulfilment to stores at reduced overall fulfilment cost Additional revenue by selling seasonal goods and fast turnover products. Short time in delivering new products to markets. Gives manufacturer regular market insights prices and trends. Preventing outofstocks. Savings on transportation costs. Key criteria to identify products for Direct Store Delivery Demand variability Products with stable or near stable demand trend make fulfilment process cost effective by predicting future demand much accurately Higher demand variability tends to increase urgent orders in system that leads to stock high buffer stock at DC Revenue contribution Group of fewer products contributes high in revenue and volume would requires optimum efforts for achieving business goals. Visibility of high margin products provides any loss on sales vs overall profitability Growth Market events Fast moving and growth in recently introduced products enables to predicts the SKUs which going to contributes to sales predominantly Solution construct Define data requirement input constraints Acquire required data build views Data transformation in required form for optimization model Products segmentation for direct store delivery to fulfill store orders from manufacturing plant Product with high in revenue with low demand variability and high RFM score or fastmoving can be qualify for Direct to store Delivery Develop mathematical and heuristics models for arriving at optimal replenishment plans At DC to fulfill nonDSD product orders and urgent DSD orders Quantity of DSD products to be stock at plant based on future orders and plant stock capacity Optimization of order allocation for urgent last minutes orders of DSD products from PlantDC and all nonDSD products from DC based on Current inv. Agelevels at DC Required delivery date Inventory levels at stores Order size Production Capacity Inventory stock at plant   High Inventory turnover products Solution construct P lan delivery routes automatically optimize delivery schedules and handle ondemand orders along with scheduled orders with rerouting on the go. Planning the shortest and most optimal routes for multiple delivery locations S orting daytoday delivery packages and improve distribution efficiencies in Direct Store Deliveries.. Routes can be printed on the shipping label to eliminate the manual dependence Product segmentation for direct store delivery Required information Product segmentation Final Interpretation Historical Demand Sales order history Retail store sales history Store geographical hierarchy Cost matrix Production cost Transportation cost Custom charges etc. External factors Seasonality Offers on weekends Holidays Promotion LocalNational Events ABC classification Using KMeans clustering XYZ classification for demand variability profit RFMFSM classification for product demand activity Vector Auto RegressionVAR for future consumption trend Ensemble modeling Multidimension segment grid to select product with high in revenuevolume AB with low demand variability XY and high RFM score or fastmoving can be qualify for Direct to store Delivery VAR modeling to capture the change in dynamics of future sales proportion of products. Classification modeling works with historical sales VAR model is to validate and change sales accordingly Twodimensional segmentation of product for illustration Replenishment Output Ship from PlantDC ShiptoStore Quantity Replenishment cycle Shipment Cost breakup Item Shipment volume DIFC Min Max boundary Shipments Lanes RDC details Item No. product group Size Color KPIs Cost of fulfilment Customer Service level Inventory levels Lead times Publish layer Consumption layer We will consider multiple layered approach to arrive at solution Cost Parameters Manufacturing cost Sales price Holding cost at DC Logistics cost Stockout cost Production Stock Production plan Throughput lot size Stock cover Product category DSDnonDSD Open RQ Priority Inputs and constraints Stores input Store stocking parameters Store geographical hierarchy Optimization of Orders Allocation Design Mathematical Model considering required objective constraints and business inputs Objective1 Maximize DSD SKUs from the plant Obj2 Minimize cost of fulfilment to stores Obj3 Minimize lead time to stores Optimize for the selected multiobjective function and measure impact on supply chain KPIs Constraints DC constraints capacity docks inventory Plant constraints Production capacity and plan finish good holding capacity line orders constraint Cartons pallets constraint Lane constraints Carrier constraints Enabled dynamic order fulfilment for eCommerce for the biggest US sporting goods retailer 1 Background and Challenges 2 Objective Sub optimal inventory fulfilling BOPIS eCom orders Ship from store Ship from DC Extend distribution network 800 stores 1 MFC as extended fulfilment centers with 2 additional MFCs Increased fulfilment costs due to splitshipments and long transit times Lack of workbench to simulate order fulfilment options from upcoming new stores Our 3step solution framework to optimize order fulfilment for quick turnaround Increased 1 day transit shipments by 40 reduced costs by 20 Business Benefits 99 shipments are within 1 2 transit days 95 MFC shipments from earlier 30 90 shipments within zone 2 4 from earlier 70 2018 Fulfilment by store type 2019 Fulfilment by store type 2018 Fulfilment by transit time 2019 Fulfilment by transit time Direct Store Delivery Plant All GTIN All Customer All Priority All FY Quarter All All Month All Product Segmentation Direct Store Delivery Plant All GTIN All Customer All Priority All FY Quarter All All Month All Gross Margin Volume Direct Store Delivery Plant All GTIN All Customer All Priority All FY Quarter All All Month All Products Products Sold To Ship To All All Illustrative purpose only Dashboards DSD Orders Customers GTIN Inventory Revenue Deliveries Settings Last 1 month Plant GTIN Customer Priority FY Month Quarter Revenue 5000 GM 70 All All All All All All All Plant All Plant All Month All Illustrative purpose only Dashboards DSD Orders Customers GTIN Inventory Revenue Deliveries Settings Last 1 month Product Category Sold To Ship To FY Month Quarter Revenue 5000 Volume 70 All All All All All All Last 1 month Plant Sold To Ship To Date All All All All Direct Store Delivery Summary Manufacturing Plant Manufacturing Plant For example the visuals should be able to answer which items are candidates for DSD why are they candidates What is the comparison of shipping these items DSD instead of from the DC impact on cost time etc.  DataPoints Challenges Highlevel actions for an agile and costeffective fulfilment approach Careful planning and management of delivery fleet is vital Rising transportation costs Lack of packaging and shipping expertise Lack of coordination on delivery status product inventory pricing changes Visibility and monitoring issues across retail locations Accommodate realtime order requests or dynamic changes in delivery schedules 02 Order allocation engine Potential enhancements Current process Manual selection of order lines based on filtering ageing orders for the desired customers No visibility into overall marginrevenueinventory holes of the order lines being selected Selection based on heuristics Eg 100 order lines1 truck. No way to optimize selection for target of trucks Due to the manual nature of the process no visibility into the tradeoff of selecting order lines based on different objectives age vs revenue vs margin vs holes Our understanding of the current allocation process Multi objective allocation considering ageing revenue margin and inventory holes to allocate orders Define a cost metric for singlemultiple objective that need to optimize in order to achieve effective dynamic order allocation model A user guided simulationbased tool to achieve optimal allocation for desired objective to achieve Work in sync with the load build solution to create an endtoend automated analytical solution Solution construct to optimize order allocation Order Allocation Solution Select Date Range Revenue Target Margin Target of SKUs of Trucks Available Soldto Shipto All Truck Type All Brand All Style All Style Family All Order Allocation Solution Summary DC 20M Revenue 50 Margin 45 Holes 600 Distinct SKUs Summary View Order Line View Soldto Shipto All Truck Type All Brand All Style All Style Family All Order Allocation Solution Order Line DC 20M Revenue 50 Margin 45 Holes 600 Distinct SKUs Summary View Order Line View Order Allocation Solution Customer Order Allocation Solution Customer Run Date DC Soldto Shipto All Truck Type All Brand All Order Allocation Solution Targets Style All Style Family All 20M Revenue 50 Margin 45 Holes 600 Distinct SKUs Revenue Holes Margin Run Date DC Soldto Shipto All Truck Type All Brand All Style All Style Family All Revenue Margin Holes Distinct SKUs Across all Metrics Order Allocation Solution Trade Offs Objective function Maximize the composite score across age revenue margin and holes Output Order allocation KPI and tradeoffs Connection to E1 for automation Dynamic order allocation model to optimize selected criteria Inventory Allocation Order Allocation model Problem Formulation Improved order allocation process for leading CPG client resulting in increased service levels at retail stores Challenge Less efficient manual order allocation process led to high OOS rates at retailers 14 of retailers in North America 70 of supply nodes 3.5 K SKUs Supply chain disruptions i.e. COVID Solution Components Impact delivered 90 Contractual Adherence while maximizing service levels 2 0 Workforce efficiency by elimination of cumbersome manual work Optimized inventory levels based on shelflife intelligence KPI Insights Intelligent allocation Cross functional collaboration Integrated and automated Scorecards Intuitive dashboards for KPI monitoring Automated diagnostics and exception management Early warning systems Intelligent allocation Allocate right SKU quantity to the right customer Simulate scenarios and final decisions C ross functional collaboration Enable weekly allocation decision process with intelligent tools Facilitate cross functional collaboration through integrated platform Integrated and automated Integrated with legacy systems Automated robust for data ingestion Scalable across new categories customers regions Transformed order allocation process KPI Insights KPI Insights Intelligent Allocation Integrated automated Cross functional collaboration Retailer scorecard for risk monitoring Drill drown and deep dive for evaluating category performance KPI Insights Intelligent Allocation Integrated automated Cross functional collaboration Exception management to identify products with constrained supply Improve allocation decisions with advanced optimization Simulate allocation Conduct simulations to view change in total delivery costs service inventory levels Assess tradeoff between different scenarios via automated scenario planner Intelligent order allocation Allocation decision based advanced multiobjective optimization algorithms i.e. improve service levels and sales reduce costs Combine business heuristics and s with information of retailer inventory product shelf life quota and order customer prioritizations Connected systems Integrate allocation recommendation at retailer level with legacy systems Build automated seamless system to upload allocation actions KPI Insights Intelligent Allocation Integrated automated Cross functional collaboration Constraints Inventory availability Order priority Target service levels MOQEOQ Contract clauses Storage capacity Input Parameters Forecast Open orders Scheduled receipts Inventory shelflife OnHand Qty Substitute products DC capacity Replenishment Visibility Optimum customer order allocation principles with maximize service levels E2e transformation of reporting allocation KPIs to provide insights on completeincomplete orders Tradeoffs between different optimization objectives to aid decision making Build right shipment projections for better warehouse and logistics planning Optimization Optimize fulfilment process by defining objective functions Obj 1 Maximize service level Obj 2 Minimize cost Explored optimization techniques like Linear programming GA simulated annealing Flexibility to incorporate business heuristics dynamics KPI Insights Intelligent Allocation Integrated automated Cross functional collaboration and operations research techniques Enabling weekly allocation decisions through cross functional collaboration Monday Tuesday Wednesday Thursday Friday Review results Review analyze performance scorecards Action on learnings from previous week Supply chain tactical direction setting to max supply availability Allocation outcomes review Review internal scorecard and allocation results Inweek order adjustments Discuss customer and order risks and opportunities Demand Supply reconciliation Review supply demand and inventory change scorecards SOE discussion Share allocation outcomes and shifts in demand mix Discuss inventory levels and any production changes Align on tradeoffs and actions from supply and commercial Integrated Tactical Planning Align on changes to production plan Summarize supply chain actions Action on learnings and outcome of tradeoffreconciliation meeting Supply Chain Sales Customer service KPI Insights Intelligent Allocation Integrated automated Cross functional collaboration Automated and connected systems for seamless data to decision journey Data Ingestion Process Organize Sample Business Questions Historic purchases retailer Quotes Supply and fulfilment Retailer inventory Volumes Source Types Manual Collection Forms ExcelCSV Entries ERP Advanced planning suites SnowflakeDB Schema What is the OOS performance in top retailers How should inventory be allocated to retailers in short term and longterm supply constrained situations What is the root cause and drivers for poor OOS performance Lead time value forecast contracts What categories are driving high OOS Data Mapping Use column names that can link tables through the keys Create a schema that comprises the Dimension and Fact tables Calculate derived metrics as per requirement of the subsequent analysis Use this base schema for modelling and visualization Data Harmonization In this layer we map the data sets and nomenclature followed in different functions of the organizations Data Mapping Transformation Sources Create SharePoint forms with input criteria for data capture Migrate to SharePoint for input criteria entry collection Use ETL jobs to make transformations and upload to cloud Use ETL jobs to create derived tables that can used for analysis Usecases KPI Insights Intelligent Allocation Integrated automated Cross functional collaboration Improved CPG clients order allocation process for priority customers Current scenario Whats the need Inventory allocation across customers is done at EOD each day by running GATP tool Customers are prioritized in categories A1 A2 A3 B1 B2 B3 etc. Current GATP tool considers constraints like Product substitution customer priorities MSTN etc. The inventory allocation for same day delivery customers is not efficient Efficient optimal ATP allocation for priority customers to enable same day deliveries Running optimizer on hourly basis to allocate inventory on the fly Incorporate multiple constraints like Service levels Product substitutions Inventory availability MSTN max stock to norm ratio Customer priorities Inter depot stock transfers What are the technical enablers Robust Advanced Analytics to optimize inventory allocation Standardize Optimization model to facilitate hourly inventory allocation shipment Interactive visibility solution to showcase future trend in inventory availability Ability to handle dynamic factors that could influence inventory allocation Allocation optimization Deep dive Output Hourly inventory allocation plans across open customer orders Same day delivery feasibility MILP Simplex or LPP based optimization Sample data requirement

Drug Similarity.pptx

Inventory Substitution by Drug Similarity Harini and Harshita 16 th Aug22 . Agenda Usecase Creating an approach on Inventory Substitution for Drugs Drug Substitute based on Similarity Main questions in this research How do we understand the properties of drugs What are the factorsproperties that needs to be considered for drug similarity How do we decide on drugs which are similar What can we recommendsuggest drugs using Similarity Benefits Recommend drugs that can be used as substitutes in the inventory Can discard drugs if they are not used Can eliminate those poor druglike compounds in advance and avoid more research and development expenses High Level challenges C ost of innovation and evolution R esearch and development expenses Effective Inventory control substitution Ways for Inventory Substitution To be effective pharmacists require to have an effective inventory control system that will Account to make substitutions . Find the demand for a drug within a set of medicines that share an active component and offer the same therapeutic effect. Drug Repositioning  Aims to identify new indications for existing drugs offers a promising alternative to reduce the total time and cost of traditional drug development. T o find potential new uses for existing drugs and apply the newly identified drugs to the treatment of diseases other than the drugs originally intended disease drug discovery The novel use of a drug is based on the assumption that if two diseases share some similar treatment profiles then the drugs used for only one of the two diseases could also be used for the other. By integrating drug or disease features information with known drugdisease associations the comprehensive similarity measures are firstly developed to calculate similarity for drugs and diseases. Then drug similarity network and disease similarity network are constructed and they are incorporated into a heterogeneous network with known drugdisease interactions. Automate replenishment by moving existing demand within a specific group of drugs from one to another when different manufacturers make orders be it because of a change in the preferred state or take in account availability or pricing problems. Pharmacists cannot simply substitute an instock item like the prescribed medicine for one that is out of stock. Suppose a medication is unavailable and the pharmacist cannot supply it. In that case they will require consultation with the medical professional who prescribed the medication to determine what substitutions are and arent suitable for the client Questions Challenges How do we decide on drugs which are similar How do we understand the important properties of drugs What are the various properties of drugs in terms of structure functionality etc What are the set of drugs labelled Preferred amongst the gambit of generic drugs How do we identify drugs that may not be suitabledangerous for switching The FDA has identified certain drugs that may be more dangerous to switch called narrow therapeutic index  NTI drugs Combinatorial drugs used in treatments estimate interactions amongst drugs combination changed for patients with allergic reactions .What drugs can be combined or substituted for each other A candidate drug with what kinds of elements and element proportion may have a greater possibility to be developed into an approved drug How do we deal with mixed data How can we come to conclusion on drugdrug interaction How do we deal with very less variability in the data What is their proper proportion Labelled data for similarity amongst existing drug types how to assess similarity amongst new drugs Choose drugs Drug lifecycle Initial Growing Maturity Sustain Decline Obsolete Based on upper and lower limit choose drugs which can be used as substitutes Clustering Perform text embeddings feature vectors Drug2vec scaling onehot techniques for mixed data and arrive at important vectors by feature selection methods Perform clustering model on the selected features to cluster the drugs Similarity Similarity within each cluster based on Drug Structure Drug functionality Drug Ingredients By drugdisease interactions drug related similarity disease related similarity etc Weighted similarity score Recommendations Rule based workflow on top of similarity scores to suggest drugs Solution Approach Drug representationlifecycle filtersFeature selectionClassification algorithmSimilarity calculationsdistance probabilistic correlation association coefficients Recommendations Calculating Similarity Similarity based on Drug Structure Drug functionality Drug Ingredients By drugdisease interactions drug related similarity disease related similarity etc Weighted similarity score Calculating Similarity for DDIs Integrated SimilarityConstrained Matrix Factorization ISCMF to predict DDIs . Eight similarities based on the drug substructure targets side effects offlabel side effects pathways transporters enzymes and indication data as well as Gaussian interaction profile for the drug pairs are calculated. Subsequently a nonlinear similarity fusion method is used to integrate multiple similarities and make them more informative. Finally we use ISCMF which projects the drugs in the interaction space into a lowrank space constrained to obtain new insight about DDIs . Drug repositioning data and approach Data requirements   Drug substructures D isease phenotypes P roteinprotein interactions PPI G ene profiles network profiles Approach C alculate similarity scores as well as drugdisease treatment priori. T he similarity matrices and the priori are integrated to construct drugdisease pair graph. Finally Semi Supervised Graph Cut algorithm is applied to predict drugdisease treatment relations How do we quantify Similarity If two drugs exhibit similar activities across multiple disease functionality then they are likely similar If two drugs has similar properties then they are likely similar structural shape color size pattern composition. Structurally similar molecules are assumed to have similar biological properties. Similar biological propertiesDrug discovery. Compounds protein binding site in the protein dockinggenerate many possible conformations of the compounds in the binding site Drugrelated similarity data has 10 networks based on   1 chemical structures 2 target protein domains 3 GO target protein annotations a target protein interactions b side effects c chemical structures 4 GO molecular functions 5 GO biological processes 6 GO cellular components 7 metabolism enzymes 8 protein sequences 9 anatomical therapeutic chemical classification codes 10 drug pairwise interactions. Diseaserelated similarity data has 14 networks based on   1 curated genes 2 HPO genes 3 literaturebased genes 4 curated variants 5 literaturebased variants 6 microRNAs 7 long noncoding RNAs 8 HPO phenotypes 9 ISA taxonomy 10 informationtheoretic similarity 11 GO terms 12 implicit semantic similarity 13 semantic and gene functional 14 curated association type ontology. Data Requirements Type Functionality Drug Molecular Structure Chemicalbiological Activity Descriptor Representation Patterns Chemical composition Date Data Requirements Categorical functionality functional group disease type cure sideeffects Text description Nature feedback sideeffects Numerical dosage sales composition proportion Chemical formula Structure Ingredients Protein No of ringsaromaticnonaromatic amino acids etc Substitution patterns Molecular Number of aromatic atomsNumber of atomsNumber of heavy atomsNumber of hydrogen atoms Number of aromatic bondsNumber of bondsNumber of double bondsNumber of rotatable bondsFraction of rotatable bondsNumber of single bondsNumber of triple bonds Composition made of chemical components Data Requirements Mockup Drug Representation Chemical representation Molecule similarity Protein structure onehot How do we choose the Suggestion High Level challenges Key Challenges in Pharmaceutical Industry 1 Demand forecasting enhance their customer service levels.. Ensuring that all the goods are produced at the desired time and delivered seamlessly poses a big challenge to the pharma companies. 2 Risk management Managing risks in the drug manufacturing process and quality systems is of great importance.Hence it becomes challenging for the pharma companies to identify possible risks associated with a product or processes involved in manufacturing development and distribution of the product. Limited clinical trials during the pandemic C ost of innovation and evolution Disruption in Availability of Raw Material for generic drugs Ways for Inventory Substitution To be effective pharmacists require to have an effective inventory control system that will Account to make substitutions. Find the demand for a drug within a set of medicines that share an active component and offer the same therapeutic effect. Drug Repositioning  Aims to identify new indications for existing drugs offers a promising alternative to reduce the total time and cost of traditional drug development. Automate replenishment by moving existing demand within a specific group of drugs from one to another when different manufacturers make orders be it because of a change in the preferred state or take in account availability or pricing problems. Pharmacists cannot simply substitute an instock item like the prescribed medicine for one that is out of stock. Suppose a medication is unavailable and the pharmacist cannot supply it. In that case they will require consultation with the medical professional who prescribed the medication to determine what substitutions are and arent suitable for the client pharmacies must be able to set up their inventory control system to recognize which manufacturers have received the preferred designation automatically or to use preestablished criteria such as prices or availability of to determine which business is the best supplier to place an order with. Based on past sales data a sophisticated system might translate the general demand for a category or the requirement for the selected product to the desired result Challenges How do we decide on drugs which are similar existing demand can only be transferred within a particular group of medications from one to another. What are the set of drugs labelled Preferred amongst the gambit of generic drugs keep track of changing local rules How do we identify drugs that may not be suitable for switching The FDA has identified certain drugs that may be more dangerous to switch called narrow therapeutic index  NTI drugs which may warrant further drug blood monitoring after a generic to branded drug substitution. What are the various properties of drugs in terms of structure functionality etc How do we deal with mixed data How do we understand the important properties of drugs How can we come to a conclusion on drugdrug interaction How do we deal with very less variability in the data W hat kinds of elements are beneficial to the druglike properties What is their proper proportion A candidate drug with what kinds of elements and element proportion may have a greater possibility to be developed into an approved drug Combinatorial drugs used in treatments estimate interactions amongst drugs combination changed for patients with allergic reactions .What drugs can be combined or substituted for each other Labelled data for similarity amongst existing drug types how to assess similarity amongst new drugs KGDDS A System for DrugDrug Similarity Measure in Therapeutic Substitution based on Knowledge Graph Curation PubMed nih.gov High level of challenges pharma industrysubstitute Drug Substitution an overview ScienceDirect Topics Questions How do we decide on drugs which are similar existing demand can only be transferred within a particular group of medications from one to another. What are the set of drugs labelled Preferred amongst the gambit of generic drugs keep track of changing local rules How do we identify drugs that may not be suitable for switching The FDA has identified certain drugs that may be more dangerous to switch called narrow therapeutic index  NTI drugs which may warrant further drug blood monitoring after a generic to branded drug substitution. What are the various properties of drugs in terms of structure functionality etc How do we deal with mixed data How do we understand the important properties of drugs How can we come to a conclusion on drugdrug interaction How do we deal with very less variability in the data W hat kinds of elements are beneficial to the druglike properties What is their proper proportion A candidate drug with what kinds of elements and element proportion may have a greater possibility to be developed into an approved drug Combinatorial drugs used in treatments estimate interactions amongst drugs combination changed for patients with allergic reactions .What drugs can be combined or substituted for each other Labelled data for similarity amongst existing drug types how to assess similarity amongst new drugs KGDDS A System for DrugDrug Similarity Measure in Therapeutic Substitution based on Knowledge Graph Curation PubMed nih.gov High level of challenges pharma industrysubstitute Drug Substitution an overview ScienceDirect Topics Similarity for drug repositioning The goal of drug repositioning is to find potential new uses for existing drugs and apply the newly identified drugs to the treatment of diseases other than the drugs originally intended disease drug discovery the novel use of a drug is based on the assumption that if two diseases share some similar treatment profiles then the drugs used for only one of the two diseases could also be used for the other. I dentify potential novel indications for a given drug. By integrating drug or disease features information with known drugdisease associations the comprehensive similarity measures are firstly developed to calculate similarity for drugs and diseases. Then drug similarity network and disease similarity network are constructed and they are incorporated into a heterogeneous network with known drugdisease interactions. Based on the drugdisease heterogeneous network BiRW algorithm is adopted to predict novel potential drugdisease associations. Overlapped KEGG pathways between Alcohol dependence and the predicted drugs. The blue hexagon nodes represent drugs predicted to treat Alcohol dependence the red vee nodes represent overlapped KEGG pathways between drugs and Alcohol dependence

Fractal_MedTech Supply Chain Art of Possible.pptx

Supply Chain Analytics solutions for MedTech industry Art of Possible June 2022 01 02 Transforming MedTech Supply Chain Art of Possible Relevant capabilities case studies Topics 01 Supply Chain Analytics solutions for MedTech industry Art of Possible Changing dynamics in MedTech industry.. Enable digitization Transforming from Doing digital to Being digital to provide connected solutions and achieve operational excellence Advance patient centricity Transitioning to patient centric ecosystem to empower and engage patients throughout their health journey Improve flexibility and sustainability Engaging in stresstesting operating models and increasing transparency to strengthen resilience and make the operations sustainable and future ready Drive value amongst customers Evolving from a traditional supplier role to a valuedriven strategic business partner to their customers posing critical Supply chain challenges 01 02 03 04 Transforming MedTech Supply Chain Art of possible Risk Resilience Maximize efficiency Control Tower End to end visibility across MedTech lifecycle Suppliers RD sites Production sterilization sites Distributors 3PL GPO Hospitals Pharmacies Enable cross functional decision making for ex. RD pre and post device launch analytics IT and design groups for demand at risk future service levels warning alerts Predict risk of delayed supply shortages and their impact to build assemble specifically complex devices Alternative options of Supply Chain planning to counter the disruptions Mitigation plans for delays through recommendations at Plant DC or HCOs such as Pharmacy Hospitals etc. Optimize inventory levels based on product portfolio Premium high value low velocity ex. Devices stents Commoditized low value high velocity ex. syringes Improved service levels by optimizing allocation decisions and thus minimize penalties Failure to Serve FTS warning alerts Optimize production and distribution processes to minimize supply chain inefficiencies across multimodal nodes Device components optimization based on criticality and Predictive maintenance to proactively predict a device failure Inventory optimization 02 01 03 04 MedTech Supply chain focus Fractals assessment of current need state Fractal research Analysis based on the challenges cited in MedTech firms annual reports investor presentations INTERNAL SLIDE A Control Towers Relevant capabilities case studies 02 Digital enablement e2e Supply chain visibility Real time metrics tracking across the Medical devices continuum of care Prevention Diagnosis Monitoring Treatment and Care Environment metrics such as sensor drift temperature etc. during device manufacturing Early warning alerts and cross functional nudges in case of critical dip in devices quality metrics Enabling decisions through Control towers Digital analytics and transformation to enable A. cross functional decision making B. shifting dynamics to homecare personalized healthcare C. Remote interactions with multiple stakeholders such as payers Build analytics engine and scalable technology plan for a successful digital adoption Transformed SOP process with Supply Chain Planning Control tower Single source of truth with right data at right time across 11 Sourcing platforms 25 Plants 13 DCs Enable collaboration across functions to drive Operational Strategic decisions Forward looking capabilities Integrated Supply chain data platform Cross functional user stories Demand at risk Early warnings alerts on upcoming risks Optimized planning scheduling asset utilization Faster decision making across Sales Operations planning Client ask Solution Result B Risk and Resilience Build resilience framework Build supply resilience Supplier analytics scorecard using key inputs such as performance spend and cycle times financial vulnerability of Tier 123 suppliers ESG performance etc. Dualalternate sourcing for raw materials redeploy from other networks Tracking Counterfeit risks at any node of the supply chain Building Risk and Resilience Build a MedTech resilience framework through the lens of 1. Regulatory factors supplier quality 2. Process factors for ex. Standard operating procedures 3. Product factors pre vs. post device launch performance 4. Structural factors manufacturing supplier location etc. 5. Operational factors sourcing concentration risk product complexity etc. Enabled Inbound supplier risk prediction and traceability Limited visibility of supply at risk Limited impact identification on supply demand reconciliation Slow tradeoffs analysis for mitigation strategies Collaborative platform for inbound shipment visibility that enabled Real Time Visibility Early warning signals on risks Recommendations to mitigate risks Product Traceability Impact of Risk across network nodes Enhanced E2E visibility to smoothen collaboration across nodes and post event analysis Client ask Solution Result C Inventory optimization Service levels Demand prediction Demand prediction by u sing AIML techniques by using the relevant MedTech input factors 1. Product Device sales price patent expiry efficacy safety etc. 2 . Market Formulary Therapeutic area spending etc. 3. External GDP growth Healthcare spending Disease prevalence etc. Manage optimized levels and prioritize inventory utilization based on MedTech variables such as Product shelf life Unique suppliers Product type highlow velocity Systematic alerts for expired and recalled inventory Inventory optimization through improved demand prediction and service levels Improve key levers impacting service levels such as Network optimization Existing alternate supplier proximity Driver analysis and RCA ex. Impact of inhouse distribution vs 3PL Improved forecast accuracy led to reduced inventory and improved service levels Poor customer service levels despite having huge inventory levels across the network static Inventory policy settings AIML solutions for improvements across Demand forecast accuracy Prediction of Inventory settings basis the volatility in Demand Supply Simulations for MoQs Service level Drivers analysis Improved accuracy with demand signals as input 20 reduction in safety stock across Top 3 brands Projected 100 service level with 27 reduction in working capital block Monte Carlo simulations to determine optimal supply replenishment from Factory to Warehouse Client ask Solution Result Improved accuracy by Demand sensing Working capital released by optimizing safety stock 14 1.9m Optimal replenishment recommendations Improved service level 1 D Maximize efficiency Optimizing distribution Optimizing p roduction Predictive maintenance to proactively predict a device failure and thus intervene prior to break down Device components optimization based on criticality Production Digital twins to enable simulation across all manufacturing levels Maximize efficiency by optimizing production and distribution T racking Multimodal shipment across multiple levels D evice components SKUs Purchase orders etc. around key metrics such as OTIF Fill rates Understanding distribution impact of productSKU fragmentation across regions owing to differing regulations approvals Improved packaging efficiency reducing both machine changeover times and demand supply gap Generate optimized packaging machines schedule within available capacity constraints A chieve 10 demandsupply gap packsize change overs minimize makespan time Optimal packaging production schedule Intelligent optimization Automated dashboard Reduced demandsupply gap by 70 Achieved average demand supply gap 2.5 vs. Current state 8 gap Reduced average changeover time by 5 Client ask Solution Result 1 plant in China 3 Packaging machines 40 SKUs 9 flavors 7 Pack sizes Impact created across Optimized Logistics Costs with e2e visibility cost drivers through AIML interventions Standardize definition of Logistics cost performance KPIs e2e visibility of cost components Standardize ways to attribute costs Forward looking view of costs rather retrospective view of costs Integrated Logistics data platform with reduced latency in data to decision making from 1QTR to 1 Week e2e cost performance management with descriptive predictive prescriptive analytics Drivers mix with customer level BBN models 25 million reduction in Logistics costs through Transportation cost drivers improving routing guide improved dynamic spot price Client ask Solution Result

Inventory RnR.pptx

A Inventory Resilience PRIORITY Risk Resilience Mapping Resilient Planning Strategic Tactical Responsive  Execution Collaborate Integrate Streamline OTIF deliveries Allocation optimization Network planning Supplier uncertainty Product  Flow Structural Inbound Our POV on supply chain risk and resilience Downtimes Material availability Production plan Operational Tactical Strategic How might we predict risks mitigate product flow across the network from sourcing lens How to monitor and control supplier performance to track ontime infull receipts How to manage inbound Supply chain nodes to ensure right product at right place and at right time How to maximize equipment utilization and optimize production schedule to meet demand surges How to minimize unplanned downtimes via predictive maintenance How to plan for material availability basis demand volatility and projected production plan What should my target stock levels be based on long term demand trends What is my OutOfStock and Excess risk based on supply and demand volatilities and how to mitigate What is my current inventory profile and how do I obtain optimal product mix What is the optimal distribution network to minimize costs and maximize fulfilment How to continuously monitor outbound shipments and enable early warnings for interventions How to find the optimal fulfilment center allocation for orders basis supply and demand constraints Inventory settings Safety stock Strategic build up Inventory related Business questions that can be addressed through the solution 01 Strategic Risk Ensuring Strategic stock levels at various nodes based on latent demand patterns and external perturbances Approach What Your Organization Can Do to Stay Ahead Analyze product life cycle to understand current phases of products Identify inventory drivers and its impacts observed in history Develop Bayesian networks to approximate impacts due to internal and external events Challenges The Challenges Behind building strategic stock Inconsistencies due to localized Inventory Management leading to inconsistencies Poor visibility on inventory drivers No recommendation on anticipatory build up for phaseinphaseout products Impact Achieving financial and service goals Improved Service Level with right inventory level Right balance between inventory and working capital Reduction in stock holding and handling cost SKU life cycle assessment to identify products in planning phase model their risk network Inventory Drivers Impact of internal or external factors on historical inventory positions Markup Adjustment factors for different internal external factors Understand impact of different markup ranges on different events Product Life Cycle List of products moving towards EndofLife EOL phase and products in EOL phase Patterns of SKUs staying in a specific life cycle stage Patterns on SKU characteristics before moving into Degrowth Forecast Production and Inventory details for products in EOL and moving towards EOL phase Bayesian Risk Networks Risk network with impact of external vs internal disruptions Impact on costs and inventory due to external internal events i.e. supplier behavior global local behavior Generate warning alerts recommend mitigation actions Recommend mitigation options in case of event occurrence Determine amount of time spend by products in product life stages Calculate average life of products within a defined grouping Using classification models determine placement of the products in its life cycle Utilize inventory data and sales forecasts along with business logics to fine tune prediction of early EOL Modelling a Bayesian network of impact of external vs internal disruptions Using a compound Poisson process to assess the probability distribution of events and estimating the joint probability of any event occurring Analytical process to recommend strategic stock adjustment factor Causal Network Conduct root cause analysis to identify impacts observed on inventory due to internal external factors Simulate impacts by creating causal network based on RCAs identified Calculate adjustment factors from historical gaps observed in inventory Segregating SKUs into planning and EOL Stage Product life cycle stages EOL prediction Classification model Determine time spend by products in each product life stages by mapping historical data Utilize inherent product hierarchy to group similar products Calculate planning stage end of life stage and transition point for product grouping Calculate average life of products within a defined product grouping Identify transition point for defined product groups Use classification models to de termine whether product is in end oflife stage or not over a period of next 24 months Utilize inventory data and sales forecasts along with business logics to fine tune prediction of early EOL C Category B Brand PL Product Line Transition Point Product Hierarchy Not in EOL EOL Bayesian Risk Network for Planning Stage SKUs Bayes Network HillClimbing algorithm to determine best network with an optimized score and arc strength Blacklist weakest arc strength andor do not make business sense Retrain the model with blacklist in place to obtain final solution Supervised model to detect probability of internalexternal event occurrence considering all evidence present Variable Node Decision Node Utility Node SC Disruption Triggers Mitigation Strategies SC Risks Decisions Consequences Utility Inventory Risk Procurement Adjustment Material Shortfall Quality Issues Transport Disruptions Sporadic Demand Inconsistent Lead Time Original vs Simulated Inventory units dollars Original vs Simulated forecast units dollars External Natural Calamity War Geopolitical Pandemic Internal Financial health Unplanned Breakdown Supplier disruption Forecast Adjustment Advance Procurement CAPA Production Adjustment Anticipatory Built Supply Risk Manufacturing breakdown Risk Transportation Risk Sovereign Risk Production Plan Adjustment Forecast Adjustment Data Required 02 Tactical Risk Review current inventory profile and optimize for assortment Approach What Your Organization Can Do to Stay Ahead Inventory portfolio assessment to segregate products based on their behavior Balancing width and depth of assortments with range and options planning Assortment planning to answer what when and how much of inventory to keep Challenges The Challenges behind assortment optimization Addiction to SKU led growth model more SKUs the merrier it is for Sales growth Fear of competitor not following suit and losing sales Tried assortment cost remained same and sales plummeted Impact Achieving right assortment mix Reduction in Nonperforming Inventory Ability to drive initiatives to reduce anticipatory build Lesser supply chain cost less picking less shipments less warehouse Optimizing assortment through range and options planning Range Options Planning Planning Categories SubCategories SKUs Styles Sizes Quantities Other Product Attributes Costs Prices Margins Deciding factors include Sales Plan Management Strategy Competition Campaigns New Stores etc. Inventory Portfolio Product classification based on ABC and XYZ segmentation Inventory norms for different categories Right balance between service level and inventory based on business requirement Assortment Plan Recommend product mix for the portfolio based on segment Optimal Inventory recommendation for identified product mix Codify business rules thresholds for alertsrisks Clustering and profiling products in categories based on historical 12 months demand impact and revenue impact Develop a multicriteria clustering system based on SKUs demand supply volatility value Volume and velocity Define a hierarchy for grouping SKUs Geographic Volume Velocity Develop costbased liner optimization model to maximize profit Determine top product attribute groups based on expected sales Demand analysis based on historical sales and growth trend to forecast expected sales for identified product grouping locations across the network Determine which SKUs to stock based on various constraints like cost space minimum quantity Analytical process to review inventory profile and do assortment optimization Develop Mixed Integer Linear Programming MILP models to maximize sell in Plant to DC efficiency for range plan and sell throughsold to customers efficiency for option plan Model outputs Optimized range plan by PPG and option plan by PPG by Cluster Range options plan and Assortment plan is more suitable for Retail businesses Optimization for Range Planning Upper Limit Lower Limit Upper Limit Mixed integer Linear programming Lower Upper Limit Range Plan Predicted Demand Historical sales Product master data Sell In SI Qty Quantity shipped from plant to DCs Optimization for Option Planning Predicted Demand Historical sales Output from Range planning Capacity data Upper Limit Lower Limit Upper Limit Upper Limit Upper Limit Mixed integer Linear programming Option Plan OP Sell ThroughST Qty Quantity of products sold to customers Optimization for Assortment Plan Predicted Demand Historical sales Output from Range planning Capacity data Upper Limit Lower Limit Upper Limit Upper Limit Upper Limit Cost based liner Optimization Retail Price 1 Cost 1 x i Min Qty 1 Retail Price 2 Cost 2 x i Min Qty 1 ... Pr ofit Data Required 03 Operational Risk Ensuring optimal stock availability at different nodes in supply chain amid demand and supply uncertainties Approach What Your Organization Can Do to Stay Ahead Assessing SKUs to understand possible risk of over stock or understock Determine shortages and overages based on generated forecast and planned supply Eventbased simulation to identify understock or overstock situations Challenges The Challenges behind inventory levels Localized operations and reporting leading to lack of visibility and collaboration. Lack of dataoriented decision making Lack of scenario analysis to simulate inventory requirements service levels at different nodes in supply chain Impact Mitigating Overstock and Understock Improved Service Level with enhanced customer satisfaction Less deadstock inventory lower holding cost and increased net revenue  Improved Cash flow for working capital Assessing SKUs at risk to simulate for overstockunderstock Simulation Time based details for understock and overstock for up to 12 weeks Inventory level situations at various predefined service levels and inventory parameters WarningAlerts for possible OOS and Excess inventory situations Risk Profiling Understand SKUs which are at risk Inherent trends in demand supply and distribution Predicted OOS and Excess situations Scenario Analysis Ability to create and compare possible mitigations scenarios Time and Cost impact associated upon exercising mitigation options Recommended optimal mitigation plan Narrowing down SKUs atrisk using Statistical analysis Demand Supply variability Demand Supply distribution Average Demand Interval Analyze historical forecast planned supply and shipment data for patterns and trends Calculate OOS and Excess inventory situations based on Forecasted demand planned supplies and shipments Scenario analysis for mitigation OOS risk by what if analysis on Inter facility transfers by overriding existing allocation Negotiation delivery timelines Tweaking production schedules Scenario analysis for mitigation excess inventory risk by what if analysis on Inter facility transfers by overriding existing allocation Tweaking production schedules Analytical approach to assess and mitigate OOS and Excess inventory Forecast Optimize Simulation of inventory levels w.r.t. Desired Service levels Predicted demand Planned Supply distribution parameters Incorporating constraints like MOQ Customer Service level DC Capacity Holding cost distribution shipment schedules Data Required Thank you

KOP - Optimization for Assortment Planning.pptx

Optimization in Assortment Planning 31 March 2022 adidas China account Contents Merchandise Planning Process adidas China project Assortment Planning AP Process Optimization techniques implementation in AP Key takeaways 01 An Overview of Merchandise Planning 1 3 Year plan by year quarter and month Sales margins markdown targets Multichannel considerations Planning on new stores brand Competition Demand Forecasting Strategic Planning Forecasting Sales Plan Inventory Plan Open to Buy OTB Buying Plan Inputs from Category Managers Focus group discussions Retail Operations teams feedback historical sales analysis are all inputs for the design team Design team picks up styles for range Production lead time for e.g. approx. 34 months Samples are sent by the design team 34 months ahead of the season Budgeting Buying Max. SKUs styles Sizes Quantities per style costs prices margins are defined in range planning Balancing width and depth of assortments with options planning. Deciding factors include Management Strategy New Stores Brands Competition Campaign etc. 2 Seasons Spring Summer April to August Autumn Winter September to March 1520 Core styles Carry forward styles 80 New Seasonal Styles Range Planning Process of selecting the products to put on the shelf in order to maximize sales and margins for every season Cluster stores by location and product attributes Plan product mix SKU rationalization. Discontinue unprofitable products What to buy how much to buy and in what size ratio Consider Historical Sales Category planning Men Women Kids etc. SKUs Style and Size Stores Capacities Store Clusters Assortment Planning Optimization Tracking sales on weekly basis and deciding on inter store transfers stock transfers between stores promotions placing orders Achieving Sellthrough targets at full prices Recalculate OTB based on trend Pricing Revising Prices and Discounts Reduce need for markdowns End of Season sale Promotions based on sellthrough targets margins inventory on hand actual sales In Season Planning Retail Merchandise Planning RANGE AND ASSORMENT PLANNING 02 About adidas China project Overall adidas project tracks Efficiency as a foundation while improving Range Assortment effectiveness Improve Range Assortment Planning OBJECTIVE approach A. Efficiency initiatives Increase analysis maturity B. Effectiveness initiatives Increase value to business Adopt efficient processes to improve speed ease agility Improve effectiveness of RA decisions Adidas Product Range 03 What is Assortment Planning Range Plan How many unique SKUs to carry in a product group Option Plan How many unique SKUs to carry in a product group within a store cluster Assortment Plan Which SKUs to carry in a store Range and Assortment Planning 12 Range and Option Plans SKUs Apparel Footwear Accessories Men Women Unisex Football Basketball Running Etc. TShirt Jacket Shorts Etc. 200300 300500 500700 Etc. Cluster 1 Cluster 2 Cluster 3 Etc. SKUs SKUs SKUs SKUs Range Plan Option Plan Division Gender Category Product Type Price Point Store Cluster Assortment Plan Assortment Plan Red Cotton Blue Knitted Black Yellow DoubleKnit DoubleKnit Colour Fabric Design Sizes SKU 1 SKU 3 SKU 11 SKU 17 Division Gender Category Product Type Price Point Store Cluster Range and Assortment Planning 22 Demand forecast Extract insights from historical performance analysis Decide on selective delisting and carryover products at subcategory level Decide on strategic listing at sub category level Split by channel Store Cluster Arrive at of SKUs ASP by Brand Category SubCategory Price Band etc. Hand off to Buying Decide at store level Decide Size Ratio Stagger units across weeks of the season aligned with Marketing calendar Create Assortment Plan Create Range Plan Option Plan Qualitative Review Input Qualitative Review Input Decide on selective delisting and carryover products at SKU level Decide on strategic listing at article level Business Goals Basic Analysis Advanced Analytics Data Science Rate of Sales ROS ST Qty per store per week ROS ST Qty 90D No. of Stores 90D 12 SI Efficiency SI Qty SKUs ST Efficiency ST Qty SKUs Sell In SI SI Qty Quantity shipped from adidas factories to DCs SI Value Yuan SI Qty Retail Price Sell Through ST ST Qty Quantity of products sold to customers ST Value Yuan ST Qty Average Selling Price Sell Through ST ST Qty 90D SI Qty Key Performance Metrics 04 Problem Statement Problem Statement Challenges Existing process are based on excel based analysis of historical sales data involving lot of manual effort Multiple decision points with subjective inputs based on experience Planning process distributed within different teams and is disconnected Scope for improvement in analytical sophistication Business Problem Improve article efficiency and thereby improving planning effectiveness Generate the Range Plan Option Plan and Assortment Plan each quarter and season SpringSummer and FallWinter Devise a analytics based methodology that does better planning as compared to existing processes Reduce cannibalization and increase cross cluster differentiation Solution Approach Understand existing planning processes data sources generate process maps and identify improvement areas Develop an automated multistage optimization solution based on MILP algorithm with An integrated workflow linking range and option planning stages Flexibility to assign various business constraints based on strategy and ART Integrated data sciencebased and flexible planning Siloed solutions. Heavy on planners experience. Efficiency NOT optimized mathematically. Cluster product differentiation NOT quantified. No quantification of product cannibalization. Very long list of KPIs. Time consuming process. Range Plan Option Plan Assortment Plan Automated Process Merchant Business Inputs Three enablers for effectiveness improvement Quick flexible and scalable process Automated process and optimized on Sales efficiency. Input of Option Plan. Visualized result and review. Optimized by cluster recommender with differentiation and insights. Input of Assortment Plan. Visualized result and review. Product attribute level recommender. Cannibalization measurement. Similar top product mapping. 1. Range optimization 2. Option plan optimization 3. Assortment insights Current state Future state with effectiveness solutions 05 Optimization in Assortment Planning Developed Range Optimization RA engine Automated Range PLAN Architecture Template for additional constraints input by users Data Processing Optimized Range Plan by PPG Optimization Process Input to OP Option Planning SI data ST data Python Based Framework Visualization of optimizer output Additional User constraints Flexible to add strategic constraints Range Plan Option Plan Assortment Plan Optimization for Range Planning Upper Limit Lower Limit Upper Limit Mixed integer Linear programming Lower Upper Limit Range Plan Range Plan Option Plan Assortment Plan Developed Option Plan Optimization Engine Automated OPTION PLAN Architecture Data Processing Optimized Option Plan by PPG by Cluster Optimization Process Option Plan Range Plan ST data Python Based Framework Visualization of optimizer output Capacity data Range Plan Option Plan Assortment Plan Optimization for Option Planning Upper Limit Lower Limit Upper Limit Upper Limit Upper Limit Mixed integer Linear programming Option Plan Range Plan Option Plan Assortment Plan Sales Transferability Cannibalization How much sales are lost when a product is delisted Or Which new articles will cause least cannibalization Sales transferability gives the amount of sales that will be lost and transferred to other products on delisting Top Product Attributes Which articles belong to the top product attribute groups based on expected sales. The attribute groups for example could before example colourfabric combinations like BlackCotton etc. Financial Sales Performance Which articles did not perform well in previous season based on metrics like sales value volume sales per store per week distribution etc. Potential Sales Which of the new articles are expected to have more sales Expected sales of an article using latent demand analysis based on historical sales and growth trend of similar articles Assortment Planning Listing DeListing Range Plan Option Plan Assortment Plan 06 Benefits and Key Takeaways Challenges Key takeaways Solution Learnings Role of art vs math in the decisionmaking process To allow manual inputs for art in the assortment consider modular solution approach instead of integrated solution New process adoption change management For business buyin multiple workshop meetings elaborating the underlying optimization logics assumptions and constraints Changing business processes meta data etc. To follow agile delivery project plan Highlight improvements and similarities by comparing with output from existing business process. Benefits Summary Improved Flexibility and Speed Planners can provide additional business constraints based on ART and Business Strategy Faster time to insights into next seasons Range Assortment Feasibility to do more planning iterations before final plan Improved Effectiveness Up to 48 shift in Range to highly efficient PPGs from low efficient PPGs Increase in Sell in Efficiency Increase in Sell through efficiency Improved cross cluster differentiation Decision Enablement Optimized range decisions considering cluster differentiation strategy and assortment insights The solution removes the effect of subjective inputs and biases A robust and scalable analytical decision methodology Potential and actual benefits Retail Merchandise Planning Flow Forecasting Range Planning Budgeting Assortment Planning InSeason Planning Demand Forecasting Sales Plan Inventory Plan Open to Buy OTB Buying Plan Planning Categories SubCategories SKUs Styles Sizes Quantities Other Product Attributes Costs Prices Margins Deciding factors include Sales Plan Management Strategy Competition Campaigns New Stores etc. Selecting Products to maximize sales and margins every season Clustering stores by location and product attributes Planning Product Mix Quantities Size ratios SKU Rationalization ListDelist Tracking sales Recalculate OTB Placing Orders Inter store transfers Price revisions Markdowns Discounts Design Planning Inputs from Category Managers Focus group discussions Retail Operations teams feedback historical sales analysis Design team picks up styles for range Production End of season sales Promotions based on sellthrough targets margins inventory on hand actual sales End of Season Planning Production with a lead time of for e.g. 34 months Merchandise is allocated to stores based on plans and inventories Merchandise Allocation Purchase Orders Sales Plan Designs Samples Finished Goods Market China Channel Wholesale Store Channels CoreSP OriginalsOCSBCS Customers Franchise Customers Project Scope Adidas sells through 12000 stores in China 400 own retail stores in China Ecommerce through adidas.com.cn TMALL JD etc.

Manufacturing Use Case - Waste Reduction.pptx

Supply chain DS Manufacturing Use cases Waste Reduction May 2022 Use Case Understanding Sustainability To improve efficiency and effectiveness To save significant costs To improve resilience and competitiveness To attain net zero To add value to the customer Procurement Supplier issues Lead times Material defects Quality E rrors in ordering or procurement Inventory Excess Inventory Cost of excess inventory Inventory bottlenecks Out of date stocks Inventory mismanagement Less Inventory turnover Production Over production Under production Unplanned downtime Usage of nonrenewable resources nonreuse Production scrap fifo Improper packaging Excess gas emission Transportation Waiting times Routing issues Logistics delays docking Delivery delays Excess gas emission What is a waste in Supply chain Why do we have to reduce waste Factors that causes wastes in Supply chain 2021 Fractal Analytics Inc. All rights reserved Confidential Proposed Solutions C ollaboration between buyers and suppliers Ratio of defective parts Supplier delivery rates Ensure quality Lean approaches like JIT Justintime logistics can help you get closer to a 100 perfectorder measure. Inventory turns visibility Adjust inventory levels based on demand Lean inventory management accurately predict customers future demand and allows your warehouse to switch to a pull model of inventory management in which stocks are only refreshed when theyre depleted below a certain threshold. Safety stock Recycling and Reuserecycled materials are used as input materials and are heavily processed into end products Use of raw materials can be reduced or expensive materials be replaced Waste can be reduced by reusing materials using less hazardous substitute materials nonrenewable resources Identify which products are generating waste minimizing raw material wastage Review production lines Packaging production replace expensive materials with cheaper ones. Monitor firstin firstout FIFO to reduce scrap Bulk order and packaging quantities to reduce the need to dismantle and repackage product Material requirements planning This demanddriven supply chain sees products being manufactured according to the amount of recent sales and the number of indemand items currently in inventory. This allows manufacturers to accurately calculate how many items are needed to replenish inventory which reduces waste in the factory Demand forecasting Reduce the need to expedite materials and the steps in processing individual orders Plan the Movement of Goods Based on Demand Lean approaches like JIT Justintime logistics can help you get closer to a 100 perfectorder measure. Inventory Procurement Production Transportation Factors which influence the use case 2021 Fractal Analytics Inc. All rights reserved Confidential Proposed Solutions Procurement Spend and Supplier analytics Analysis of Dailymonthly Inventory to increase visibility Inventory optimization techniques Distribution center capacity optimization cross docks logistics managementalternate routes Streamlining portfoliodata Forecasting demand Digital transformation Analysis of procurement chain to understand purchase patterns raw materials and supplier groups Predict and optimize Spend across suppliers. Check supplier clusters in high demand regions Understand spend patterns and improve visibility. Supplier analytics performance spend and cycle times Dual sourcing and alternatives for raw materials Redeploy raw materials from other networks Placing atrisk products on customer allocation to avoid stockpiling Analysis of dailymonthly raw material inventory for cash flow and bottom line improvement Inventory monitoring Inventory management Optimization Safety stock and Inventory cost analytics Waste reduction and improve performance using lean and six sigma techniques Analysis of batchwise production patterns raw material consumptions and plant utilization Manufacturing capacity management increase inhouse manufacturing Efficient production Supply levels to match the needs Supply Demand Planning S OP Shift production to most critical products Demand Forecasting based on history data to predict the future demand using AIML techniques Recommendations to mitigate risks Analyze trends to spot spikes and anomalies in inventory levels demand production and distribution Logistics Port alternate routes Trip and Tradelane analytics Manage lane Expedite shipments air S OE Updates on product availability to distribution centers Cross docks stocking locations distribution center capacity Merge physical and digital supply chain to increase visibility Create digital supply chain twins Enable MedTech with endtoend supply chain visibility Predict risk of delayed supply and their impact Mitigation strategies or alternatives Appendix What Waste be it material gas or excess inventory etc is the enemy of efficiency and effectiveness. expensive waste process inefficiencies communication gaps lack of or delayed responses and even errors in ordering or procurement transport inventory motion waiting overproduction overprocessing and defects Unplanned downtime is a giant problem The cost of excess inventory Defective products they produce more than they have to Packaging reducing pershipment packaging weight . Challenge Statistical tests were the first piece but they are essentially only useful when products have already been shipped in more than one package type textbased data that customers find on the Amazon Store the item name description price package dimensions and so on Our model detects the packaging edges to determine shape identifies a perforation a bag around the product or light shining through a glass bottle Why improve their resilience and competitiveness saves significant manufacturing related costs. Many benefits can be realized by waste minimization or source reduction including reduced use of natural resources and the reduction of toxicity of wastes. Dry cleaning example petroleumbased compounds reduce costs Impact to reduce costs improve return on investment manage resource scarcity and reduce the environmental impact of operations fail to manage their waste will find their operations become less sustainable those who do not operate in an environmentally friendly way pay higher costs leading to waste when outofdate groceries and pharmaceuticals had to be disposed of Overproduction of products means youll have to pay for warehouse space to store the excess. Worse still if the products in question are perishable such as food goods you could end up having to throw some away if they expire Reasons 1. Lead Times 2. TransportationRouting Issues Waste in this area includes the obvious like extra fuel costs due to idling trucks incomplete loads and unpredictable costs of the fuel itself. 3. Inventory Mismanagement Weve hinted at issues in this area already but  inventory mismanagement  is a huge source of logistics delays. Go back to our initial example of the common item stored at the opposite end of the warehouse from the loading docks. Thats a delay in stocking the item when it arrives and another delay waiting for it to be brought out and loaded for delivery. 5. LastMile Overproduction of products means youll have to pay for warehouse space to store the excess How switching to electric vehicles  Prevention or elimination of waste is of course the golden standard however redistributing unpreventable surpluses to alternative markets is the next best option. What needs to be avoided is sending waste to landfill. Recycling is therefore viewed as the third best option in the UK Waste Hierarchy. collaboration between buyers and suppliers developing lean methods for supply chain management. Recycling and Reuse recycled materials are used as input materials and are heavily processed into end products use of raw materials can be reduced or expensive materials be replaced Review production lines and administrative processes for unnecessary procedures and rework Review equipment and delivery policies INBOUND OPERATING HOURS PALLET LOSS AND BREAKAGE AVERAGE DELIVERY LEAD TIME AVERAGE TIME IDLE DURING DELIVERY FUEL COSTS INVENTORY ONHAND SUPPLIER DELIVERY RATES SHIPPING HANDLING RATES RATIO OF DEFECTIVE PARTS The manufacture of products is often based on the demand of consumers. This demanddriven supply chain sees products being manufactured according to the amount of recent sales and the number of indemand items currently in inventory This allows manufacturers to accurately calculate how many items are needed to replenish inventory which reduces waste in the factory Analyze Product Design One way of not just  reducing  waste but also optimizing your production is to examine and reevaluate your products design. Identify any areas or methods to reduce raw material use or replace expensive materials with cheaper ones. If you can shave off small costs they might result in substantial savings. You should also  evaluate your product packaging options  and see if theres any way you can use cheaper materials. Manage Resources Apart from just looking at just using cheaper materials you should examine each of your production processes to identify which ones are generating waste. Redesign processes that are creating nonrecyclable or nonreusable waste. Even recyclable waste should be assessed and you should take the cost of recycling into account. Improving Production Quality Quality control is often focused on finished products but one of the goals of quality management should be minimizing raw material wastage Inventory Management Lean approaches like JIT Justintime logistics can help you get closer to a 100 perfectorder measure. Apart from reducing order errors it can also help you cut down tremendously on a lot of associated costs of inventory management like warehousing utility costs rentals and even insurance and taxes. However placing multiple orders may increase transport costs and your vendors might charge higher rates if each order is of a lower value eight types of waste in lean manufacturing Overproduction Waiting Inventory Transportation Overprocessing Motion Defects Workforce eKanban increased inventory turns by as much as 91 and cut inventory costs in half  Materials can spend as much as 90 of cycle time in the queue learn how to measure and accelerate queue turns Reduce the need to expedite materials and the steps in processing individual orders Automatically adjust inventory levels based on demand Monitor firstin firstout FIFO to reduce scrap Transform purchasing into a more strategic function waste can be reduced by reusing materials using less hazardous substitute materials or by modifying components of design and processing This method of streamlining and reducing waste in processes is called lean supply chain management. Radio Frequency Identification RFID  allows automatic capture of data This allows you to more accurately predict customers future demand and allows your warehouse to switch to a pull model of inventory management in which stocks are only refreshed when theyre depleted below a certain threshold. use of nonrenewable materials  Sustainability Inventory Transportation Delivery Common areas that you can discuss with suppliers to reduce product waste are Ratio of defective parts Recyclability weight and volume of packaging materials Packaging takeback options Bulk order and packaging quantities to reduce the need to dismantle and repackage product ML Solutions Cost efficiency due to machine learning which systematically drives waste reduction and quality improvement Lastmile delivery Machine learning in supply chain can offer great opportunities by taking into account different data points about the ways people use to enter their addresses and the total time taken to deliver the goods to specific locations. ML can also offer valuable assistance in optimising the process and providing clients with more accurate information on the shipment status. Industrial waste makes up at least 50 of global waste. Much of that waste is the result of poor quality products from manufacturing processes. AI can help workers reduce waste and lower the cost of quality. Whats also clear is that poor quality is one of the most preventable sources of manufacturing waste. where to send the goods so we optimize inventory levels and how to make sure we dont use excess resources along the way.  achieved a 20 reduction in forecast error and a 30 reduction in lost sales by using machine learning to predict demand One of our customers achieved a 4 gain in gross profits because of a 30  drop in unplanned downtime . Optimize the Speed of the Supply ChainMachine learning can analyze timings and handovers as products move through the supply chain. It can compare this data to benchmarks and historic performance to identify potential holdups and bottlenecks and make suggestions to speed up the supply chain. Forecast Likely Demand From CustomersData can be sourced from many areas like the marketplace environment seasonal trends promotions sales and historic analysis. Machine learning will combine this data to predict demand for specific goods and help to manage the sourcing and manufacture of those products. Plan the Movement of Goods Based on Demand Ensure Quality from Suppliers Products and AssetsQuality is vital to good SCM as waste and faulty products create unnecessary rework and increase costs. Machine learning can monitor how quality varies over time and suggest improvements. This doesnt just apply to materials and products. It can track other areas such as shipping supplier and thirdparty quality. Machine learning delivers several benefits for SCM Organizations in the supply chain do not need to hold as much inventory because machine learning optimizes the flow of products from one place to another Costs are reduced due to machine learning driving quality improvement and waste reduction Products arrive in the marketplace just in time for sale as a result of upstream optimization Supplier relationship management becomes easier due to simpler proven administrative practices Stakeholders get more insight into meaningful information allowing for continual improvement and easier problem solving 2021 Fractal Analytics Inc. All rights reserved Confidential

Med Tech Supply Chain analytics - Art of Possible July 2022.pptx

Supply Chain Analytics solutions for MedTech industry Art of Possible July 2022 Changing dynamics in MedTech industry.. Enable digitization Advance patient centricity Improve flexibility and sustainability Drive value amongst customers Transforming from Doing digital to Being digital to provide connected solutions and achieve operational excellence Transitioning to patient centric ecosystem to empower and engage patients throughout their health journey Engaging in stresstesting operating models and increasing transparency to strengthen resilience and make the operations future ready Evolving from a traditional supplier role to a valuedriven strategic business partner to their customers posing critical Supply chain challenges Global complex Supply chain Actionable realtime data is often unavailable lacking or siloed and hence difficult to track devicesproducts across its lifecycle sourcing manufacturing warehousing and distribution MedTech must adapt quickly to rapidly shifting pandemic conditions while planning for multiple potential next normal scenarios Predict right demand and provide optimal service levels to customers need to optimize multiple factors such as demand volatility supply disruptions excess product on hand or shortages product recalls product portfolio etc. MedTech requires optimal supply chain processes to eliminate inefficiencies and adhere to high quality products to avoid product recalls and thus achieve cost optimization reduce revenue loss and brand reputation risk ShortLack of Resilience Lack of end to end Supply chain visibility Inventory optimization Demand prediction and optimal service levels Supply chain inefficiencies 01 02 03 04 Transforming MedTech Supply Chain Art of possible End to end visibility across MedTech lifecycle Suppliers RD sites Production sterilization sites Distributors 3PL GPO Hospitals Pharmacies Enable cross functional decision making for ex. RD pre and post device launch analytics IT and design groups for demand at risk future service levels Predict risk of delayed supply shortages and their impact to build assemble specifically complex devices Alternative options of Supply Chain planning to counter the disruptions Mitigation plans for delays through recommendations at Plant DC or HCOs such as Pharmacy Hospitals etc. Optimize inventory levels based on product portfolio Premium high value low velocity ex. Devices stents Commoditized low value high velocity ex. syringes Improved service levels by optimizing allocation decisions and thus minimize penalties Failure to Serve FTS Optimize production and distribution processes to minimize supply chain inefficiencies across multimodal nodes Device components optimization based on criticality and Predictive maintenance to proactively predict a device failure 02 01 03 04 Risk and Resilience Demand prediction and service levels Control Tower Minimize inefficiencies Digital enablement e2e Supply chain visibility Real time metrics tracking across the Medical devices continuum of care Prevention Diagnosis Monitoring Treatment and Care Environment metrics such as sensor drift temperature etc. during device manufacturing Early warning alerts and cross functional nudges in case of critical dip in devices quality metrics Digital analytics and transformation to enable A. cross functional decision making B. shifting dynamics to homecare personalized healthcare C. Remote interactions with multiple stakeholders such as payers Build analytics engine and scalable technology plan for a successful digital adoption Enabling decisions through Control towers Transformed SOP process with Supply Chain Planning Control tower Single source of truth with right data at right time across 11 Sourcing platforms 25 Plants 13 DCs Enable collaboration across functions to drive Operational Strategic decisions Forward looking capabilities Integrated Supply chain data platform Cross functional user stories Demand at risk Early warnings alerts on upcoming risks Optimized planning scheduling asset utilization Faster decision making across Sales Operations planning Client ask Solution Result Build resilience framework Build supply resilience Supplier analytics scorecard using key inputs such as performance spend and cycle times financial vulnerability of Tier 123 suppliers ESG performance etc. Dualalternate sourcing for raw materials redeploy from other networks Tracking Counterfeit risks at any node of the supply chain Build a MedTech resilience framework through the lens of Regulatory factors supplier quality Process factors for ex. Standard operating procedures Product factors pre vs. post device launch performance Structural factors manufacturing supplier location etc. Operational factors sourcing concentration risk product complexity etc. Building Risk and Resilience Result Enabled Inbound supplier risk prediction and traceability Limited visibility of supply at risk Limited impact identification on supply demand reconciliation Slow tradeoffs analysis for mitigation strategies Collaborative platform for inbound shipment visibility that enabled Real Time Visibility Early warning signals on risks Recommendations to mitigate risks Product Traceability Impact of Risk across network nodes Enhanced E2E visibility to smoothen collaboration across nodes and post event analysis Client ask Solution Service levels Demand prediction Demand prediction by using AIML techniques by using the relevant MedTech input factors 1. Product Device sales price patent expiry efficacy safety etc. 2. Market Formulary Therapeutic area spending etc. 3. External GDP growth Healthcare spending Disease prevalence etc. Manage optimized levels and prioritize inventory utilization based on MedTech variables such as Product shelf life Unique suppliers Product type highlow velocity Systematic alerts for expired and recalled inventory Improve key levers impacting service levels such as Network optimization Existing alternate supplier proximity Driver analysis and RCA ex. Impact of inhouse distribution vs 3PL Inventory optimization through improved demand prediction and service levels Client ask Solution Result Improved forecast accuracy led to reduced inventory and improved service levels Poor customer service levels despite having huge inventory levels across the network static Inventory policy settings AIML solutions for improvements across Demand forecast accuracy Prediction of Inventory settings basis the volatility in Demand Supply Simulations for MoQs Service level Drivers analysis Improved accuracy with demand signals as input 20 reduction in safety stock across Top 3 brands Projected 100 service level with 27 reduction in working capital block Monte Carlo simulations to determine optimal supply replenishment from Factory to Warehouse Improved accuracy by Demand sensing Working capital released by optimizing safety stock 14 1.9m Optimal replenishment recommendations Improved service level 1 Optimizing distribution Optimizing p roduction Predictive maintenance to proactively predict a device failure and thus intervene prior to break down Device components optimization based on criticality Production Digital twins to enable simulation across all manufacturing levels T racking Multimodal shipment across multiple levels D evice components SKUs Purchase orders etc. around key metrics such as OTIF Fill rates Understanding distribution impact of productSKU fragmentation across regions owing to differing regulations approvals Maximize efficiency by optimizing production and distribution Improved packaging efficiency reducing both machine changeover times and demand supply gap Generate optimized packaging machines schedule within available capacity constraints A chieve 10 demandsupply gap packsize change overs minimize makespan time Optimal packaging production schedule Intelligent optimization Automated dashboard Reduced demandsupply gap by 70 Achieved average demand supply gap 2.5 vs. Current state 8 gap Reduced average changeover time by 5 Client ask Solution Result 1 plant in China 3 Packaging machines 40 SKUs 9 flavors 7 Pack sizes Impact created across Optimized Logistics Costs with e2e visibility cost drivers through AIML interventions Standardize definition of Logistics cost performance KPIs e2e visibility of cost components Standardize ways to attribute costs Forward looking view of costs rather retrospective view of costs Integrated Logistics data platform with reduced latency in data to decision making from 1QTR to 1 Week e2e cost performance management with descriptive predictive prescriptive analytics Drivers mix with customer level BBN models 25 million reduction in Logistics costs through Transportation cost drivers improving routing guide improved dynamic spot price Client ask Solution Result