Segmentation in Demand Planning for Enhanced …

1 Jim Davis Jeff Metersky CHAIN alytics Segmentation in Demand Planning for Enhanced forecast Accuracy Jim Davis, CPIM Director, Demand Planning & customer Service at Colgate-Palmolive (Global customer Service & Logistics) Jim currently serves as Director of Demand Planning and customer Service at Colgate-Palmolive. Jim has more than 34 years of supply chain experience at Colgate including manufacturing, Planning , customer service and logistics. Prior to joining Colgate s global supply chain team, he lead the US customer Service and Logistics organization. His current responsibilities include global process ownership for Demand Planning and customer supply chain collaboration. Jim holds a Bachelor of Science in Industrial Engineering from Lehigh University and an MBA in Operations Management from Fairleigh Dickinson University.

2 Jeff Metersky Vice President, S&OP Practice Chainalytics Jeff is a co-founder of Chainalytics and Vice President of the Sales & Operations Planning Practice. His global consulting experience which spans more than 100 clients across a variety of industries is focused on supply chain design and analysis, inventory strategy and optimization, Demand Planning , and cost-to-serve analytics. Jeff has authored multiple articles and is frequently cited by leading industry publications and analysts. In 2006, Jeff was recognized as a Pro to Know by Supply & Demand Chain Executive. Jeff holds a Bachelor of Science in Industrial Engineering from The University of Illinois and a Master of Business Administration in Materials and Logistics Management from Michigan State University. Applying Segmentation to Evaluating forecast Accuracy Forecasts Drive the Majority of Demand & Supply Planning Decisions in Most Manufacturers Product FlowForecastsOfProductSalesOrdersto Suppliers,& forecastsof ordersDeploymentsTo FieldDistributionCentersProductionSchedu lesProductionPlans& Financial Planning typically uses aggregated forecasts Demand Planning relies on a mix of aggregate and SKU-specific forecasts Supply Chain management mostly uses quite detailed forecasts.

3 OSKU / Shipping location / day or weekCustomerordersInventory& POSDataVMI service of Retailer DC s&DSD serviceto storesPlant ship to Retailer DC sFinancial PlanningMarketing and Promotion PlanningProduction StaffPlanningOutboundTransPlanningDC StaffPlanningImproving Item-Location forecast Accuracy Drives Operational Efficiency CompanyChannelsSales RegionsCustomersShip-To sCategoriesBrandsProduct GroupsItemsNetworksEchelon LevelsLocationsTime BucketsYearlyQuarterlyMonthlyWeeklyToday s Demand Planning Environment (1)Supply Chain Insights 5/13 Survey of 92 Companies (2) IBF Research Summer 2011 Supply Chain Pain Points(1) Demand and Supply Variability Top Pain of Supply Chain its increasing 2000-20042005-20091 Month Lag2 Months Lag4 Months Lag1 Month Lag2 Months Lag4 Months Lag1 Month Lag2 Months Lag4 Months LagConsumer ProductsFood and BeverageIndustrial Products75%72%72%76%72%73%74%68%74%74%72 %68%74%77%74%79%69%67% Demand Planning performance improvements are down/flat SKU level forecast Accuracy(2) Traditional Benchmarking has Not Provided a Path for Improvement 63%81%79%77%83%54%43%40%30%40%50%60%70%8 0%90%100% 0 Lag 1 Lag 2 Lag 3 forecast Accuracy (FCA) Can I really improve this much?

4 If so, where and how can I improve? Questionnaire-based Participants self-report forecast accuracy as they measure it Forecasting process checklist Attempt to define best practice Limited root-cause and comparative analysisConventional SurveysThere Must be a Better Way: Sales & Operations Variability Consortium (SOVC) Industry: Non-Durable Consumer Product Goods, Food & Beverage Geography: US customer Demand Members: 40+ Participants Item-Locations: 300,000+ What is our underlying Demand uncertainty? Is our forecast accuracy and bias reasonable compared to competitors and peers using common metrics? What is causing our challenges in forecasting? How well do we forecast , relative to that inherent uncertainty? Does the way we compute error distort comparisons? How do we prioritize improvement opportunities?

5 Are we better at forecasting some types of products than others? What are the underlying drivers of error, such as product portfolio, customer order patterns, economic cycles, seasonality, new product launches, Food & Beverage 51%Personal Care 30%Home Care 12%Pet Care 7%What is Segmentation ? Process of dividing a large unit into various small units which have more or less similar or related characteristics How can this help me improve Demand Planning ? 81%61%54% of Units Shipped in PatternStableTrendingSeasonal/UpliftInte rmittentLaunch/EndOtherFCABiasMember 2 Member 3 Member 1 Demand Pattern Mix Influences forecast Accuracy My Planning world is more dynamic! Forecastability of Demand Patterns Stable RangeRound UpliftTrend UpTrend DownSharp UpliftPhase OutPhase InIntermittent ConsecutiveIntermittent Non-ConsecutiveWeekly ForecastersMonthly ForecastersEasierto ForecastHarderto ForecastHarderto ForecastEasiestto ForecastI have the most difficult patterns to forecast .

6 Member 3 FCA 54% High Variability Low Variability High Velocity, Low Variability High Velocity, High Variability Ship QtyLowItem LocVelocityVariabilityShip Variability and Velocity Influence forecast Accuracy Ship QtyLowItem LocVelocityVariabilityShip 1 FCA 81% % High Variability % Low Variability High Velocity, Low Variability High Velocity, High Variability My Demand is way more variable. High-Velocity,Low-VariabilityMedium-Velo city,Low-VariabilityHigh-Velocity,Medium -VariabilityLow-Velocity,Low-Variability High-Velocity,High-VariabilityLow-Veloci ty,Medium-VariabilityMedium-Velocity,Hig h-VariabilityLow-Velocity,High-Variabili tyMedium-Velocity,Medium-VariabilityWeek ly ForecastersMonthly ForecastersEasierto ForecastHarderto ForecastHarderto ForecastEasiestto ForecastForecastability of Demand Variability & Velocity I have the most difficult variability to forecast LagNearly 1,600 Benchmarks of forecast Accuracy & BiasMonthlyNetworkLocationWeeklySegmenta tion Framework for Benchmarking Product Portfolio Characteristics Drive Performance Benchmarking Based on Forecastability Index What should my business accuracy be?

7 30%40%50%60%70%80%90% Item-Location FCAS&OVC Forecastability IndexMonthly Forecasters Lag 1 ExpectedUnderOverI m over performing! Improvement Opportunities & Setting Realistic Targets Finally, Differentiated Accuracy Targets Based on Reality Colgate-Palmolive s Demand Planning Segmentation Strategy Who is Colgate-Palmolive? Started in 1806 in the $17+ Billion in global sales 35,000 People worldwide Operations in over 80 countries Selling products in 225 countries Four core categories: Oral Care Personal Care Home Care Pet Nutrition Pet Nutrition Home Care Personal Care Oral Care Business Challenges Rapidly evolving retail environment in developed and developing markets Greater demands for customized products and new product innovations Increasing competitive pressures in global and local markets Volatile Demand with less lead time Managing information flow across multiple networks Limited Planning resources challenged to manage Demand Effective Demand Planning across Colgate s cross functional team is critical to overcoming these challenges.

8 Demand Planning Segmentation is the tool that has allowed us to drive effectiveness. What is Segmentation in Colgate Demand Planning ? A process that splits a portfolio up into SKU segments that have similar Demand characteristics SKUs with the same Demand characteristics can be forecasted using similar approaches Segmentation Objectives Manage portfolio complexity Identify variability or volatility in Demand Prioritize and focus Planning activities and resources Leverage statistical models Increase Demand Plan Accuracy and Effectiveness Not Every SKU Behaves the Same Way Channel of Distribution and Retail Environment Item Usage (Impulse vs. Everyday) Seasonality Volatility of Demand Volume or Velocity Our approach to segmenting SKUs is based on criteria that Volatility & Volume SKU Segmentation HIGH LOW Volume LOW HIGH High Priority SKUs LOW VOLUME HIGH VOLATILITY Up to 50% of SKUs < 10% of Volume HIGH VOLUME LOW VOLATILITY 10 to 20% of SKUs Up to 40% of Volume LOW VOLUME LOW VOLATILITY ~10% of SKUs 10% of Volume HIGH VOLUME HIGH VOLATILITY < 20% of SKUs Up to 40% of Volume Volatility Threshold CV < 40% Segmentation Matrix Volatility (CV%) vs.

9 Volume Segmentation Matrix Volatility (CV%) vs. Volume HIGH LOW Volume LOW HIGH LOW VOLUME HIGH VOLATILITY Hard to Predict Low Impact Items HIGH VOLUME LOW VOLATILITY Easy to Plan/ forecast High Impact Items LOW VOLUME LOW VOLATILITY Easy to Stat forecast Low Impact Items HIGH VOLUME HIGH VOLATILITY Hard to Predict High Impact Items Volatility Threshold CV < 40% HIGH LOW Volume LOW HIGH LOW VOLUME HIGH VOLATILITY Use Inventory Strategies to Manage Volatility HIGH VOLUME LOW VOLATILITY Model Baseline Volume and Collaborate on Uplifts LOW VOLUME LOW VOLATILITY Use Statistical Models & Manage by Exception Focus Efforts & Collaboration on High Volume & Volatility SKUs customer Inputs CPFR Focus Collaborative Demand Planning Efforts & Resources Segmentation Works! business decreased Demand Planning Error (weighted MAPE) by over 5% Global subsidiaries that have applied Segmented Demand Planning : DPA Inventory Coverage Keys to Success Segment your category SKU portfolio to: Improve accuracy Increase DP efficiencies Need cross functional understanding of benefits to drive ownership and focus Start by prioritizing high volume SKUs with higher volatility and lower DPA Focus organization on quick wins!

10 0%5%10%15%20%25%30%35%40%45%50%0%5%10%15 %20%25%30%35% Demand Planning Error Volatility Opportunity to Improve DPA Error aligned with Volatility in Demand Demand Planning Error vs. Volatility Focus on Higher Volume & Volatility SKUs with High Error Keys to Success Leverage statistical forecasting Stable Demand Baselines for higher volatility SKUs Evaluate lower volume and higher volatility SKUs Rationalize where possible Cover with safety stock Move to make-to-order Next Steps Expand global roll-out and application Leverage Chainalytics SOVC analysis to expand Segmentation Increase stat modeling to cover some predictable Demand volatility Fully implement integrated DP Segmentation tools in SAP APO Questions & Answers Jim Davis Jeff Metersky Survey