Solution-Driven Integrated Learning Paths Educational Sessions Lean Global Supply Chain Basics of Operation Management Demand Management, Forecasting, and S & OP Professional Advancement Special Interest Topics Plant Tours Networking Events/Peer Interaction Materials for Sale at the APICS Bookstore Make the Most of Your Educational Experience Develop a Learning Plan Assess your learning needs Use teamwork Prepare to learn Create your own Action Plan Live Learning Center Sync-to-Slide www.softconference.com/apics
Steven Crane, CSCP, is Director Strategic Supply Chain Management NAFTA, Wacker Chemie AG, where he is responsible for strategic supply chain management for the Wacker Polymers Division. This includes demand planning, supply planning, and sales and operations planning. Currently, Crane is working on the implementation of APO and S&OP for several businesses in the division. A Roadmap to World Class Forecasting Accuracy Six Keys to Improving Forecast Accuracy Stephen P. Crane, CSCP Wacker Chemical Corporation Director Strategic Supply Chain Management Content WACKER Company Overview Why Forecast? Forecasting Challenges Six Keys to Improving Forecast Accuracy Process, People, & Tools Statistical Forecasting Forecasting Segmentation Data Aggregation Sales Adjustments to Statistical Forecast Measurement & Exception Reporting Forecasting Accuracy Results Conclusions
Content WACKER Company Overview Why Forecast? Forecasting Challenges Six Keys to Improving Forecast Accuracy Process, People, & Tools Statistical Forecasting Forecasting Segmentation Data Aggregation Sales Adjustments to Statistical Forecast Measurement & Exception Reporting Forecasting Accuracy Results Conclusion Over 90 Years Of Success Wacker Chemie AG Founded in 1914 by Dr. Alexander Wacker Headquartered in Munich, Germany WACKER Group (2007) Sales: 3.78 billion EBITDA: Net income: Net cash flow: R&D: 1.00 billion 422 million 644 million 153 million Capital expenditures: 699 million Employees: 15,044 We Are Committed To Benchmark-Quality Products Designed For Our Focus Industries Industries Adhesives Automotive and transport Construction chemicals Gumbase Industrial coatings and printing inks Paper and ceramics Products: Polymer powders and dispersions for the construction industry Polyvinyl acetate solid resins, polyvinyl alcohol solutions, polyvinyl butyral and vinyl chloride terpolymers
Content WACKER Company Overview Why Forecast? Forecasting Challenges Six Keys to Improving Forecast Accuracy Process, People, & Tools Statistical Forecasting Forecasting Segmentation Data Aggregation Sales Adjustments to Statistical Forecast Measurement & Exception Reporting Forecasting Accuracy Results Conclusion Forecast Types Most companies use three types of forecasts: Sales or channel forecast Corporate planning forecasts Supplier forecasts These forecasts are very different in their use, frequency, and definition Need consistent demand signal across these three forecasting processes, but most companies don t know how to align them Why Forecast? The forecast drives supply planning, production planning, inventory planning, raw material planning, capital planning, and financial forecasting Companies that are best at demand forecasting average: 15% less inventory 17% higher perfect order fulfillment 35% shorter cash-to-cash cycle times 1/10 the stockouts of their peers 1% improvement in forecast accuracy can yield 2% improvement in perfect order fulfillment 3% increase in forecast accuracy increases profit margin 2% Source: AMR Research 2008
Inventory Level Vs. Forecast Accuracy 120 Inventory Days of Sale 100 80 60 40 20 0 20% 30% 40% 50% 60% 70% 80% 90% 100% Source: Aberdeen Group 2008 Forecast Accuracy at SKU Location What Is A Good Forecast? World class forecasting accuracy performance 95% currency by product line 90% by product family 85% product mix level Goal Source: Buker Management Consulting Content WACKER Company Overview Why Forecast? Forecasting Challenges Six Keys to Improving Forecast Accuracy Process, People, & Tools Statistical Forecasting Forecasting Segmentation Data Aggregation Sales Adjustments to Statistical Forecast Measurement & Exception Reporting Forecasting Accuracy Results Conclusion
Management Is Often the Problem Forecasting Challenges Typical forecasting process involves historical demand data loaded into a database using software to generate statistical forecasts Statistical software is rarely allowed to operate on its own Management usually overrides the statistical forecast before agreeing to final forecast Forecasting is often a difficult and thankless endeavor with high inaccuracies Companies react to inaccuracies with investments in technology, but do not guarantee any better forecasts Forecasting Process There are often fundamental issues that need to be addressed before improvements can be achieved Forecasting Challenges When businesses know their sales for next week, next month, and next year, they only invest in the facilities, equipment, materials, and staffing they need There are huge opportunities to minimize costs and maximize profits if we know what tomorrow will bring but we don t. Therefore we forecast!
Content WACKER Company Overview Why Forecast? Forecasting Challenges Six Keys to Improving Forecast Accuracy Process, People, & Tools Statistical Forecasting Forecasting Segmentation Data Aggregation Sales Adjustments to Statistical Forecast Measurement & Exception Reporting Forecasting Accuracy Results Conclusion World Class Forecasting Accuracy Requires Making The Right Decisions Six Keys To Improving Forecast Accuracy Step 1: Defining the Process, People, & Tools
Step 1: Defining the Process, People & Tools Process Forecast accuracy improvement occurs with the proper blending of process, people, and IT tools Overemphasis on any one leads to an imbalance that can defeat the desired result The process should be defined first, then followed by roles and responsibilities, and then IT applications The more people that touch a forecast, the greater the bias and the greater the forecast error Sales & Operations Planning Process Forecasting & Demand Planning WD 0-16 Demand Planning Data Preparation Phase [WD 0-7] Upload New Add New Review Review Update Sales History & Demand Business Historical Sales Demand Business Change Entries in R3 Data Combos Metrics Summary Create Unconstrained Demand Plan [WD 8-16] Apply Historical Apply Planning Review FC Create APO Data Sales Data Type Adjustments Statistical Passed to BW Adjustments Assignments with Sales Forecast Managers BT/BU Review DP Passed to Apply Future Review Final Publish Demand & Approve CO for Sales Manager Unconstrained Changes Unconstrained Financial Figures Demand Plan Demand Plan Forecast Supply Planning WD 13 -End Supply Network Planning Preparation Phase Submit Off Update Loc Preferred Review Supply Update Supply System Shift Table For Sources Planning Change Demand Future Source Assigned to Metrics Summary Figures Changes Demand Refresh Adjust Supply Analyze Draft Release FC to Update Master Inactive Planning Supply / R3 For MRP Data & Upload Version / Inventory Release DP to Constraints Capacity Plan SNP Supply Networking Planning WD 1-6 Sequential Finalize & Business Unit Approve Planning Supply Plan [WD 1-5] [WD 6] Review Supply Upload WD 1 Shortage & Inventory Capacity Overload Alerts Enter Adjusted Resolve Supply Demand Alerts Submit Supply Adjust Target Plan with Stock Levels Documented Options Fix SNP Approve Planned orders Supply Plan Develop Updated Supply Plan Proposal Approve Supply Chain Plans WD 7-8 Partnership Executive Meeting S&OP [WD 7] [WD 8] Review Action Items from Last Month Review Unconstrained Demand Plan Exceptions Review Supply Options & Cost Projections Review Financial Plan Review Performance Metrics Key Business Issues & Resolution Summarize Supply Chain Plans Review Action Items from Last Month Review Performance Metrics Review Supply Chain Plans Review Revenue Projections Agree & Communicate Approved Plans Communicate Implications to Financial & Sales Plans Automated Jobs Step 1: Defining the Process, People & Tools People People are critical to the forecasting process and how it s used within the organization They need to understand how their role fits with the work process and how to make improvements Position descriptions need to be defined with clear responsibilities that are accepted by the organization Full-time positions are essential Limit number of people making decisions about final forecast
Step 1: Defining the Process, People & Tools IT Tools Forecasting software is sometimes sold as the answer to forecasting issues Software does not solve forecasting problems. Processes and people solve problems Implementing forecasting software should not be considered an IT project, but a business process improvement project Forecasting applications can eliminate much of the manual work associated with forecasting if configured properly Six Keys To Improving Forecast Accuracy Step 2: Establish Statistical Forecasting Capability Step 2: Statistical Forecasting Capability Many supply chains are too complex to manually generate forecasts for all products and customers Forecasting engines are widely used to improve forecast accuracy by generating statistical forecasts Statistical forecasting uses sales history to predict the future by identifying trends and patterns within the data to develop a forecast Need to decide at what level the forecasting should be done Product family Individual product Product/customer Product/customer ship-to Plant/product/customer ship-to
Step 2: Statistical Forecasting Capability Need to decide how often to forecast Quarterly Monthly Weekly Daily Determine how much sales history is required for meaningful statistical forecast (minimum 2 years) Sales history master data must be correct especially when migrating from legacy systems. Very difficult to do correctly Analyze the forecast error associated with using available forecasting algorithms to optimize accuracy of forecast Step 2: Statistical Forecasting Capability Typical statistical forecasting methods include Multiple regression analysis Trend analysis Seasonal Simple moving average Weighted moving average Exponential smoothing Recommended Automatic selection Step 2: Statistical Forecasting Capability It is critical to make appropriate sales history adjustments to generate an accurate statistical forecast Adjustments Necessary to History Data errors Lost product volume Lost customers One time customer outages Discontinued products Non-optimal sourcing
Six Keys to Improving Forecast Accuracy Step 3: Forecasting Segmentation (80/20 Analysis) Data Aggregation Business Intelligence Exceptions Collaboration Step 3: Forecasting Segmentation - 80/20 Analysis It is crucial to distinguish the high-value items for special attention while automating the not-as-valuable items Statistical Forecastability (measured by 1/COV) High Low Notes COV (Coefficient of Variation) = STD Deviation/Ave. Demand Non-High Impact Items Use Data Aggregation in Statistical Model for all Non-HI Items Manage by Exception using Exception Reports Low Sales Volume/Impact High Impact Items Gather Business Intelligence for all HI items Collaborate with Customer (if possible) ~ 80% Total Volume High Step 3: Forecasting Segmentation Example
Six Keys to Improving Forecast Accuracy Step 4: Data Aggregation Step 4: Data Aggregation SAP APO 4.1 used for forecasting and demand planning Forecasting done at product/customer ship-to level Thousands of unique customer / product combinations exist to manage Too much data for Planners to review monthly Sales history for many combinations (~80%) was sporadic and difficult to forecast, i.e., high forecasting errors So how do you get a good forecast for sporadic combinations? Forecast at a more aggregate level Step 4: Data Aggregation Getting a good forecast for sporadic combinations Compile non-high impact items from segmentation analysis Program forecast model to aggregate non-high impact items to logical planning source (plant, warehouse, prod. unit, etc.) Generate statistical forecast at aggregate planning source Disaggregate statistical forecast to lowest forecast level in model based on past history (plant/product/customer ship-to) Aggregation to the plant/product level reduced number of combinations to review by 80% Aggregation produces a more accurate forecast for sporadic items allowing more time to focus on high impact items
Six Keys to Improving Forecast Accuracy Step 5: Sales Adjustments to Statistical Forecast Step 5: Sales Adjustments to Statistical Forecast Accurate forecasting is not just getting a forecast from the customer. The customer isn t always right! Source: Agere Systems Inc.; Operations Management Roundtable Research Step 5: Sales Adjustments to Statistical Forecast Since statistical forecasting is based on previous sales, it is necessary to account for factors that can affect future sales Seasonality of the business State of the economy Competition and market position Product trends It s also important to ask customers the right sales questions to validate forecast assumptions Where does this project rank today? When are you looking to make a purchasing decision? When will you be implementing? What would you like to see happen as a next step?
Step 5: Sales Adjustments to Statistical Forecast Adjustments Made to Forecast Pre-buying Promotional impacts Upside and downside volumes Customer plants expected to be down New business New customers Adjustments Made to History Data errors Lost product volume One time customer outages Packaging changes Discontinued products Decide where adjustments should be made Six Keys to Improving Forecast Accuracy Step 6: Measurement & Exception Reporting Step 6: Measurement & Exception Reporting If you can not effectively measure forecasting performance, you can not identify whether changes to the process are improving forecast accuracy Effective measures should evaluate accuracy at different levels of aggregation (plant, warehouse, sales region, etc.) Measuring and reporting forecast accuracy helps to build confidence in the forecasting process Once people realize that sources of error are being eliminated, the organization will begin to use the forecast to drive business operations decisions
Step 6: Measurement & Exception Reporting A variety of analyses and exception-based measurements are needed to understand where the biggest sources of forecast error are Identifying High Impact exceptions for Sales review Accuracy of adjustments made to statistical forecast Identifying non-high Impact exceptions Current month variances (Forecast Actual) Forecast but no sales Sales but no forecast No sales in last 12 months High Impact Exceptions for Sales Review Identify High Impact Exceptions to focus forecast review Decreasing or increasing sales rates Aggressive statistical forecast based on historical run rates History S&OP Forecast AUG 2006 SEP Sold To Loc Product MAY 2006 JUN 2006 JUL 2006 JUL 2006 2006 Customer A MO Product 1 142,800 80,948 101,251 88,863 80,000 80,000 Customer B GA Product 2-20,276 - - 10,000 10,000 Customer C TX Product 3 101,577 101,378 81,284 82,009 86,990 87,060 Customer D AZ Product 4 60,899 81,166 39,208 60,000 60,000 60,000 Customer E TN Product 5 60,954 20,348 81,247 53,012 60,520 60,582 Customer F TX Product 6 141,131 121,971 141,321 120,000 140,000 120,000 Customer G NY Product 7 101,496 81,438 82,019 81,210 87,681 87,738 Customer H FL Product 8 100,761 120,302 100,788 103,993 104,148 104,239 Customer I GA Product 9 121,753 142,392 121,817 116,505 117,554 117,554 Customer J NJ Product 10 163,003 142,709 61,217 155,800 155,800 155,800 Accuracy of Forecast Adjustments (Units in KGS) Jun06 Stat Stat Fcst Jun06 S&OP S&OP Abs Ship-to Customer Jun06 Act Fcst Abs Error Final Error Impact +/- Customer A 2,167,904 1,465,968 701,936 1,691,155 476,749 225,187 Customer B 286,650 494,531 207,881 281,859 4,791 203,090 Customer C 316,315 454,387 138,072 313,809 2,506 135,566 Customer D 743,619 906,002 162,383 680,000 63,619 98,764 Customer E 20,266 0 20,266 50,000 29,734 (9,467) Customer F 244,023 370,466 126,443 205,698 38,325 88,117 Customer G 332,211 252,729 79,482 330,000 2,211 77,271 Customer H 20,312 40,547 20,236 113,400 93,088 (72,853) Customer I 50,000 130,416 80,416 80,000 30,000 50,416 Customer J 301,221 111,502 189,719 159,757 141,464 48,255 Customer K 40,479 56,471 15,992 102,000 61,521 (45,529) Customer L 142,709 84,879 57,830 155,800 13,091 44,739 Customer M 163,592 237,196 73,604 193,000 29,408 44,196 Customer N 0 1,603 1,603 40,000 40,000 (38,397) Customer O 40,615 0 40,615 37,800 2,815 37,800 Customer P 0 52,893 52,893 16,000 16,000 36,893 Customer Q 152,434 194,153 41,719 140,000 12,434 29,285 Statistical Forecast Final S&OP Forecast (Includes Forecast Adjustments) Impact +/-
Accuracy of Forecast Adjustments If adjustments to the statistical forecast do not improve forecast accuracy, why make them? Actual Demand Stat Forecast Stat Forecast Error (Units in KGS) S&OP Forecast S&OP Forecast Error Month Impact Jan06 6,091,955 7,593,962 5,196,370 5,918,751 1,886,262 54.3% Feb06 9,147,987 8,661,173 4,497,213 9,061,139 3,013,993 16.2% Mar06 11,570,962 11,932,441 3,992,114 12,448,817 2,885,691 9.6% Apr06 10,625,650 12,512,901 5,348,524 13,477,086 4,550,418 7.5% May06 10,815,034 9,840,902 3,791,501 11,297,905 3,059,782 6.8% Jun06 8,817,693 9,041,067 3,697,356 9,311,634 2,569,887 12.8% Last 6 Month Ave. 57,069,281 59,582,446 26,523,078 61,515,332 17,966,033 15.0% Accurate customer intelligence provided by Sales can significantly improve forecast accuracy Current Month Variance Analysis These items have the biggest overall impact to forecast accuracy results Determine the source of the variances Forecast Variances -- By Ship-to Customer JULY 2006 Product Customer JULY Actual JULY Forecast JULY Variance JULY Fcst Accuracy Product 1 Customer A 2,134,314 KG 1,611,990 KG 522,324 KG 67.6% Product 2 Customer B 1,230,867 KG 900,000 KG 330,867 KG 63.2% Product 3 Customer C 433,674 KG 147,542 KG 286,132 KG -93.9% Product 4 Customer D 0 KG 240,000 KG 240,000 KG 0.0% Product 5 Customer E 61,117 KG 300,000 KG 238,883 KG 0.0% Product 6 Customer F 165,799 KG 330,000 KG 164,201 KG 50.2% Product 7 Customer G 431,901 KG 268,665 KG 163,236 KG 39.2% Product 8 Customer H 424,536 KG 262,575 KG 161,961 KG 38.3% Product 9 Customer I 334,660 KG 174,096 KG 160,565 KG 7.8% Product 10 Customer J 20,040 KG 175,167 KG 155,127 KG 11.4% Product 11 Customer K 490,342 KG 355,456 KG 134,886 KG 62.1% Product 12 Customer L 40,642 KG 171,132 KG 130,490 KG 23.7% Product 13 Customer M 264,235 KG 143,679 KG 120,557 KG 16.1% Product 14 Customer N 181,589 KG 77,946 KG 103,644 KG -33.0% Forecast but No Sales Zero out the history for these customers in the forecast model Last 3 Mo. Stat Fcst Final Fcst Final Fcst Final Fcst Average Ave Fcst Next 3 Ave Fcst Next 3 Monthly 09/2006 10/2006 Mo. Mo. Demand Product Customer Ship-to Product 1 Customer A 0 0 125,000 125,000 125,000 Product 2 Customer B 0 15,694 80,000 123,000 101,333 Product 3 Customer C 0 0 87,000 87,000 87,000 Product 4 Customer D 0 0 72,000 72,000 72,000 Product 5 Customer E 0 34,619 34,485 34,676 34,619 Product 6 Customer F 0 0 1 20,000 20,000 Product 7 Customer G 0 18,047 18,047 18,047 18,047 Product 8 Customer H 0 17,293 17,282 17,298 17,293 Product 9 Customer I 0 10,654 10,624 10,667 10,654 Product 10 Customer J 0 9,763 9,744 9,772 9,763 Product 11 Customer K 0 8,223 8,223 8,223 8,223
Sales but No Forecast Add forecasts for these customers Act Sales Act Sales Act Sales Last 3 Mo. Stat Fcst Stat Fcst Final Fcst Final Fcst Average Monthly 05/2006 06/2006 07/2006 09/2006 10/2006 09/2006 10/2006 Demand Product Customer Ship-to Product 1 Customer A 0 0 84,550 28,183 0 0 0 0 Product 2 Customer B 0 0 61,117 20,372 0 0 0 0 Product 3 Customer C 29,402 20,330 0 16,577 0 0 0 0 Product 4 Customer D 13,789 12,093 20,339 15,407 0 0 0 0 Product 5 Customer E 0 0 40,760 13,587 0 0 0 0 Product 6 Customer F 0 0 40,587 13,529 0 0 0 0 Product 7 Customer G 5,700 11,400 22,800 13,300 0 0 0 0 Product 8 Customer H 0 0 39,635 13,212 0 0 0 0 Product 9 Customer I 0 0 39,608 13,203 0 0 0 0 Product 10 Customer J 0 0 35,150 11,717 0 0 0 0 Product 11 Customer K 654 0 34,382 11,679 0 0 0 0 Content WACKER Company Overview Why Forecast? Forecasting Challenges Six Keys to Improving Forecast Accuracy Process, People, & Tools Statistical Forecasting Forecasting Segmentation Data Aggregation Sales Adjustments to Statistical Forecast Measurement & Exception Reporting Forecasting Accuracy Results Conclusion % Forecast Accuracy 100 90 80 70 60 50 Forecast Accuracy Improvement Process, People, Statistical Forecasting + 7% (Product Mix Level) Forecast Segmentation Data Aggregation World Class Exception Analysis Forecast Adjustments 40 4Q03 1Q04 2Q04 3Q04 4Q04 1Q05 2Q05 3Q05 4Q05 1Q06 2Q06 3Q06 + 6% + 15%
Production Plan Adherence Supply Planning Accuracy 100 Production Plan Adherence (%) 90 80 70 60 38% Improvement 50 2004 2007 Financial Forecast Accuracy 100 Direct Profit Accuracy (%) 90 80 70 60 50 40 21% Improvement 30 2004 2007 Forecast Deviation Trend Forecast Deviation Forecast deviation at chemical product level APO Go-Live.5 APP Wacker.4.3.2.1 Target.0 Jul-07 Aug-07 Sep-07 Oct-07 Nov-07 Dec-07 Jan-08 Feb-08 Mar-08 Apr-08 May-08 Jun-08 Jul-08 Aug-08 Sep-08 Oct-08 Nov-08 Dec-08 Jan-09 Feb-09 Mar-09
Forecast Deviation Trend Forecast deviation at chemical product level 100% Forecast Deviation % 80% 60% 40% Forecast Deviation Reduced 56% 20% Target 0% Sep-08 Oct-08 Nov-08 Dec-08 Jan-09 Feb-09 Mar-09 Apr-09 Forecast Deviation Vs Inventory Levels 50% 60 Inventory Decreased 29% Forecast Deviation % 40% 30% 20% 10% 50 40 30 20 Inventory DOS 0% 10 Nov-08 Dec-08 Jan-09 Feb-09 Mar-09 Apr-09 Forecast Deviation Inventory DOS Forecast Deviation Vs On-Time Delivery 50% 100 Forecast Deviation % 40% 30% 20% 10% On-Time Delivery Increased 2% 98 96 94 92 0% Nov-08 Dec-08 Jan-09 Feb-09 Mar-09 Apr-09 On-Time Delivery % 90 LDU Forecast Deviation Calvert On-Time Delivery
Forecast Deviation Vs Revenue Accuracy Aligning demand plan with financial forecasting process 50% 100 Forecast Deviation % 40% 30% 20% 10% Revenue Accuracy Increased 29% 0% Nov-08 Dec-08 Jan-09 Feb-09 Mar-09 Apr-09 90 80 70 Revenue Accuracy % 60 Forecast Deviation Revenue Accuracy Content WACKER Company Overview Why Forecast? Forecasting Challenges Six Keys to Improving Forecast Accuracy Process, People, & Tools Statistical Forecasting Forecasting Segmentation Data Aggregation Sales Adjustments to Statistical Forecast Measurement & Exception Reporting Forecasting Accuracy Results Conclusion CONCLUSION If you follow the Roadmap to improve your company s sales forecasting practices, you will experience reductions in costs and increases in customer and employee satisfaction. Costs will decline in inventory levels, raw materials, production, and logistics. But the first step a company must take before realizing these kind of benefits is to recognize the importance of sales forecasting as a management function, and to be willing to commit the necessary resources to becoming world class.
The WACKER Group CREATING TOMORROW'S SOLUTIONS Stephen P. Crane Director Strategic Supply Chain Management stephen.crane@wacker.com THANK YOU FOR YOUR ATTENTION