Statistical Sales Forecasting using SAP BPC Capgemini s unique statistical sales forecasting solution integrated with SAP BPC 10.0 helps global fortune 1000 company built robust & accurate sales forecasting model 1. Executive Summary Budgeting, Planning & Forecasting is not new to companies. A company must have a robust & effective process to avoid any mismatch in strategic initiatives, capital allocation, inventory management, revenue guidance etc. But even though companies engage in different types of planning for sales, operations, expenses, HR, finance etcm all of these forecasting processes rely on historical data to come up with forecast numbers, assuming external factors don t change. However, there are many limitations in this approach. First and foremost being the assumption that external factors do not change. In today s global business environment, external factors play a key role, and planners need to consider the impact of external factors while generating forecast numbers. In this article, we discuss: 1. A new approach to sales forecasting the external market-driven statistical sales forecasting solution. 2. The reason why we believe this statistical sales forecasting solution the most accurate and ideal way of forecasting. In addition, we will explain how this model can be incorporated with existing ERP infrastructure (SAP BPC), and we discuss the business benefits of implementing such a solution. In Summary, Capgemini s unique statistical sales forecasting method brings following business benefits Accurate sales prediction Better Inventory management Better capital allocation Better insights about factors influencing sales Accurate earning guidance Reduced budgeting & forecasting cycle time Integrated with other planning models
2. Challenge Our client is a Fortune 1000 company which has a global presence across multiple end-market industries. The Client s current sales forecasting process was based on a large number of offline spreadsheets. Planners used to download the actual data from SAP and manipulate the forecast numbers on a case-by-case basis, often projecting an average of historical data. Multiple iterations were done to error out any discrepancies. The whole process was time consuming, error prone and required a lot of manual effort. Keeping the above problem in sight, Client wanted to implement a robust product & region-level sales forecast model based on external factors which are likely to influence sales for a particular region or a particular product family. In addition, client wanted see the correlation between the external factors and which factor(s) has the most influence for a particular product in a particular region. Finally, client wanted to automate the whole process by integrating it with existing ERP. 3. Solution Keeping above requirements in our sights, we developed a BPC-based statistical sales forecasting solution. The solution considers external factors and uses multivariable regression analysis to correlate these external factors. Based on the regression formula, sales forecast numbers are generated. The whole process is implemented in SAP BPC NW 10.0 and integrated with existing SAP ERP system. Multiple reports and dashboards with drill down functionality were developed to have better insights. Figure 1: Solution Mapping Having talked about how we addressed each of the requirements, let s talk about the solution in detail.
4. What is Statistical Sales Forecasting? The statistical forecasting solution is different from the traditional forecasting approach which uses historical data for forecasting. In addition to historical data, the statistical forecasting approach uses external economic indicators like GDP, total industry output, disposable income, etc to predict future sales. Using this statistical method for predicting sales not only gives an accurate prediction but also gives insights to business about what trends can help them increase their sales in that particular region. Capgemini s unique methodology helps business to build robust business model integrated with existing SAP Infrastructure 5. How it is Implemented? Capgemini has developed a unique and proprietary method for implementing this solution. As shown below, the methodology consists of 4 steps: Figure 2: Implementation Process Understanding Requirement The first step is to interview the stakeholders to understand the business requirements Research Economic Indicators The second step is to analyze the economic indicators. Capgemini assists clients in selecting the data that is the right fit for their business model and the forecast accuracy needs Run Multivariable Regressions Statistical regression analysis is the exercise of analyzing the fit of a time series of dependent (sales) and independent (economic indicator) variables to a linear historical pattern. Regression analysis can measure the correlation between a dependant variable (sales) w.r.t to a number of independent variables (economic factors).
For instance, in below example (sales) is dependant variable while X 1, X 2 etc are independent variables (economic indicators). E( Y X ) o 1 * X1 2 * X 2 3 * X3... Figure 3: Regression The fit of an equation can be measured using adjusted R-squared (R 2 ). R 2 provides a measure of how well observed outcomes fit to a multivariable regression formula, and it is equal to the proportion of total variation of outcomes (residuals) which are explained by the regression formulas vs. taking a simple average of the historical data. Between two regression equations with the same dependent variable and different independent variables, the equation with higher Adjusted R-Squared has a better fit. However, R 2 in itself is not the only goal of a good-fit regression for the predictive modeling tool. The first lesson one learns in Statistics 101 is that correlation does not necessarily imply causation. R 2 can be manipulated by adding more explanatory variables, which can improve the appearance of historical fit while having no effect on the model s predictive quality. Therefore, managers interested in bringing a data-driven approach to their firms should always ask for verifying evidence that data analysis results are true in the real world. A rigorous statistical regression methodology needs to be customtailored to a firm s unique business profile. Capgemini differentiates its statistical regression analysis approach by applying modern statistical methods to siphon out noise and identify the most accurate economic relationships. Capgemini recommends repeating this regression exercise quarterly to reflect changing economic conditions.
Integration with SAP BPC The last step is to automate the whole process by integrating it with existing ERP solution. SAP BPC NW 10.0 is one of the most prominent budgeting & forecasting tool available. Once the model is developed, it can be easily with SAP BPC. Economic Changes Quarterly Model Tune-Up Integrated with SAP-BPC Figure 4: BPC Design
Ref. Steps Technical Solution 08.10.05 Load economic Indicators into Statistical accounts were created in Account BPC Dimension to store economic indicators in BPC 08.20.00/ 08.20.10 Load BW Actual into BPC A separate Z*BPC cube was created with 1:1 map with BPC model. Using ETL, Actual from ECC were loaded into BI and formatted based on BPC model 08.20.20 Generate Forecast Sales Regression formulae were created using excel. BPC input schedule workbook was used to save the sales data into BPC 08.20.25 Granular Level Data For reporting, sales data were allocated using BPC allocation engine into more granular level. Allocation were made based on last year allocation % values 08.20.30 GM% GM% was entered manually by business users. A BPC input schedule was designed with using statistical account to store allocation % values 08.30.35 COGS Calculation Cost of goods sold was calculated using SAP BPC script. The formula was developed in BPC to generate the COGS 08.20.40 Reports Variance reports like Actual Vs Forecast were developed using BPC EPM add in 08.20.50 Push to other Models Finally, SAP BPC script was used to push data from BPC Forecast Model to other models like Operations Table 1: BPC Process Steps
Reports & Dashboards Reports and dashboards are last but key element of the methodology. Multiple reports were developed to give compare Actual Vs Predicted sales number. In addition, dashboards were created for top management to have better insights about the product. EPM Reports & Dashboards gives better insights Variance Report: Actual Vs Plan( Region wise) Figure 5: Variance Report Waterfall Report : Variance between old method Vs Statistical Sales Forecast Method Figure 6: Waterfall Report
6. Why SAP BPC based Statistical Sales Forecasting? Given the intertwined nature of the global economy, external factors influence sales of a particular product in a particular region. BPC-based statistical forecasting relying solely on historical sales to predict forecast numbers will often prove wrong. Best practice should be for planners to consider external economic factors that influence sales, though it is difficult to gauge the impacts of disparate economic factors on a company s business. Therefore, developing a model which considers both historical data as well as external economic factors will generate more accurate sales numbers, and automating and integrating this model into SAP-BPC saves management time and effort. Let s look at some of the benefits and their impacts on business. Solution Benefits Accurate sales prediction Better Inventory management Better capital allocation Better insights about factors influencing sales Accurate earning guidance Automated short term ( 1 year) and long term ( 5 year) forecast process Reduced budgeting & forecasting cycle time Integrated with other planning models like operational planning Dashboard with variance analysis Business Impact More Informed Corporate Financial Decisions Better External Guidance Reduced time for budgeting & forecasting Reduced total cost of ownership Better insights
7. Capgemini s offering Capgemini s unique offering comprises of leveraging the expertise of Capgemini Consulting to develop the sales forecast model and Capgemini s EPM practice to implement the sales forecast model into SAP BPC NW 10.0. With both teams working together to leverage their strengths, this unique service offering has delivered tangible results for our clients that have been received with astonishing success. The current project was implemented within a timeframe of 12 months. Based on its success, we are working on implementing a similar model for operational planning for new clients. Figure 7: Cagemini s offering 8. Going Forward As we observed, the statistical based sales forecasting solution is much more accurate and efficient for business than historical forecasting alone. This model is not restricted to a single industry. It can be expanded to all the sectors and business models. As a technology company, we are looking to expand the scope of the model to offer similar solutions across all planning processes that may include operations planning, financial planning, HR etc. In addition, we are also looking at SAP s predictive analytics and open source R to improve the efficiency of the process.
9. Authors Gleb Drobkov, Senior Consultant Gleb.Drobkov@capgemini.com Pratyush Panda, Senior Consultant Business pratyush.panda@capgemini.com & Technology Innovation, North America