Historical data can be used in conjunction with predictive methods to populate a forecast

Description

In Corporate Performance Management (CPM) software, leveraging historical data with predictive analysis techniques is an optimal way to generate a forecast. This method uses a blend of existing data and advanced algorithms to project future trends, patterns, and results. These methods vary in technique, with the leading systems offering multiple methods based on the trends and patterns in the historical data being used.

Example Use Case

Scenario: A retail company implements CPM software to improve their inventory management. They have 5 years' worth of sales data that accurately reflects their business operations over time, seasonality, and market trends. Solution: The CPM software uses the historical sales data in conjunction with predictive analysis methods to generate an efficient inventory forecast. For instance, the system uses the past sales of winter clothing to accurately predict the amount of stock needed for the upcoming winter season, thus reducing overstock and shortages.

Considerations

Predictive analytics, sometimes called predictive methods, can produce a forecast without the use of cumbersome drivers or manual inputs. For established businesses with a consistent pattern of sales, these methods can produce an accurate forecast in seconds. The best systems will automatically apply different statistics methods like ARIMA, exponential smoothing, moving averages, and so on based on the historical data that is has been fed. Better yet, if the system allows you to pick and put the results of different methods side by side, that is icing on the cake.

A best practice is to blend predictive analytic-driven forecasts with a more traditional driver or bottoms up approach. For example, one version of the budget was entirely generated via different statistical methods. The other version was created using drivers and other KPIs. They are then placed side by side to detect significant variations in the techniques. Those variations are red flags that should be investigated by the analyst.

Questions to Ask a Vendor

• Seasonality and Trends: Business operations often have seasonal patterns and trends. Does the system take this into consideration when applying the method to our forecast?
• Method Customization: Can the choice of predictive methods be customized according to our specific needs?
• Versions: Can the methods populate an entire forecast version at once, or just one account at a time?
• Time Granularity: At what detail will the predictive methods function with our historical data? We have information down to the day. Can it process that?