External Scripting Use Case : Product Forecasting
The complexity of the forecasting can take many forms incorporating:
- daily patterns
- event driven variations such as promotional activity
- continuously variable data such as temperature, sunlight, social media trending
Regardless of the complexity of the forecasting model, there is often a need to be able to carry out the process over thousands of individually changing data sets. Traditionally the approach to this type of analytics is to run these forecasts by transferring the data sets out of the data store into the forecasting engine and push results back. Why not take the forecasting to the data? Having the ability to run complex external scripts in the language of your choice in a massively parallel environment means this complex, time consuming, (often manual) process can be removed.
Take the example of a retailer with 1,000s of products and 100s of stores. Forecasting sales volume could be:
- At product level to gauge the effects of promotional activity
- At product-store level if you are considering stock-outs
- For online sales with a whole different attribute set, such as product position on page etc.
All these sales forecasting requirements can be run from the same script with minor changes to the script interface and query that feeds the script.