What makes me an analyst and not just a prognosticator? I verify that my predictions can be trusted—that the number of bananas in my forecast for next month, for example, is a number that can be used to positively influence business decisions. But due to the nature of forecasting, we’re forced to wait for the future to evaluate my present forecasts. So how can we validate the effectiveness of current forecasting methods? It’s a simple matter of traveling back in time.
From a data input standpoint, backward time travel is simple: You choose the date you want to travel to and ignore any data you collected after that date. If we ignore that data when we create forecasts, we simulate what we would have forecasted on the date chosen, with the added benefit of knowing the future.
We can use our sample data to validate the accuracy of forecasts made with current forecasting methods. We can choose another date to travel back to and run another forecast and again, we have actual sales to use to evaluate that forecast, for example. We can repeat this process as many times as we can find new dates to travel back to.
In practice, we analysts will repeat this evaluation process over a 52-week period. That way, we can analyze our forecasting methods across time to see if there are weak spots in time, geography and/or products.
We can see if certain products are harder to forecast at certain times of year, or if a specific geography experienced a dramatic shift in behavior. We can brainstorm solutions to these concerns, adjust our methods, and re-run the 52 weeks of sample validations. Through this process, we can be highly confident that the predictions we provide to clients will have a positive impact on their bottom line.
This is just one of the many examples of our constant improvement process. Every day, we’re making our forecasts better—more accurate, more flexible and more encompassing.