To that, Mike and John agreed, as it depends on the context. But they both concluded that adjusting the history would make more sense for Ethan. As John put it, “The history of what should have happened (barring Covid) is clearer to Ethan’s company than the unknown of what might happen in the future.” He added that forecast data is based on historical data anyway, so by adjusting history, Ethan’s setting himself up for a more accurate forecast as well. Plus, forecasts eventually become historical data, so if Ethan adjusts his forecast, he might have to make amends to it as time progresses.
Mike also suggested that, for some organizations in a situation similar to Ethan’s, they could extend the window of history they’re using for planning. This would help their algorithm accumulate more demand that’s in line with what they expect in post-covid times: Just make sure to take a straight average of these months without assigning extra weight to the COVID era, John advised. That said, Ethan will also have to monitor and adjust his history based on what actually transpires.
Forecasting Accuracy – Is a Forty-Percent Forecasting Accuracy Really Bad?
To this second question from a customer, John and Mike answered, “No, and it might not matter that much anyway–not in the Service Parts Planning environment, at least.”
They went on to explain that forty percent accuracy would indeed be horrible in a retail environment, for instance, but it’s common in Service Parts Planning. This is because, in most service parts environments, demand is dispersed incongruently across your array of service parts. For example, if you plan ten thousand service parts, eighty percent of that demand may fall on just two hundred of those parts, with twenty percent spread across the other 9,800 parts. With that kind of asymmetry and unpredictability, high forecasting accuracy in the service parts world just isn’t realistic, especially on the long tail of very slow moving parts. In fact, the best accuracy rate concerning Service Parts Planning that John’s ever heard of was in the sixty-percent range.
Luckily, both John and Mike agree that accurate forecasting in the service parts world is just a “nice to have.” Standard forecast accuracy measurements don’t do a good job of measuring service parts forecasting, and should be viewed in that light. While it does make sense to focus on accuracy for things like a sales forecast, other metrics, such as off-the-shelf fill rates are more apt for Service Parts Planning. At the end of the day, the primary goal for service organizations is to make sure they’re satisfying the customers.
Submit Your Own Question!
On that note, this concludes our summary of the second episode of Spare Talk. Have a question you’d like Mike and John to answer in a future episode? Send us an email at firstname.lastname@example.org!