The rollover technique

- A simple, but effective, sales forecasting technique is what I call the rollover forecast. It's easy to use because the technique does all the work for you. No calculations, no guesswork. Here's how a rollover forecast works. The forecast for the next time period is simply the actual sales results from the last time period. Let me say it again using different wording. To forecast sales, you take the actual sales results from one time period, say last month, and make that number your forecast for next month. Now, I know what you're thinking. How could such a simple approach be accurate? Well, as with any forecasting technique, this one will have a certain level of error, but you'd be surprised how this very simple approach can be effective for many business models. If you're in a business that is fairly stable with little or no seasonality, this technique might be perfect for you. But here's what I really like about this forecasting model, regardless of how accurate it is, the results from it can serve as a benchmark for you to compare other forecasting techniques. Hey, if you find a technique that produces less error, meaning it's less wrong on average, you should use that technique, and if you're considering a technique that produces a more error on average, well, forget about it. You'd be better off using the rollover approach. Let's look at how to calculate the error between actual and forecasted sales. You can follow along using the 03_02 exercise file. Now, in column C you'll find actual results for 36 months. In column D, you'll see the rollover forecasted results, which is the previous month's actual results. For example, you can see that the rollover forecast for February is the same as it was in January, 4,398. Now to determine the absolute error I'm going to subtract rollover forecast from actual results. I'll use the absolute value function in Excel because we don't care if the difference is positive or negative. Now what you have here is the distance between the actual results and the forecasted result for the period. I can drag the cell down to quickly calculate the rest of the period's errors. Finally, I want to find the average error for the rollover forecast. To do this, I'm going to find the average of all the period's absolute errors. I'll use the average function and select all the absolute errors. Now, it turns out the rounded average monthly error is 1,335 units. Hey, pretty good. Depending on what your business model can tolerate. Let's keep this number in mind. As we evaluate other forecasting techniques to see if we can improve it.