How to choose the most appropriate trending approach for a sales forecast
When you are creating a sales-based forecast there can be a wide range of sophisticated forecasting tools available to you.
How do you decide on the best methodology and how do you check the appropriateness? Kris Barker, senior consultant at J+D Forecasting discusses:
As a starting point, we would recommend using your preferred trending method for all the SKUs/ products/ regimens in your model. This may be a simple growth rate, linear regression, Holt-Winters etc. There are then a couple of checks we would suggest to review the ‘appropriateness’ of your preferred trending option for each product:
1) Review, where applicable, the R-squared value. If this is high, say 60% or above, then you can take comfort in the fact that the trending option selected is doing a reasonable job of using the historic data to predict the future. If this value is low, then this is the first indication that you may want to change your trending option
2) Ultimately, trending algorithms are only mathematical formulas and they cannot think or apply any market understanding that you as the forecaster will possess. As a result, we would certainly recommend reviewing the trending option for any product where the past is not likely to be a very accurate predictor of the future. For example, in cases where a product has only launched in the last 12 months or where a product has had stocking issues over the last 12 months and you know that the volumes are going to ramp up in the coming periods.
Trending algorithms are a very quick and easy way of creating your baseline forecast, before you then go on to overlay your events. But they cannot think, and the knowledge and expertise of the forecaster cannot and should not be underestimated when it comes to accurately trending the market in a Sales+ forecast model. But as a general rule of thumb, the screen shot below provides an overview of when each of the different trending options should be considered.