Sales Forecasting Model

Predictive Sales Model For Major Chain

The situation: Our client, a retail chain that operates stores in three western states, wanted to continue its expansion program at a rate of 20 to 30 stores per year.

The issue: Over the past two years, the chain had been forced either to close or to operate at a lower ROI a number of stores it had originally opened based on the sales predictions of its model.

The problem: Management had no difficulty finding plenty of potential new sites, but did not have adequate tools for determining which of these sites would be profitable. The chain could not predict a site’s potential profitability with sufficient accuracy because the range of error associated with its sales predictions was too great. Neither could it accurately assess the risk associated with any site because it had no way of determining the probability that a given site would achieve a range of sales volumes.

The solution: DeForest & Company recognized that the information the client needed could not be generated by the client’s model. That model employed traditional linear regression, which does not yield sufficiently specific results when confronted with situations such as this, and analog methodology, which is far too subjective. To satisfy the client’s need, we built a custom predictive model that incorporates our state-of-the-art proprietary methodology. This model was more accurate than linear regression and did not incorporate the subjectivity of the analog process. Not only did our model lower the range of error associated with sales predictions, it also provided the client with the probability that the site would achieve them.

We have applied our model to over 100 client sites in the last three years and to date, all of the stores opened have performed within an acceptable range.