Using data modeling to predict marketing’s effect on sales

We use data modeling to predict the effects of marketing activities before they happen, so we can test and apply the most successful strategic approach. In this case, data modeling results informed an updated social media plan for our client that increased expected sales by 7%.

The problem: As with many marketing teams, our client was struggling to make a direct connection between marketing activity and sales. They wanted to know what was working, what wasn’t, and most importantly, how their advertising and media spend would affect sales before they opened their wallets.

The solution: We used our Intelligence Platform to develop a predictive data model of expected sales for our client.


We looked at historic store-by-store sales, marketing and media spend for this client’s 400+ stores across the U.S. After assimilating, processing and analyzing all their historical data, we were able to project the future sales trajectory for each store. Among the projected data, we predicted a decrease of 5% in comp sales for the stores in a single large DMA in Texas.

Armed with valuable information from our data model, we conducted a test in this DMA using very different social media tactics than the client had ever implemented before. If this test was successful, it could be implemented in other stores to help boost sales. This is one way our CALLAHAN INTELLIGENCE PLATFORM is particularly useful for our clients: we can run tests of new media ideas or approaches and measure the results on a store- or market-level basis before rolling them out systemwide. And, we can phase the rollout, which creates a perpetual testing environment to continue to prove results and add to the model data over time. This further improves the model’s accuracy and enhances its predictions.

The aha:

This 2% increase in comp sales year over year took on new meaning given the projected 5% decrease in comp sales our model had predicted. So the campaign’s 2% increase was actually a 7% lift in predicted sales.

This metric wouldn’t have been observable, nor would its significance have been understood, without the predictive model.


The results:

When we measured the results of the social media test in that DMA, we saw an increase of 2% in comp sales versus the same period the year prior.


Services this client used:

If you’re interested in how data modeling could help you understand past brand performance, predict future performance and test new approaches, fill out the form below. We can’t wait to hear from you!

Contact us to learn how
Callahan can help