Contact Us
Choose a Category

Establishing true incrementality of your marketing efforts using predictive analytics

Dev Ganesan President and CEO

Here  is a common scenario faced by restaurant marketers. They have developed a fantastic offer that is increasing traffic and net sales, but are not sure about true incrementality.  Questions often asked included:  Are we rewarding our loyal customers who would visit us anyway? Are we just pulling forward or cannibalizing visits from a future date?   Is the offer driving future sales? And, these are tough questions to answer. Most of the measurement tools measure sales attribution i.e. we send  1,000 offers and those offers drove $100,000 in sales, but they do not get to true incrementality.

Take a look below  – measuring the impact of a $10 off $30 offer. The chart shows the visits of two groups of customers to a casual dine restaurant.The test group (in blue) received the offer and control group (in red) did not. The black line shows when the offer was received by the test group.Look how closely the two groups match each other in visit frequency before the offer and how they differ once the offer is received. While tradition says you need to establish a hold-out group prior to launch of any campaign in order to measure these kinds of changes in behavior, we have leveraged a sophisticated statistical algorithm that allows us to identify naturally occurring test and controls, regardless of pre-formed hold out groups.  It establishes that we have the right control group that looks exactly like the group that receives the offer after the campaign is in flight. UpliftIn this case we found the promotion lead to a sharp increase in Visits (21%) and Gross Sales (30%), generating significant lift above the control. Also because this was incremental sales lift it made a strong case for rerunning this offer, much stronger than just using the sales attribution approach.

Another nice thing about doing statistical matching instead of doing a holdout group is that you do not have to stop sending offers to any of your customers. For the situation below you can test an offer after it has already been in market or analyze historical offers. One key element of success is ensuring we have some way of connecting the customer to their transactions.  In this case we utilized payment and POS data.  But other means may include loyalty program transactions – both traditional or mobile app, dining reservations or online ordering.   And since we have access to this kind of data from multiple sources we can apply this learning to your full marketing channel like eClub, SMS or Loyalty program. In the next series of blogs I will share further examples of how we, at Fishbowl, are using our Guest Analytics platform and data science to measure incremental sales grounded in statistical analysis and predictive modeling.