For years, Direct Marketers and Fundraisers have relied on Predictive Data Modeling to find new, responsive audiences to acquire customers. These prospect data models, built by analyzing the responders to prior campaigns, combines art and science to successfully identify those likely to respond to offers.
Typically, once a Predictive Data Model is built, validated, and tested thoroughly, there is minimal maintenance required. Sometimes a high-performing model can be used successfully for years. We have seen such longevity behind many Predictive Data Models because a simple, underlying assumption has held true – the profile of the responder to a given offer or product type has remained stable. Demographic attributes, motivations, buying history and lifestyle habits that characterize a typical responder have remained relatively constant, and therefore predictable.
Unfortunately, over the last few weeks our world has been anything but predictable, meaning that all assumptions made prior to the spread of COVID-19 may now have to be altered. What does that mean for your current data models? Here’s what we are telling clients.
Data Models in the Age of COVID-19
While we believe that the axiom of model building, described above, is still true in the age of COVID-19, the profile of a COVID-19 responder has changed for virtually all marketers. In many cases it has changed significantly enough to question the validity of any data models that were built prior to the pandemic. For many organizations, it may be too soon to confirm this assumption. However, there is anecdotal evidence that most companies are seeing differences in consumer spending habits and needs.
For example, many businesses needed to make significant changes to their order processing workflow to adapt to the COVID-19 environment. Curbside and “no-touch” order pickups are becoming a new norm, and industries are offering delivery services for the first time. While taxing on the businesses themselves, this new convenience for consumers could open your brand to an entirely new customer segment.
This one change alone in how consumers interact with your brand may contribute to the change in customer profile. Think of all of the other factors! As a result of COVID-19, many organizations are seeing an influx of new customers that would not have matched their previous customer profile. Inversely, some organizations are seeing a decline in activity as people prioritize and alter their buying habits accordingly.
Ask us how LiftEngine can help you build your model today!
What Does This All Mean?
It is impossible to know if this is the new normal, or if things snap right back once COVID-19 becomes less of an impact on our lives. What we do know is that all marketers are living in an unstable environment, and they should expect to adopt their business and marketing processes as the environment changes.
From a modeling standpoint, this means that you should no longer assume that models that have been working and providing stable results in the past will continue to do so. Therefore, we have a few recommendations for marketers that rely on models for customer acquisition.
1. Pay Attention to Results
Continue to monitor the performance of existing models closely. If you are seeing declining performance, you should rebuild them with an emphasis on purchases made after March 1, 2020.
2. Update Your Data Sources
Make sure the third-party data you use to build your models is as fresh as possible. Using the most current data available will allow the modeler to more accurately define the profile of a current customer and should provide better results in the near term.
3. Consider Multi-Channel Data Models
If you are a retailer with the majority of sales in-store, clearly times have changed. Consider building models of customers who bought instore AND online to identify the profile of these multi-channel buyers. Apply this model to those who have only bought at retail and you might find a nice pocket of potential online buyers.
4. Re-Consider Your Modeling Technique
In Data Modeling, one of the preferred techniques is known as Mailed Regression. These models are built by analyzing the audience contained in prior mail files. While great to leverage in times of stability, these may not be your best choice right now. Why? The audience contained in prior mail files, which were used to build the model, was likely derived from a customer profile that is vastly different than it is today.
Look-Alike models may be the better alternative since they cast a wider net and are less sensitive to changes in an organization’s customer profile.
5. Embrace the Instability
Rebuild and test models more frequently, perhaps every four to six months. This will add work, but we are in unprecedented times and we simply don’t have consistency and stability in the marketplace.
As society evolves in a COVID-19 world and beyond, the only consistent thing we can count on is instability. Just as some organizations needed to adapt their product and service delivery to meet customer expectations, marketers must adapt as well.
Published on Apr. 09, 2020, Last Updated on Nov. 08, 2023