brandon mathias blog

Predictive models show cost-effective results

The global shutdown caused by the covid-19 pandemic has put significant strain on American businesses. As marketers, we can’t help but notice the effect the pandemic has on the marketing plans across a wide range of industries. Even companies fortunate enough to be well positioned for such a widespread emergency (such as food services) have had to dial back marketing efforts because they have reached capacity to fulfill orders. The majority, however, are reducing or stopping marketing spend entirely to cope with other more immediate needs caused by the shutdown.

There are reasons for optimism, however, as some early affected countries show signs of recovery. Companies in the U.S., if not already started, will soon make their own plans for returning to normal operations.  Marketers, too, in these companies will begin to think about how they restart their programs and reengage their customers. I think this is a perfect opportunity to revisit these programs and ensure that they have the best chance for success.

As a model developer at Speedeon, I naturally think a good starting place for improving marketing programs is – surprise! – to build a model. Programs built on “business rule” selects or some other manual method for determining who to target almost always lead to a suboptimal solution. Using a predictive model, trained on examples of historical customer data, can yield better results for less money spent. If budgets are tight during the first phases of the recovery, using models for marketing programs can be a great way to show cost-effective results.