Utilizing rigorous data science techniques to answer difficult business questions.
FinancialServices logo 1
For a regional financial institution, we conducted an assessment of the data they had available, the scale and variability of their spend, and their primary outcome of interest, which was an auto loan application. In this case, which is, unfortunately, not uncommon, the limitations in the data necessitated that the most viable option was a multivariate analysis.
Search Discovery conducted a regression analysis using spend by channel by month and seasonal factors as the independent variables and the number of auto loan applications (same month, one month lagged, and two-month lagged) as the dependent variable. We used 10-fold cross-validation to reduce the risk of overfitting in the resulting model.

While this analysis returned a model that explained 65% of the month-to-month variation in the number of loan applications, it also revealed that seasonality was far and away the dominant driver of loan applications while the marketing spend potentially had long-term impact of driving brand awareness, the scale of that spend was not significantly (detectably) moving the loan applications from month to month.

The result was that we recommended the organization continue to use their existing last touch attribution where possible (~50% of their spend was in out-of-home and radio), focus on increasing the quality and completeness of the data they were collecting, and consider controlled experiments if they were planning to increase their marketing investment substantially.
Scroll to Top


Catch the latest industry trends we’re watching and get new insights from our thought leaders delivered directly to your inbox each month.