Several years ago, I was visiting an old colleague who works for a large technology company. Over dinner, our conversation slowly turned to work and he asked me about data modeling and outliers. Apparently, his team views data outliers as noise that should be removed from all models to increase confidence and predictability. For some outliers, that is the correct process. But, pruning every outlier can severely limit your potential for growth. Sometimes, an outlier is an opportunity.
Opportunities masquerading as outliers can take many forms. In the database, they could be customer attributes that don’t match the target profile or product sets that don’t correlate to the standard mix. Outside of the database, they could be an engagement point that differs from the current conversion path or a new media channel. At the surface, these outliers may seem interesting, but not worth the effort to pursue. However, spending a little time and digging a bit deeper could yield quite a bit of gold.
A while ago, I was helping a large retailer with their loyalty program. The member base was mostly moms who were purchasing apparel for their children. Via profile and purchase analysis, an outlier appeared in the transaction mix: during major traffic periods, like Back to School, moms occasionally purchased something for themselves. The retailer knew about this behavior and attempted to capitalize on it by including a few additional product shots of apparel for moms in each mailing.
After a bit of convincing, we tested a different approach. We pulled the best target and tested a separate, additional communication stream focusing on moms. The ROI was similar during key traffic periods. But, outside of the key periods, the ROI skyrocketed. Incremental store and online visits increased from three times per year to five times per year. These additional visits drove stronger ROIs per piece and dramatically increased lifetime values.
The biggest hurdle with outliers is not the identification, but the justification of pursuing them as opportunities. Once an outlier is found, we use primary and secondary research to add qualitative insight to the outlier. Via the combination of quantitative and qualitative insights, we are able to determine the value of the outlier and propose a plan of action. This plan usually contains a test-and-learn process along with a detailed ROI expectation generated via our proprietary calculator. This calculator combines objectives, KPIs, target quantities, revenue and cost expectations based on several what-if scenarios. We like to call it our Return on Dare calculator. You can call it your justification for pursuing a different path.