Client Use Cases: 1-800 CONTACTS Identifies Drivers of Churn
Updated: Sep 16, 2019
The following is the first post in a five-part series on how 1-800 CONTACTS uses Compellon20|20 to drive change within their organization. Thank you to Neil Wieloch, PhD, Director, Marketing Strategy and Insights, for writing these posts and for sharing 1-800 CONTACTS’ successes with us.
It’s not me – it’s you: Why do customers who have had a great experience not return?
Since 2005, 1-800 CONTACTS had been looking to identify the drivers of customer churn. On two different occasions, they hired statisticians to build models – and both times they failed. In both instances, a blend of customer transactional behavior data and data from satisfaction surveys were used.
In 2016, in an attempt to once again see if they could predict customer reorder rates, 1-800 CONTACTS worked with Compellon. And in this case, they were successful. Running the 1-800 CONTACTS data through the Compellon predictive engine revealed two incredible insights:
1-800 CONTACTS customer order behavior could in fact be predicted, and
The reason customers returned had very little to do with what the customer purchased or how satisfied they were with the experience – and had almost everything to do with who the customer was (demographically & psychographically).
One of the reasons Compellon was so successful in identifying drivers of churn when data scientists were not is due to Compellon’s “kitchen sink” approach to analysis. Because the Compellon20|20 platform is able to identify relationships between variables without any apriori modeling, analysts can dump in any and all data and let the engine reveal the drivers. This is an amazingly different and powerful approach to model building.
The reason this approach is unique is that, unlike the traditional approach, the Compellon methodology does not require that the analyst first build the model to be tested. In the case of 1-800 CONTACTS, the initial and unsuccessful models that were built by data scientists were constructed based on the assumption that purchase and satisfaction were the drivers of re-order – so those were the only variables from which the predictive models were built. But, as the Compellon analysis revealed, those were false assumptions. In contrast, Compellon’s unbiased approach encourages the use of as much data as possible in the analysis. In the case of 1-800 CONTACTS, the otherwise discarded data (demographic and psychographic) turned out to be the primary predictors.
1-800 CONTACTS now has a very clear picture of:
who to prospect,
who to incentivize, and
how many of each segment to target in order to reach a definitive final business objective.