Text-Based Predictive Analytics
The Brief: Other research has shown how customer experience is related to membership retention. This client was curious as to what parts of their customer experience are detracting from membership length versus driving a longer membership period. The client’s previous data sets had proven difficult to find any relationship between customer experience and membership duration. But surely something had to predict when people would leave. The question was how could we find data to predict membership duration.
The Research: Since previous data sets had proven to yield no results, Origin decided to focus on the open-ended answers of the survey. Using natural language processing, Origin took the open-ends, broke them into individual words and phrases, and used those words and phrases to predict membership length.
The Insights: The result was a significant text-based predictive analytics model. Our model was able to predict the approximate duration of an individual’s membership based on their open-ended response to a question about their recent experience with the brand.
The Implications: After the validation of the model, an application could be built to read members open-ended responses and “flag” those members who were at risk of leaving. Once the client knew who was at risk of leaving, they could take action (e.g., send a coupon) to increase the at risk customers’ duration and ultimately increase customer retention.