Challenge
A European market leader in online survey and feedback software acquired complementary companies in different Wester European countries, each of which had its own survey platform.
This disparity became problematic when data showed sales performance was inconsistent across platforms. One survey product grew to 50% of the companies overall revenue over five years, while the other product was shrinking due to customer churn at a rate of 15% per year.
The software company needed a fresh approach.
Using data analytics, they sought to determine the reason for the customer loss, find ways to stop the bleeding, and renew product growth through enhancements that align with customer need, not guesswork.
Solution
Partnering with EastBanc Technologies, the company sought to disciver insights in the sea of product performance and consumer behavior data that the company had collected. We employed a Minimal Viable Prediction (MVP) approach. An MVP is an outcome that addresses a primary problem and is the starting point for any journey to predictive analytics - whether it’s predicting consumer behavior, future revenues, etc. in the fastest way possible.
To get quick, actionable results, we identified low hanging fruit (data that was both the easiest to analyze, and most likely to lead to valuable insights). We focused on conservative and long-established events to study user behavior patterns. Looking at user behavior for a specific survey product, our goal was to identify how user activity changed over time, which product features had been used and for how long, and which features or combination of features might be used as predictors of churn.
To boost the predictive power of data, we analyzed absolute and relative user behavior. These perspectives provide unique insights into seasonal factors, product learning curve, etc.
To empower non technical users the data was presented with intuitive descriptive statistics and visualizations. Dynamic modeling and unsupervised learning was used to explore and cluster user behavior patterns, while supervised machine learning was applied to predict customer churn based on system usage logs and CRM and data insights gathered from exploration.
Our initial findings showed that the amount of product usage was the most important predictor of customer churn. But we dug deeper. Was it viable that if a customer uses a product X times a month or less, that they would abandon the product at contract end? If proved correct, the first actionable insight, or MVP, is realized and action can be taken. If a change in usage patterns occurs, the software company would have X months to act and re-engage the user (through marketing, training, etc.).
Furthermore, distinct user groups were identified each with discrete behaviors suggesting that each group should be handled differently.
Business Impact
Using a combination of agile processes, scientific insights, and a strong technology foundation, our method was unique. With EastBanc Technologies’ help, the organization could develop a realizable goal to reduce churn from 15% to 10% within 12 months.
Future enhancements include the analysis of data usage patterns to inform product roadmaps and enhance consumer engagement, drive product improvements, improve perceived product value, measure marketing campaign effectiveness, and more. In addition, deep learning and data mining (intelligent assistant) will be applied to speed up workflows and empower users.
Technologies Used
Microsoft Power BI, Azure Machine Learning Studio, Apache Spark, Google TensorFlow