3P Learning

Case Study | 3P Learning

Helping a Global Leader in Online Education Improve Customer Retention With Machine Learning

The Project

3P Learning is a publicly listed ASX organisation and a global leader in online education. 3P’s suite of learning resources is designed for schools and families, covering mathematics, spelling and literacy. 3P has tens of thousands of schools as subscribing customers and millions of students and teachers as users of its products across the world. In light of its strategic objectives and increased competitive environment, 3P’s leadership team decided to allocate resources in data science to tackle a variety of business challenges. In a combined effort with Polynomial Solutions, 3P engaged with Felipe Rego to specifically develop and enhance a customer churn model using machine learning.

Objectives

Identify customers likely to leave before they do

3P’s customer retention rates were already considerably high, but management wanted to make sure its market share position was sustained in the long-run. As part of a larger data and analytics strategy, 3P’s customer retention analytics project focused on:

• Identifying customers that were unlikely to renew their subscription before the due date;
• Providing an easy-to-understand solution for sales teams to adopt and embed in their retention strategies;

Methodology

Aligning industry domain knowledge, cutting edge technology and a robust solution

The fact that 3P’s customer churn was historically low meant that applying predictive customer churn models was a harder-than-usual task. Additionally, there were many different source systems in which data had to be used. For it to be successful, the approach had to consider:

• Individual markets and geographies to cater for regional nuances and specific educational domain knowledge;
• A set of unknowns that could not realistically be captured in the dataset and model results;
• The class imbalance of historically low churn plus the possibility of a non-parametric nature of the dataset;
• The relative cost of misclassification;

Over 30 different families of statistical models were tried and tested with hundreds of parameters optimised. The final solution incorporated an ensemble learning method with a multitude of decision trees yielding good model performance results.

Results

Improving customer retention with high ROI

Over a few weeks, Felipe Rego helped 3P Learning define, develop and deploy a robust customer retention solution using cutting-edge machine learning with an estimated ROI of c.250% in the first year alone.

"Felipe was recommended to us by Anna Russell who is the Director of Polynomial Solutions. We needed someone who could come in and help define and build a cutting-edge analytics solution for customer retention as part of a larger analytics investment. We at 3P have been laser-focused on our customers and further utilising advanced analytics for that was an easy decision. I was impressed by Felipe’s self-directed approach. Differently from other data science professionals we worked with, Felipe was really good at maintaining constant communication with the executive team and stakeholders. He also made sure that the results were easily interpretable for non-technical folks such as product, sales and design teams. But the most valuable thing for me was that he aligned technical solution to commercial benefits. At the end of the engagement, we had a clearer picture of the potential benefits of the solution he built which made it easier to adopt."- Simon Perry, Chief Information Officer, 3P Learning

• Get in touch