The Challenge
The client, a subscription-based news aggregator, wanted to focus retention efforts on their highest-value subscribers to minimise churn and grow lifetime value across the existing base.
While the company maintained a solid data infrastructure, they lacked a reliable mechanism to spot exactly which subscribers were drifting towards cancellation. Consequently, retention campaigns remained broad and untargeted, spending just as much on highly active users with no intention of leaving as on those genuinely at risk.
A company-wide cost-saving drive put acquisition budgets under pressure, making revenue protection an urgent priority. The business needed to isolate likely cancellations before they occurred and allocate their defensive marketing budget effectively.
The Solution
An audit of the client’s customer data confirmed that subscription histories, marketing touchpoints, and digital consumption patterns were clean and linkable at the individual subscriber level.
We then built a propensity model to score each subscriber’s likelihood of churning based on recent platform interactions. Content consumption frequency, overall time spent on platform, and interaction with premium features emerged as the strongest predictors of cancellation, carrying significantly more weight than historical tenure.
Once validated, the scores were loaded into the client’s CRM and reporting platforms, feeding an evidence-based ranking that allowed the retention team to prioritize outreach within their existing day-to-day workflows.
The Impact
Concentrating outreach on high-value, at-risk subscribers lifted retention rates by 20% within the first two quarters, immediately cutting the campaign budget previously wasted on safe accounts.
The churn scores also revealed an unexpected avenue for growth, identifying highly engaged, low-risk subscribers who were prime candidates for premium tiers. Redirecting targeted tier offers toward this cohort drove 15% increase in upgrades, contributing more than £500k in additional EBITDA.
The predictive model now operates on a standing quarterly refresh cycle, ensuring risk scores adapt naturally as content features and pricing structures evolve.
Key Takeaways
- Churn Propensity Modelling - Scored every subscriber on their likelihood to cancel, replacing broad campaigns with focused effort.
- Prioritised Retention - Concentrating outreach on high-value, at-risk subscribers lifted retention by 20% and cut wasted contact spend.
- CRM Integration - Deployed scores directly into existing platforms so the retention team could act on insights without changing how they worked.
- Business Results - Secured a 15% increase in plan upgrades and over £500k in additional EBITDA from the existing subscriber base.
Tools and Techniques
- Propensity modelling
- Exploratory data analysis - profiling subscription history and digital consumption metrics
- Random Forest classification - executing the predictive machine learning model to calculate risk probabilities
- SQL - data extraction, cleansing and user-level record transformation
- R - data handling and predictive model training
- Salesforce - operational CRM integration and campaign activation
Related services
No matching items