SaaS Subscription Intelligence Platform
45-day churn early warning with 89% accuracy — fires early enough for CS teams to actually intervene
Overview
Multi-tenant SaaS analytics platform with AI-powered churn prediction, expansion revenue identification, and product usage intelligence. Deployed for 6 SaaS companies ranging from seed-stage to Series B.
The Challenge
Every SaaS client was losing customers they could have saved — not because they lacked data, but because their data was fragmented across product analytics, CRM, and billing systems with no unified intelligence layer. Churn was only visible after the cancellation. Expansion revenue was left on the table because no one knew which free-tier users were ready to convert.
What We Built
A unified intelligence platform that ingests event data from any SaaS product (via JS snippet or API), normalises it, and produces four outputs: a product usage dashboard (feature adoption, session patterns, activation rates), a churn risk feed (ranked list of at-risk accounts with reason codes), an expansion opportunity feed (free-tier users ranked by conversion probability and predicted deal size), and a 90-day MRR forecast with confidence intervals.
Tech Stack
Key Outcome
89% churn prediction accuracy · 6 SaaS clients
The AI Layer
The churn model is trained per-client on their historical subscription data — learning the specific usage patterns that precede cancellation for their product and customer segment. Common features include login frequency decline, feature abandonment sequences, support ticket sentiment, and NPS response patterns. The model fires 45 days before renewal for accounts showing early churn signals — giving CS teams a full engagement cycle to intervene. The expansion model identifies free-tier users whose usage patterns match the early-stage behaviour of the client's most successful paid conversions — ranked by predicted deal size so CS can prioritise outreach.
Results
- •Churn prediction accuracy: 89% at 45-day advance warning
- •Expansion revenue identified: average $42K per client in first quarter
- •Deployed for 6 SaaS companies (seed to Series B)
- •MRR forecast accuracy: within 4% over 90-day horizon
- •Delivered in 28 days per client (multi-tenant architecture)