← Research · Retention
Retention and Engagement in AI Products
Retention improves when a product keeps solving a real job. Define the first useful result, watch where people stop, and compare cohorts after changing the workflow.
By Jake Bauman · Revised
Jake Bauman's measurement framework. Any example metric is a proposed instrument, not a benchmark.
Name the return reason
What recurring job makes a customer come back? Identify the moment they first receive value and the interval at which the job recurs.
Instrument the journey
- Track activation separately from sign-up.
- Group users by start date and intended use.
- Record corrections, abandoned outputs, and support requests.
- Ask churned customers what the tool failed to do.
Avoid a false win
More generated outputs can coexist with lower trust. Pair usage with quality, repeat use, and the customer's own account of value.
A cohort review
Take users who started in the same week and had the same intended job. Compare those who reached a first useful result with those who did not. Review their correction patterns and support questions. If the returning cohort is small, interview the people who left before assuming a feature caused churn. A result can be valuable even when the usage frequency is low.
Sources and context
These sources inform the framework. They do not validate the suggested actions for every business.
- Brian Balfour, Four Fits for $100M+ Growth: Original source of the Four Fits framework.
Want a practical starting point? Enter your email to get the free Growth Agent Starter Kit download link on the page.