Analytics can show where people drop off or stop engaging, but deciding what to do next still takes work. At Revnest, I needed a clearer way to turn usage across the free marketplace and paid product into product priorities.
I built an automation that reviews PostHog analytics and recommends what to improve at Revnest. The automation compares recent usage with the previous period and looks for changes in conversion and engagement.
It ranks possible improvements by impact, supporting evidence, confidence and effort, then sends them to Slack for review. A separate automation checks for approval before creating Linear tickets.
The architecture, at a glance.
Aggregate events, funnels and comparison periods
Evidence checks, strategic filters and ranking
A person approves the proposed ranked opportunities
Duplicate check and a ticket with formal write-up
Each ticket includes a metric and a review window, so later results can be compared with the original data.
I built the ranking around what people actually do in the product. Small samples, missing totals and guesses about why something happened all reduce confidence. If the evidence is too weak, the automation makes no recommendation.
The analyzer can recommend work, but it can’t change the product or create tickets. A separate automation checks that someone has approved the recommendation before creating work in Linear and then moving to development.
Each opportunity gets an identifier so repeat runs can check whether it has already been raised. The ticket keeps the original evidence and measurement plan for review later.