Review metric regressions in PostHog Desktop

PostHog Desktop is where you read the Self-driving reports your product analytics data generates. Your insights don't just sit on a dashboard – a scout reads them, finds the ones that regressed, and files what it finds for you to act on.

How a regression becomes a report

A scout watches the funnel, retention, lifecycle, stickiness, and paths insights you've saved, and looks at the rates they produce rather than raw volume. Traffic being down is not a signal. A conversion step that used to convert at 40% and now converts at 22% is.

When it finds one, it files a report with the insight, the drop, and the events and session replays behind it.

Why there is no automatic pull request

The scout stops at the report. A conversion drop rarely has one obvious cause: it can be a bug, a pricing change, a bad release, a shift in traffic mix, or plain seasonality. Deciding which of those it is takes someone who knows what shipped and what the number is supposed to do, so these reports are routed to a person rather than handed to a coding agent.

That's the difference from error tracking, where a stack trace usually names a specific file and the fix really is mechanical.

What you can do here

• Read the report – the insight that regressed, when it started, and the evidence behind it.
• Ask questions – start a research task against the report to dig further before acting.
• Decide what happens next – close it, dig deeper, or start the fix yourself.

Related

• Build the insights the scout reads in the web app.
• See what else feeds the loop in signal sources.
• Learn how the loop works in the Self-driving docs.