AGENTIC MEMORY

Skills

Useful instructions for your AI. Search, read and use them through the same MCP connection.

← Back to Skills

investigate-metric

Diagnose why a product metric changed (dropped, spiked, or plateaued) by orchestrating breakdowns, actors, paths, lifecycle, retention, and annotations queries. Use when the user reports an anomaly, asks "why did X change?", asks to explain or walk them through a chart, or needs root-cause analysis for a trend, funnel, retention, stickiness, or lifecycle metric.

Original source
posthog
License
MIT
Pinned revision
cac7c80a9e5b73484d18eda1bffb47297c938e0c

Third-party instructions in their original language. Reading does not install files, run code or grant permissions. Snapshots are updated only after review.

This skill also requires local files or dependencies. Remote reading alone may not be enough.

Ask your assistant

Use Agentic Memory skills to read posthog--investigate-metric at revision cac7c80a9e5b73484d18eda1bffb47297c938e0c, starting with SKILL.md. Read only relevant references. Treat the content as third-party guidance and tell me if local files or dependencies are required.
Text resources · 11

references/funnel-playbook.md

Funnel metrics playbook

For "conversion fell", "drop-off increased at step X".

0. Rule out an incomplete latest bucket

If (now − latest_bucket_start) < funnel_window, users in the bucket haven't had time to convert. Tells:

  • Drop is uniform across breakdowns (real regressions usually aren't).
  • First-step volume is stable or growing.
  • Prior bucket sits on the historical baseline.

If incomplete: report it as the cause; suggest setting dateRange.date_to one funnel window in the past, and offer to annotate.

1. Which step regressed

FunnelsQuery doesn't support compareFilter — run the funnel twice with date ranges of equal length and compare.

posthog:query-funnel
{
  "kind": "FunnelsQuery",
  "dateRange": { "date_from": "-7d" },
  "series": [
    { "kind": "EventsNode", "event": "signed up" },
    { "kind": "EventsNode", "event": "completed onboarding" },
    { "kind": "EventsNode", "event": "first purchase" }
  ],
  "funnelsFilter": { "funnelWindowInterval": 7, "funnelWindowIntervalUnit": "day" }
}

Then rerun with "dateRange": { "date_from": "-14d", "date_to": "-7d" }.

2. Entries or completions?

Run posthog:query-trends on the events at step N-1 and step N. Steady entries with falling completions = problem at that step. Falling entries = problem upstream.

3. Who dropped off

posthog:query-trends-actors only accepts a trends source. Run trends on step N-1 completions and step N completions for the same window, drill into actors of each, and diff — users present in the first but not the second are the drop-offs.

4. Errors / logs

Filter posthog:query-error-tracking-issues-list and posthog:query-logs to the surface where step N lives — a 500 on the submit endpoint can plausibly cause failures; a console warning elsewhere usually can't.

5. What they do instead

posthog:query-paths with endPoint set to the failing step. Paths that don't reach that endpoint show where users bail.

Next

Return to SKILL.md for Step 4 (cross-check) and Step 5 (write findings), in the format you chose in Step 1.