Analyze

/sunboard.analyze grounds every "why aren't users activating?" question in real data, then turns the diagnosis into a concrete plan that hands off to the right skill. It reads analytics and proposes changes — it never edits specs, mutates segments, or deploys.

When to use it

• An experience is live and you want to know where users drop off.
• After shipping a fix, to confirm the funnel actually moved.

What happens

1. Finds the experience (or reads sunboard state to pick one).
2. Fetches analytics — sunboard analytics experience <key> --json, defaulting to a 30-day window.
3. Reads the diagnostics first — noViews (nothing wired or bound), missingGoal (no activation event), largestDropoff (the primary funnel signal), noTourStarts (launcher problem).
4. Cross-references the spec to form a hypothesis — wrong route/selector, an event firing from the wrong path, a step demanding too much before value, tour copy, or the wrong audience.
5. Slices when needed — by version, segment, environment, or origin.
6. Writes an optimization plan naming the fault, the numbers, the proposed change, and which skill to run next.

Related