Funnel Analyzer

A funnel analysis maps the steps users or customers pass through from first touch to core value (and optionally revenue), measures conversion between stages, and prioritises where to fix leaks. It combines:

• Stage definitions aligned to user mental model.
• Baseline conversion per step and time window.
• Segment cuts — channel, persona, plan, geography.
• Impact ranking — drop-off volume × gap vs benchmark.
• Testable hypotheses — cheapest experiment first.

Teams run A/B tests on headlines while 50% of users die at HRIS connect. Funnel diagnosis focuses experiment budget on steps that move the north star, not vanity UI tweaks. It also exposes instrumentation gaps — if you cannot measure a stage, you cannot improve it.

When growth stalls, onboarding underperforms, PLG conversion drops, or before a major UX or pricing change. Pair with /datastrat if event taxonomy is unreliable.

1. Define stages — 4–7 steps from acquisition to activation (extend to revenue if needed).
2. Instrument events — one owner per stage metric.
3. Baseline conversion — overall and by key segment.
4. Quantify leaks — absolute drop-off count, not only percentage.
5. Rank by impact — volume × distance from benchmark or prior period.
6. Draft hypotheses — falsifiable; one primary change per test.
7. Queue experiments — /abtest on top leak; measure guardrails.

• Stages match how users describe their journey.
• Segment view reveals leaks hidden in averages.
• Hypotheses state expected lift and guardrail metrics.
• Follow-up tests documented; losers retired.

• Too many micro-steps — noise without insight.
• No segment view — mobile vs desktop, SMB vs enterprise.
• Optimising top of funnel when activation is broken.
• Missing instrumentation — guessing from support tickets.

Northvale Systems portal funnel: invite → account → docs uploaded → ERP sync. 100% → 71% → 48% → 39%. Largest leak: doc upload (48% of lost volume). EU suppliers −22 pts vs domestic. Hypothesis: checklist unclear → guided wizard vs PDF. Next: /abtest.

PulseWell PLG: landing → signup → HRIS connect → first nudge → manager action. 100% → 12% → 6% → 4% → 1.4%. Half of loss at HRIS connect. Hypothesis: OAuth errors opaque → better failure copy + CSV fallback.

Harbor diagnostic funnel: ad → form → paid diagnostic → workshop. Leak at form (62% abandon on plant size field). Hypothesis: two-step form with email capture first.

Clearwater volunteer funnel: training invite → first check-in → 30-day retention. Leak post check-in — photo upload on low bandwidth. Hypothesis: offline queue or USSD-only pilot on 20 sites.

Run /funnel for FUNNEL-001 with staged metrics and hypotheses. Follow with /abtest on the top-ranked leak; align activation stage with /northstar inputs.