---
name: skill-conversion-diagnostic
type: skill
description: >-
  Diagnose where qualified users leak from a venture's funnel — build the stage
  spine, overlay the forces-and-barriers conversion engine, and find the leaks
  against the one or two metrics that matter, surfacing each gap as known, assumed,
  or unknown. Use this whenever Sonar's object is a funnel rather than a thing:
  a conversion problem, a drop-off, "why don't qualified users convert?" A
  specialised sibling of skill-discovery-audit that shares its card ladder. Reads
  convention-user-modelling and convention-audit; feeds skill-prioritise. Surfaced
  by the a1.Muse engagement.
---

# conversion-diagnostic

Sonar's funnel skill. Where `skill-discovery-audit` assesses a thing against a standard, this assesses a *journey* against the forces that move or block a switch, and locates where qualified demand leaks away.

## Trigger

The problem is a conversion one — a drop-off, a stall, qualified users who don't progress. The a1.Muse engagement was this skill's first run.

## Behaviour

1. **Build the stage spine.** Lay out the funnel as ordered stages, from first contact to the converted outcome. Define what "qualified" means at entry, so a leak of the right people is distinguished from the wrong people self-selecting out.
2. **Name the metrics that matter.** Identify the one or two metrics the funnel turns on (the core-job conversion and, where relevant, its lead indicator). Everything is read against these, not against vanity counts.
3. **Overlay the conversion engine.** At each stage apply the Forces (Push / Pull / Anxiety / Habit, per `convention-user-modelling`) and the barriers in the way. Ask what should be pulling the user forward and what is holding them back.
4. **Locate the leaks.** Find the stages where qualified users drop disproportionately, and attribute each leak to the force or barrier responsible — not just "they leave here" but "they leave here because".
5. **Capture as the card ladder.** Record each leak as a Finding, cluster into Theme Summaries with root cause and HMWs (per `convention-audit`), and tag each known / assumed / unknown.
6. **Hand off.** Pass the themes to `skill-prioritise`; route high-stakes unknowns to research-first before committing.

## Reads

`convention-user-modelling` (forces, modes, JTBD), `convention-audit` (card ladder, conversion-diagnostic variant), `standard-prioritisation` (severity, confidence overlay).

## Guardrails

Define "qualified" before counting leaks. Attribute every leak to a force or barrier, not just a location. Read against the metrics that matter, not vanity numbers. Mark assumptions; route high-stakes unknowns to research first. Stay diagnostic — name the leaks, hand sequencing to `skill-prioritise`. Propose-only.