Diagnose Why Customers Don't Come Back — Then Plug the Leaks

You are a retention analyst for an Australian small business. Work out why customers don't come back — with hypotheses tested before money is spent — then plug the confirmed leaks.

<context>
BUSINESS: [TYPE + NATURAL PURCHASE CYCLE — e.g. hairdresser ~6 weeks / accountant ~yearly]
REPEAT PICTURE: [ANY NUMBERS OR HONEST ESTIMATE — e.g. maybe 3 in 10 return]
JOURNEY: [FIRST VISIT TO HOPED-FOR SECOND, STEP BY STEP]
EXIT SIGNALS: [COMPLAINTS, REVIEW GRUMBLES, THINGS SAID ON THE WAY OUT]
COMPETITOR PULL: [WHO THEY MIGHT GO TO INSTEAD AND WHY]
TRIED ALREADY: [PAST RETENTION ATTEMPTS AND RESULTS]
CONTACT DATA: [WHAT YOU HOLD — emails? numbers? consent to contact?]
</context>

Before diagnosing, sort non-returners into three kinds — defected (unhappy, went elsewhere), drifted (happy enough, just forgot), and natural one-offs (tourists, one-time need) — because each needs a different fix and the mix determines where effort goes.

<task>
1. Estimate the likely mix of the three kinds from my evidence, stating the reasoning and confidence; unknowns become [NEEDED: …].
2. Build a leak-hypothesis table: journey point → hypothesised leak → evidence status (supported / contradicted / untested from my inputs).
3. Design cheap tests for the top 3 untested hypotheses. Include a lapsed-customer call/text script (5 customers, 3 questions, non-defensive wording — asking permission first if consent status is unclear) and one observational test that needs no contact.
4. For each SUPPORTED leak: the fix, its cost basis from my inputs, and who does it.
5. Drifted customers get a reminder rhythm matched to the purchase cycle (only where I hold contact permission — messages must carry an opt-out under Australian spam rules, stated as fact).
6. Define the repeat-rate metric to track from now on — simple enough for a spreadsheet: definition, source, review day.
</task>

<output_format>
Mix estimate → hypothesis table → 3 test designs with scripts → confirmed-leak fixes → drifted-customer rhythm → metric definition.
</output_format>

Rules: never treat an owner hunch as a confirmed leak; the tests exist because hunches about 'why they leave' are usually wrong. En-AU spelling, plain language.

Copy the block above straight into Claude — anything in [BRACKETS] is yours to fill in.