The Three-Question Filter: Auditing a Referral Program Before It Ships
Objective: Given a real referral program design brief, apply the lesson's three-question ethics filter (is the fact true, is reversal cost equal to action cost, would it survive being explained out loud) to decide what ships as-is, what needs a fix, and what needs a legal review before launch.
You're the growth marketer at Walker & Company Brands, the grooming company (Bevel) that Procter & Gamble acquired in 2018. Product is proposing a referral program for a Q4 launch and wants your sign-off by Friday.
Run all 5 elements of the proposed program through the three-question filter and write a go/no-go memo with specific fixes for anything that fails.
Before you start
What you'll need
Free path (everything below is enough to finish)
Free, shareable directly with product and legal for sign-off
The process
1 step
Step 01 of 01
The lesson's practical framework runs three checks in order: is the underlying fact true, is the cost of reversing the action equal to the cost of taking it, and would the tactic survive being explained out loud to the user it targets.
The brief proposes: (1) a 'limited spots' referral cap that isn't actually limited, (2) a reward that pays out only after the referred friend's 3rd purchase with no disclosure of that condition, (3) an opt-out link buried in a footer FAQ. Which of these ship, which get fixed, which get killed?
Procedure
- List every element of the referral program brief in its own row
- For each, answer question 1: is the asserted fact (cap, reward, deadline) actually true?
- Answer question 2: is opting out or reversing as easy as opting in?
- Answer question 3: would this survive being explained in plain language to the user?
- Any element failing one or more questions gets a specific fix, not a vague 'make it clearer'
REFERRAL PROGRAM AUDIT 1. 'Limited spots' cap -- Q1 FAIL (no real cap exists) -- VERDICT: fix, remove fabricated scarcity or implement a real cap 2. Reward requires friend's 3rd purchase, undisclosed -- Q1 FAIL (headline implies immediate reward) -- VERDICT: fix, disclose the condition in the same font size as the headline 3. Opt-out buried in footer FAQ -- Q2 FAIL (signup is 1-click, opt-out requires finding a footer link then reading an FAQ) -- VERDICT: fix, add opt-out to the same settings screen as signup ...2 more rows
Healthy
Every failing element gets a specific, shippable fix, not a blanket 'run it by legal'.
Unhealthy
The whole program gets killed because one of 5 elements failed, wasting a legitimate growth mechanic.
What this means
The filter isolates which specific claim or asymmetry is the problem, so the fix is surgical instead of scrapping the entire program.
So what do I do about it?
| Symptom | Action | Effort |
|---|---|---|
| Team debates whether the whole referral program is 'too risky' without pinpointing which element | Run every program element through all 3 questions individually before any go/no-go decision | 30 min |
Final deliverable
A go/no-go memo: each program element, its verdict (ship, fix, kill), and the specific one-line fix for anything that failed a question.
See a reference example
Drunk Elephant referral audit (excerpt) ELEMENT: 'Give $15, Get $15' headline paying out as store credit Q1: FAIL, headline implies cash-value discount, actual reward is time-limited store credit VERDICT: fix, headline must read 'Give $15 credit, Get $15 credit' or the reward must become cash-equivalent ELEMENT: cancel-anytime subscription toggle in account settings Q2: PASS, toggle is one click, same surface as the original signup CTA VERDICT: ship as-is
Success criteria
You're done when you can:
- Every program element gets an answer to all 3 questions, not just an overall verdict
- Fixes are specific and shippable, not generic
- Memo distinguishes 'fix and ship' from 'kill entirely'