AI discovery brings trials. Does it bring growth?
Compare retained buyers and returns before you reward the new channel.
The question
Fictional consumer brand Vale compares two acquisition cohorts with the same product, prices and observation window. Acquisition costs and customer outcomes differ.
| Paid search | AI shopping assistants | |
|---|---|---|
| First-time buyers | 1000 | 1000 |
| Acquisition spend USD | 20000 | 15000 |
| Repeat buyers within 60 days | 300 | 180 |
| First orders returned | 50 | 120 |
Invented data. Same product, price and observation window; margin is not supplied.
- Unknown
- Gross margin, return-processing costs and the accuracy of third-party product descriptions have not yet been checked.
“Would you move most of the budget to AI shopping assistants?”
Not saved by this site. Your browser may restore text when you navigate back.
A strong answer, step by step
Open each step in turn. This is one strong answer, not the only one.
Step 1: Compare the right denominator
What would you do here? Open the move you’d make.
Move budget because first-order acquisition is 25% cheaper.
Tempting, but it misses something. That compares only acquisition spend per first order and ignores weaker repeat buying.
Compare repeat buyers, returns and contribution.
The strong move. Repeat-buyer conversion is 30% versus 18%, with the same cohort size.
See how a strong candidate says it
“Paid search costs about USD66.67 per repeat buyer; AI discovery costs about USD83.33. Lower first-order acquisition cost does not win on this measure.”
Why it works: Repeat-buyer conversion is 30% versus 18%, with the same cohort size.
Step 2: Inspect why customers return
“Returns are 5% versus 12%. Sample the reasons and audit the claims AI shopping assistants display.”
Why it works: A channel can recruit the wrong expectations, even when first orders look cheap.
Step 3: Check contribution, not only conversion
“Calculate contribution after acquisition, product margin, refunds and service costs by cohort.”
Why it works: We cannot infer lifetime value from two orders or compare profit without margin.
Step 4: Run a bounded channel test
“Keep a small test allocation. Correct unsupported descriptions and measure a fresh cohort with the same 60-day window.”
Why it works: This preserves learning without scaling a potentially misleading channel.
The answer, in short
- Decision
- Do not move most of the budget yet. The AI cohort costs more per repeat buyer and has more returns.
- Because
- The apparent acquisition saving disappears on the repeat-buyer measure; profit remains unknown.
- What would change it
- Higher contribution after returns, or a fresh cohort with better retention after correcting claims, could justify more budget.
Common traps
- Calling the cheapest first order the best channel.
- Ignoring returns.
- Comparing cohorts with different observation windows.
Check your answer
- Did I use the same cohort denominator?
- Did I compare cost per repeat buyer?
- Did I include returns and margin?
- Did I define the next test?
Last reviewed Nov 11, 2026
