Is the AI pack distinctive or just familiar?
Check whether an asset points to your brand or a rival.
The question
Luma is a fictional tea brand. In each random cell, 200 category buyers see one pack with the product name covered. Each respondent names one brand or answers other/do not know. The goal is recognition of Luma, not recognition of tea packaging.
| Luma | Rival | Other or do not know | |
|---|---|---|---|
| Current pack | 146 | 10 | 44 |
| AI pack | 92 | 64 | 44 |
Each row totals 200 people. The name is hidden in both cells; this tests the remaining design assets.
At least 70% must identify Luma, and at most 10% may identify the rival. These are invented rules for choosing a limited pack test.
“Choose your next action and explain the evidence behind it.”
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: Choose the recognition measure
What would you do here? Open the move you’d make.
Measure correct and rival attribution separately.
The strong move. Category familiarity is not brand recognition.
Treat any familiar-looking pack as a strong Luma asset.
Tempting, but it misses something. A familiar style can be attributed to the rival instead of Luma.
See how a strong candidate says it
“I would count correct Luma attribution and rival attribution separately. A familiar category style can make a pack easy to process while sending people to the wrong brand.”
Why it works: Category familiarity is not brand recognition.
Step 2: Check both gates
What would you do here? Open the move you’d make.
Retain the distinctive assets and revise the illustration.
The strong move. The design must identify the right brand.
Approve the AI pack because it looks professional.
Tempting, but it misses something. Professional appearance does not repair 46% correct attribution or 32% rival attribution.
See how a strong candidate says it
“The current pack scores 73% Luma and 5% rival. The AI pack scores 46% Luma and 32% rival, failing both gates. I would restore Luma’s recognisable crest and colour relationship before retesting.”
Why it works: The design must identify the right brand.
Step 3: Reassess the new pack
A revised AI pack retains the crest and colour relationship. A new comparable cell of 200 records 150 Luma responses, 8 rival responses and 42 other/do not know responses.
What would you do here? Open the move you’d make.
Advance this tested revision.
The strong move. A changed design deserves a new evidence-based decision.
Approve every variant generated from the same prompt.
Tempting, but it misses something. This result covers one design; it does not test the other generated variants.
See how a strong candidate says it
“The revision scores 150/200 = 75% Luma and 8/200 = 4% rival. It passes both gates and can move to the limited pack test.”
Why it works: A changed design deserves a new evidence-based decision.
Step 4: Protect the production version
What would you do here? Open the move you’d make.
Check the actual production pack before release.
The strong move. The tested design must survive production.
Allow the printer to redraw the crest to fit.
Tempting, but it misses something. Redrawing the crest changes an asset whose recognition the test depended on.
See how a strong candidate says it
“I would name the approved asset version and brand owner, then check shelf legibility and print reproduction. Small production changes can alter the assets that passed.”
Why it works: The tested design must survive production.
The answer, in short
- Decision
- Advance the revised pack with its tested assets intact.
- Because
- Correct and rival attribution both pass the stated gates.
- What would change it
- Poor shelf legibility or altered production assets would reopen the decision.
Common traps
- Confusing category cues with ownable assets.
- Extending one test result to untested variants.
Check your answer
- I checked that each row totals 200.
- I separated Luma and rival attribution.
- I applied both gates to the revision.
- I protected the tested asset version.
Last reviewed Nov 10, 2026