All work
05 / 06  Iterative Testing
PrevNext
Iterative Testing  ·  2026

Iterating on Performance Data

Shikaku Creative Testing

No competitor existed in the market yet to benchmark against, so there was no reference point to start from. The brief was to reposition the game for an older audience around a "brain training" angle, positioned as a daily habit app. Three rounds of testing followed, each one built on what the last one got wrong.

First test: Brain Age Angle
Flower USP
Chill gameplay

The first creative led with "What's your brain age?", showing Day 1 versus Day 30 progress (solving a 3x3 grid growing into a 10x10 in the same time). It targeted non-gamers, particularly people in their late 40s and 50s concerned about cognitive decline.

Result: CPI of $7.16. Breaking the Meta data down by demographic showed the ad had reached exactly the intended audience, but that audience itself had a high CPI. The opposite demo, men in their 20s and 30s, had the lowest CPI of anyone.

Second test: Confirming the Male Audience
Men UGC
Couple UGC

Round 1's finding suggested men were the lower-CPI demo, but that hadn't been confirmed. This round tested that hypothesis directly: comfort, sleep, and "for women" language was removed, and the creative was rebuilt in a more gender-neutral direction. UGC-style hook intros were tested as a variation, and voiceover was added to teach the game's rules, a pattern seen often in top-performing creative for puzzle games with less intuitive rules, like Meowdoku or Codebusters.

Result: CPI dropped to $1.9, confirming men in their 40s and 50s as the lowest-CPI demo.

Third test: Targeting Men Directly
Throwing rock intro

With the male demo confirmed, this round committed to it fully: an older male voice, and interests skewed toward US men 35 to 64. Even without deliberately targeting that group in Round 2, men 35 to 64 had already pulled 40% of spend at a $1 CPI, suggesting more room for CVR to grow once the creative spoke to them directly. The same speaker and UGC format were kept, and the effectiveness of the UGC intro format was re-validated.

Result: to be confirmed.