Fractional UA lead for Salto Games. D0 ROAS doubled, day one payer conversion rose 50 percent, and blended CPI dropped from £9.43 to £5.38 in one month. Then the creative got tested against an outside studio on the same account.
This is the case for paid UA for mobile games and the account behind how I make ads. The second half is a direct comparison between my creative and an external studio's, published with the caveat that a careful reader would raise, so that they do not have to.
What was built
- Value optimization on Meta across iOS and Android, switched on once the signal could carry it.
- SKAN and AEM set up so iOS conversions carried value rather than counts.
- The cohort stack on Adjust and Tableau, so payback was visible by market and channel.
- Structured creative concept tests, rough first, with a spend rule per ad, and only the winners remade. The method is written out.
What happened
D0 ROAS went from 8 to 12 percent to 17 to 23 percent, roughly double. Day one payer conversion rose 50 percent. Blended CPI went from £9.43 to £5.38 in one month.
Two sources of creative, one account
From January to April 2026 the account ran ads from two sources side by side: mine, made with a creator I work with, and a widely known external creative studio brought in for comparison. Same Meta install campaign, same country, same spend rule. Here is what the account recorded, and one caveat about it that matters.
| Install test, Jan to Mar 2026 | My team | The studio |
|---|---|---|
| Ads tested | 175 | 16 |
| Ads over the £10 spend threshold | 41 | 4 |
| Spend | £1,208 | £186 |
| Installs | 993 | 125 |
| Blended CPI | £1.22 | £1.49 |
| Best ad, CPI | £0.65 | £1.21 |
| Ads under £1.00 | 14 | 0 |
The caveat, before anyone else raises it
My team had been testing since December. The studio’s ads entered in March. Compare March alone, like for like, and the gap is smaller: £1.34 blended CPI on my side against £1.49 on theirs, 11 percent rather than the 22 the headline suggests. Their sample was also thin: £186 of spend and four ads over the threshold. So I would not lean on the install phase on its own.
Where the algorithm decided
The studio’s best ads were then moved into the purchase campaigns running on value optimization, where Meta decides which ads get budget based on which ones drive value. Over 22 days from late March, the studio’s top ad received £4,763 of real spend, so it had a fair run. Meta walked its share of daily budget from a peak of 30 percent down to 4.5 percent, while my team’s ads went from around 70 percent to over 95 percent. Daily spend on the studio’s ad fell from £463 to £85. That is the algorithm’s verdict on a proper budget, and it is the number I would defend.
| Value optimization, 22 days | My team | The studio |
|---|---|---|
| Spend | £23,357 | £4,763 |
| Share of budget | 83.1% | 16.9% |
| Share trend | about 70% to over 95% | 30% to 4.5% |
What it says
More rough ads found a lower floor than fewer polished ones, and the ads that found it held up when the algorithm had real money to allocate. That is the whole argument behind how I price ads, and this account is where I learned it. What happened when the ads that spent the most were removed is the question that follows.
In the client’s words
We've worked closely with Growth Therapist, led by Samet, for the past two years as we scaled our first product. He led creative and UA strategy through our soft launch and beyond, and consistently drove fast, high-quality creative experimentation. He's proactive, communicative, and very transparent.
Questions people ask
Who was the studio?
A widely known creative studio. I have chosen not to name them, because the point is the method rather than the loser of a comparison. The client is named because the result is theirs.
Was the budget fair?
Yes, and it is the first thing I checked. The studio’s top ad received £77 in the install phase, more than most of my individual ads, and £4,763 in the value phase. Average spend per ad was higher on their side.
Why does the March comparison matter?
Because my December and February months were stronger than March, and they pull the blended figure down in a period the studio was not running. The honest gap is 11 percent on the install phase. The value phase needs no adjustment.