Most UA best practice is survivorship

Almost everything the industry calls UA best practice is a description of what the winners did, filtered through the winners' memory of it. That isn't the same as knowing what works. If a hundred studios ran the same creative volume programme and four of them scaled, the four give the talks; the other ninety-six don't get a slot at PGC London, and nobody writes a case study called "we shipped 2,500 variants a quarter and closed the studio".

The uncomfortable version of this thesis is that most of what a UA team copies from a conference deck isn't causal. It's a correlate of being the kind of studio that survived long enough to present, and something else mostly decided survival: the game, the cash runway, the timing of a genre wave, or a platform relationship that never appears on a slide.

Why the filter is so strong in this industry

Survivorship bias exists everywhere, but mobile UA has three features that make it unusually severe.

The first is churn of studios. Games businesses fail quietly and quickly, and there's no bankruptcy filing for a 15-person team whose second title never got past soft launch; it simply stops posting. The population of failures is invisible. So the denominator is unknowable.

The second is that the platforms curate the evidence. Case studies on ad network sites and MMP blogs, or on store developer pages, are commissioned marketing, and a network has no incentive to publish "advertiser doubled creative spend, saw no change in ROAS, churned". AppsFlyer's State of Gaming for Marketers 2026 report, published in January with Unity and Newzoo, noted that top advertisers produce 2,400 to 2,600 creative variations per quarter, up roughly a quarter year on year. That describes who is at the top. It isn't a finding that variant volume put them there, and the report doesn't claim causation; readers supply it.

The third is compensation. UA leaders get promoted on wins, and a win is easy to narrate, whereas nobody's promotion packet says "we avoided a mistake that would have cost us the year", because the counterfactual doesn't appear on any dashboard.

The mechanism: what actually drives the correlation

Take the most repeated best practice of the last two years: high creative velocity. The observed fact is that large advertisers ship far more variants than small ones, and the story told about it runs volume drives learning, learning drives winners, winners drive scale.

The mechanism is mostly the reverse. Scale drives volume. A studio spending heavily can afford a 20-person creative team and a testing budget that yields statistically clean reads on each variant, while the same volume at a small budget produces variants that each get a few thousand impressions, which is noise. Copying the output of a big team without the spend that makes the output meaningful is cargo cult UA. The plane doesn't land.

The same reversal applies to the "diversify off Meta and Google" advice. The studios that diversified successfully were mostly studios whose games had broad enough appeal to convert on rewarded inventory; diversification didn't make the games broad, the breadth made diversification survivable.

The second-order cost most coverage misses

Here is the trade-off nobody puts in a slide. When an entire category adopts the same survivor playbook, the playbook stops working for everyone, including the survivors.

If every mid-core studio decides that UGC-style creative is the winning format because the top grossers use it, the auctions on that inventory fill with near-identical ads. The winners' advantage was partly that they were early. The followers pay full freight for a diluted signal, and the earlier movers now face rising costs on inventory they used to own cheaply. Best practice, by spreading, consumes itself, which is why the half-life of a UA "trick" keeps shortening and why a team that reads a lot of case studies can end up systematically late.

There's a subtler cost too. Following survivor practice narrows the range of things a team tries. If your creative brief says "match the format the category leaders use", you've removed from the portfolio the very outliers that generated the original winners, and you're optimising within a space somebody already mined out.

A decision rule for reading any case study

The practical question isn't whether to ignore case studies. It's how to weigh them. A workable rule is to ask four questions before adopting anything presented as best practice.

• Would this have looked like a good idea before the result was known, using only the information the team had at the time?
• What is the plausible denominator? How many teams likely tried the same thing, and do I have any evidence about what happened to them?
• Does the practice depend on a resource I don't have: budget, first-party data scale, a specific game shape?
• Does it depend on an early-mover position that has already closed?

If a practice fails the first question, it's hindsight dressed as strategy. If it fails the second and you have no denominator, treat it as a hypothesis for a small test, not a policy. If it fails either of the last two, the practice isn't transferable regardless of how clean the case study looks.

A worked illustration

Consider an illustrative mid-sized puzzle studio deciding how to allocate a creative budget for the second half of the year. The team has read that the category leaders run hundreds of concepts per month and that "fail fast" testing is the standard.

Applying the rule: the practice would have looked plausible in advance, so it passes the first test. On the second, the team knows of at least a dozen comparable puzzle studios that ran heavy testing programmes in 2024 and 2025, and can name only two that are still growing. On the third, the studio's monthly spend means most concepts would never reach a confident read.

The rational response isn't to abandon testing. It's to size it to the budget: fewer concepts, each given enough spend to produce a real signal, with the saved production money redirected to holdout measurement of whether the winners are incremental at all. That's a different programme from the one on the conference slide, and a better one for this studio, because it takes what the team can actually observe as its starting point rather than what a survivor remembered.

Still go to the talks. Just listen for a different thing. The most valuable content in a survivor's presentation is rarely the playbook; it's the list of things they tried that didn't work, the moment they nearly ran out of money, and the piece of luck they credit only when pushed in the Q&A. Those parts give you the denominator the slides leave out.

And when your own team wins, write the version of the story that includes the failures. Somebody else's decision a year from now depends on whether anyone in this industry is willing to publish the base rate.

Related archive reading

These articles provide related context and remain subject to their stated review status.

• Bending Spoons files its S-1: a portfolio roll-up read
• Anatomy of a creative that scaled, then died