Media Buying Is Solved. User Acquisition Is Not.

Buying media used to be the whole job

20 years ago the advantage lived in execution: knowing which network to use, how to structure campaigns, how to negotiate rates, how to move bids on a Tuesday afternoon.

Most of that has been absorbed into the platforms. Meta's Advantage+ collapses campaign structure down to a handful of inputs. Google's App Campaigns pick placement, bid, and creative rotation on their own. Meta's own pitch is that creative is the new targeting, which is a polite way of saying it took the targeting controls away.

The machinery for finding players got very good. If you have enough conversion volume, these systems will find more people who look like your payers, and they'll do it faster than any human buyer working a spreadsheet.

Distribution got industrialized. Growth didn't get easier. The work moved.

"If it's commoditized, why does everyone get different results?"

That's a fair question, and the answer is the actual argument.

Two studios can run the same platform, the same budget, and the same objective and finish 2 to 3x apart. On Meta, creative is estimated to account for 60 to 80% of the variance in CPA, and top-quartile creative pulls roughly 2 to 3x the CTR of median creative on the same channel.

The gap comes from inputs rather than buying skill: sharper creative, a cleaner conversion signal, a more accurate LTV model, a product that holds the players it gets sent.

The platform is a search engine running over your inputs. Commoditizing the search doesn't commoditize the inputs. It raises their value, because they're the only variable you still control.

The new bottleneck is learning

The studios pulling ahead answer questions faster than everyone else. That's most of the trick.

• Which concept creates real demand instead of curiosity?
• Which fantasy attracts players who spend?
• Which onboarding step loses them?
• Which creative brings payers instead of tourists?
• Which update actually extends lifetime?

Every answer narrows the range of things you might be wrong about. Answer faster than the studio next door and the gap compounds quarter over quarter.

The hard part you can't fix with a bigger budget

Here's the tension buried inside "learn faster."

In games, the decision and the data arrive at different times. You commit spend on Day 1. The signal you need shows up on Day 7, Day 30, sometimes later. In strong-retention genres like RPG and strategy, D30 captures only a slice of eventual LTV, and reliable extrapolation often wants D90 cohorts.

Then privacy took a bite out of what does arrive. SKAdNetwork reporting is delayed, aggregated, and probabilistic. SKAN 4 and AdAttributionKit improved the picture without restoring it. Most teams now treat platform attribution as one input among several rather than the source of truth.

So the practical version of "build a faster learning loop" is narrower and harder than the phrase suggests: get a trustworthy read on cohort quality in days instead of months, and stay honest about the error bars while you do it.

That's why early LTV prediction, incrementality testing, and creative-level signal quietly became the most valuable capabilities on a growth team. Each one is an attempt to buy back time.

The failure mode is familiar. A creative oversells a mechanic the game doesn't really have. It posts a great CTR, a CPI well under your blended average, and a D7 retention number that clears your threshold.

It clears the threshold on the way down. That cohort went 40% at D1 to 15% at D7. The one from your honest gameplay creative went 28% to 15%. The two share a D7 number and nothing else, and only one of them is alive at D30.

The level is what your dashboard reports. The slope is what predicts D90, and it's readable on day 7 as long as you're looking at cohorts instead of a blend.

Creative became the research instrument

Creative used to arrive after the game was built. Now it arrives first, and it's doing a different job.

You can put 200 concepts in front of a real audience before production art is locked. Different value propositions. Different hooks. Different promises about what the game even is.

Sometimes the market tells you your positioning is wrong. Sometimes it tells you your game is. Both are worth knowing, and the second one is worth more.

There's a limit to what this buys you. When every studio can generate 100 variants a week, volume stops separating anyone. The advantage sits in how sharply you form the hypothesis and how honestly you read the result.

AI shortens the loop

Most AI conversation in games is about unit cost: more ads, more copy, more variants, cheaper.

That's real, and it's the least interesting part.

A pipeline that generates concepts, pulls themes out of player reviews, explains yesterday's ROAS drop, predicts cohort value from Day 1 behavior, and proposes the next experiment worth running is doing something other than making assets cheaper. It raises the number of good decisions a team makes per week.

The job shifts from producing the work to choosing which questions are worth asking.

What this does to publishing

The old publisher question was: which games should we sign?

The emerging one is: which ideas deserve another round of evidence?

That reads like a small change. It reorganizes the whole company. Signing 6 big bets a year is a portfolio strategy. Running 60 cheap experiments and killing 54 of them is an operating system, and it needs different people, different tooling, and a very different relationship with being wrong in public.

The moat

Ad platforms will keep automating. AI tools will keep getting cheaper and more available. Creative production costs will keep falling toward zero.

None of that creates an advantage, because it happens to everyone at the same time.

What's left is the rate at which a company converts spend into knowledge, and knowledge into the next decision. The studios that win the next decade will have the tightest loop between product, creative, analytics, and what players actually do.

The best growth teams 10 years from now will describe themselves as learning systems rather than media buyers. A few already do.