UA and product teams optimise against each other

In most mobile game studios the acquisition team and the product team report to different people, answer to different numbers, and are each capable of hitting their target by making the other's target harder. This isn't a communication problem, and another cross-functional sync won't fix it. It's an incentive design problem, and it's present in the structure of almost every growth organisation we have looked at.

The argument here is that the standard split (UA owns payback, product owns retention) isn't a neutral division of labour. It's a mechanism that predictably produces two specific distortions, and until a studio measures both teams on at least one number they cannot game against each other, the friction between them is the system working as designed.

How each side games the other

Start with UA. A team measured on day thirty return on ad spend has a lever nobody talks about in the review. Who it acquires. Creative that pre-selects for spenders, geo mixes that skew towards high-ARPU markets, and network choices that favour existing payers all improve early payback, and they also tend to shrink the pool of long-tail retained players who make a game's social features feel alive, its leaderboards and live events included. UA hits its number. Product inherits a thinner and more transactional population, then watches its engagement metrics soften for reasons that look like design failures.

Now product. A team measured on retention has its own lever: friction. Longer onboarding, gated features, harder early levels, slower resource pacing: all of them improve day seven and day thirty retention among the players who stay, because the players who leave early were never counted as retained in the cohorts product looks at. But every one of those changes lowers early conversion and early monetisation, which is exactly what the network's algorithm and the UA team's payback model are optimising for. Product hits its number. UA watches its CPIs rise as the network learns that installs from this game convert slowly, and gets told to fix its creative.

Neither team is acting in bad faith. Each is responding correctly to the metric someone handed it.

The mechanism underneath

The reason no amount of talking resolves this is that the two metrics have different denominators and different time horizons. Payback runs over acquired players and resolves in weeks; retention runs over players who reached some early milestone and resolves in months. Any decision that shifts players between the denominators, which is most decisions, will show up as a gain on one side and a loss on the other.

The second-order effect that most coverage misses is what this does to experimentation. When UA runs a creative test and product runs an onboarding test in the same fortnight, each team's treatment contaminates the other's control group. Both tests read noisy. Both teams conclude their change had a small effect. Over a year the studio's experimentation programme quietly loses power, and nobody attributes that to the org chart.

There's also an accounting consequence. Finance sees UA spend as a cost and product as a fixed headcount line, so when blended payback slips the pressure lands on UA. A product change that lengthened payback by a week can cost the acquisition team its budget for the following quarter, and the product team will never see the invoice.

A shared metric that resists gaming

The fix isn't to merge the teams, and it isn't more meetings. It's to give both teams one number they are jointly accountable for, chosen so that neither can improve it at the other's expense.

The candidate that works in practice is cohort-level net contribution over a fixed window: for each weekly acquisition cohort, total revenue minus acquisition cost minus an allocated share of variable service cost, measured at a horizon the studio agrees on, say day ninety. UA can't improve this by acquiring a thinner, richer population if that population churns before the horizon and leaves the social layer empty. Product can't improve it by adding friction if the friction lengthens payback enough to make the cohort unprofitable.

Illustrative worked example. A hypothetical mid-core title acquires a weekly cohort at a cost that the UA team's own model says will return in forty days. Product ships an onboarding change that lifts day seven retention by a modest margin among players who complete the tutorial, but lowers tutorial completion. Under the split model, product reports a win and UA reports a payback slip. Under cohort net contribution, the studio sees a single number move, and the question becomes whether the retained players are worth more than the lost early converters. That's the right question, and in the split model nobody was asking it.

Operating rules that make it stick

A shared metric only changes behaviour if it changes decisions. A handful of rules help.

• Any product change that touches the first session or the first purchase gets a UA sign-off on expected payback impact before it ships. Sign-off, not veto.
• Any UA change to geo mix or creative targeting gets a product sign-off on expected composition impact, on the same terms.
• Experiments go on a shared calendar so that no two tests share a cohort without both teams agreeing to accept the noise.
• The cohort net contribution number appears on both teams' dashboards in the same position, and the quarterly review discusses it before either team's own metric.

The trade-off is real. A shared metric resolves slowly, and teams need faster feedback to do their jobs. Payback and retention remain as leading indicators; they simply stop being the thing anyone gets paid on.

Who has to give something up

This lands hardest on whoever currently benefits from the ambiguity. In studios where UA has been the visible cost centre, the product team has enjoyed a decade of deniability about acquisition outcomes, and a shared metric ends that. In studios where product has carried the blame for softening engagement, UA loses the ability to buy a good quarter by narrowing the funnel.

The person who can impose this is usually the one to whom both teams report, and the reason it rarely happens is that this person currently receives two clean stories every quarter and would be trading them for one messy one. That messy story is the studio's actual position. It has been there the whole time, split across two dashboards nobody built to be read together.

Related archive reading

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

• ROAS targets versus payback windows
• Retention is a UA metric