UA is a capital allocation problem

The UA team at most studios sits in the marketing org chart, reports to a CMO or a head of growth, and answers for ROAS. That's a category error. Paid acquisition is the single largest discretionary use of cash in a live mobile game, it produces a return that arrives over months, and it competes for that cash with product and live ops, and with the plain option of leaving the money in the bank. That makes it a capital allocation decision, and the people making it should face the test an investment committee faces: return on capital deployed, adjusted for how long it's tied up and how sure they are of getting it back.

Plenty of experienced buyers will bristle at that. Their view is that UA is a craft; creative and bidding decisions live at a tempo finance can't follow, and treating installs as loans slows everything down. The craft point is right, but the conclusion is wrong: the craft is how you execute the allocation, not a reason to skip deciding what the allocation should be.

Why the marketing frame produces bad decisions

Studios set marketing budgets top-down, as a percentage of revenue or a fixed annual number, and then spend them. The incentive that creates is to deploy the budget, because an unspent budget is next year's cut. A UA manager with a fixed monthly number and a ROAS target will spend to the number and steer ROAS toward the target, which usually means buying cheaper, lower quality installs late in the month to hit volume while the blended figure still clears the bar.

An allocation frame flips the incentive. The question is no longer "did we spend the budget at target ROAS" but "of every pound we could have deployed, did we deploy it where it returned the most, and did we stop when the next pound returned less than the alternative". Under that frame an unspent budget isn't a failure. It's a correct decision about a poor opportunity set.

When installs get scarcer and dearer, how you ration capital matters more than how you spend it.

The mechanism: installs are loans with uncertain repayment

An install is a payment made today for a stream of revenue that arrives later, which makes it, structurally, a loan the studio makes to its own future. Three things determine whether the loan is a good one, and only one of them shows up in a ROAS dashboard.

The first is the size of the return, which is what ROAS measures, imperfectly, at fixed horizons.

The second is time. Cash returned in 30 days can go straight into the next cohort; cash returned in 360 days can't. Two campaigns with the same 12-month ROAS aren't equal if one repays in three months and the other in eleven, because the faster one compounds. A studio funding growth from its own cash flow feels this acutely; a studio with a large external raise can afford to ignore it, which is exactly why well-funded studios so often overpay for installs.

The third is certainty. A predicted 12-month LTV built on a 30-day cohort is a forecast, and forecasts on paid cohorts from a new channel are wider than forecasts on a mature one. An allocation frame charges for that uncertainty. A marketing frame doesn't, which is why "scale the new channel, the early ROAS looks great" is one of the most expensive sentences in the business.

A second-order effect: the cash you never see

The cost most coverage misses is what the capital could have done elsewhere. When a studio pushes UA spend past the point where the marginal cohort repays inside its funding horizon, it isn't only earning a thinner return on those installs. It's starving the alternatives: a live ops team that could have lifted retention across every cohort, a localisation push that could have opened a market with cheaper installs, or a web shop that could have raised the margin on every existing payer.

Those alternatives share a feature that paid UA lacks: they improve the return on installs already bought, whereas paid UA at the margin buys more installs at a worse return. When a growth head has one pot and two doors, the door that raises LTV for everyone usually deserves the first try, ahead of the one that lowers ROAS for the next cohort. Most org charts make that comparison impossible because the two doors belong to different people with different budgets.

A working framework: the hurdle rate

Borrowing from how any capital committee works, give UA a hurdle rate rather than a budget.

Set the hurdle as the return the studio could earn on the same cash elsewhere, plus a premium for forecast uncertainty. For a self-funded studio the "elsewhere" is often the next live ops release or simply holding cash for a launch; for a studio with expensive external money it's the cost of that money. State the hurdle as an annualised figure so you can compare it across horizons.

Then take every channel and geo cell and, each fortnight, ask three questions of it. What is the predicted 12-month return on the last increment of spend, not the average? How many months until that increment returns its cost, on the current cohort curve? And how wide is the confidence band on that prediction, given the cohort's age and the channel's maturity?

Convert the answer into an annualised return on the marginal pound. Spend where it clears the hurdle; cut where it doesn't, and leave the remainder unspent or hand it to whichever team cleared the hurdle with a non-UA project.

An illustrative example makes the difference visible. Suppose a channel returns its cost in four months and then another 40 percent by month twelve. Its 12-month ROAS of 140 percent looks similar to a second channel returning 145 percent, but the second one only repays its cost in month ten. The first channel's cash can go round twice more in the year. To annualise it, count the turns: a pound back at month four can go out again three times a year, so the 40 percent tail on each turn compounds to roughly three turns of return against the second channel's one. On an annualised, redeployable basis the first channel is far ahead, and no ROAS-at-D365 target will tell you that.

What changes in the room

Running UA as capital allocation changes who is in the review meeting and what is on the screen. Finance is in the room, not as a cost-controller but as the owner of the hurdle rate. The UA lead presents marginal return by cell, not blended ROAS by channel; the live ops lead presents the same metric for their next release, and someone chooses between them.

It also changes the UA lead's career incentives, which is the part people underestimate. A buyer judged on deployed volume at target ROAS will always find more volume. A buyer judged on return on capital, with permission to hand money back, will start saying no to channels, and that is the behaviour that separates studios that survive a downturn in install supply from studios that discover their ROAS was a mirage the quarter after the cheap installs ran out.

Related archive reading

• The platform tax and the true cost of an install
• UA teams should own store conversion