Set UA budgets by marginal return, not share
Budgeting UA as a share of revenue is a governance habit dressed up as economics. The last dollar spent, not the average one, should set the ceiling.
The most common way a mobile studio sets its UA budget is as a percentage of revenue: trailing or forecast, thirty percent or fifty, adjusted at the quarterly review. It is easy to explain to a board, easy to compare across studios and easy to defend. It is also, from an economic standpoint, close to meaningless. The number that should set a UA budget is the return on the last unit of spend, and a share-of-revenue target has no relationship to it whatsoever.
A sceptical operator will say that everyone knows this and the percentage is just a planning anchor. In practice the anchor becomes the ceiling, and the ceiling becomes the target, and by the third quarter the team is spending to hit the percentage rather than to clear a return.
Average return is not what the budget buys
The mechanism is diminishing returns. Every channel has a spend curve: the first tranche of budget buys the cheapest, best-matched users, and each subsequent tranche buys users that are slightly harder to find, slightly less likely to convert and slightly more expensive to reach. The average return across all spend can look healthy while the return on the most recent increment has dropped below the studio's hurdle.
An illustrative curve makes it concrete. Suppose a channel returns 180 percent of spend at ninety days on its first 200,000, 140 percent on the next 200,000 and 95 percent on the 200,000 after that. At 600,000 of spend the average ninety-day return is roughly 138 percent, which most teams would call a good campaign. The marginal return on the last 200,000 is 95 percent, which is a loss before fixed costs. A budget set at 600,000 because that is what thirty percent of revenue came to has funded 200,000 of value destruction that the average conceals entirely.
Share-of-revenue budgeting has a second, structural flaw: it is pro-cyclical. Revenue rises, budget rises, spend pushes further along the flattening curve into worse marginal returns. Revenue falls, budget falls, spend retreats from parts of the curve that were still clearing the hurdle. The rule systematically over-buys when the market is good and under-buys when it is cheap, which is the opposite of what an investor would do with any other asset.
What the platforms want you to measure
The incentives here point away from marginal thinking. Ad platforms report average metrics, because average metrics look better and because the platform's revenue scales with total spend rather than with the advertiser's marginal return. AppLovin's fourth quarter, reported in February, showed revenue up 66 percent year on year, and the whole self-serve industry has been built around making it frictionless to add budget. Automated bidding products, which Mintegral's February trends report noted had grown spend by more than half, are typically tuned to hit a target average return, not a marginal one. A campaign holding a 130 percent target ROAS can be quietly spending its last twenty percent at 90.
The trade-off that most coverage misses is that measuring marginal return is itself expensive. Estimating the slope of a spend curve requires varying spend, deliberately, in ways that cost money in the short term: pushing a channel above its comfortable level to see where returns fall off, or pulling it back to see how much return is lost. A team that never varies spend never learns its curve and defaults, by necessity, to averages. The budget for learning the curve is a real cost and should be booked as one.
A framework for marginal budgeting
The practical version has four parts.
Set the hurdle first, independent of any budget number. The hurdle is the minimum acceptable return on an incremental unit of spend at the studio's chosen payback horizon, after platform commission and the fixed layer of UA cost. It is a financial decision, not a marketing one.
Estimate the curve per channel, roughly. Step tests do this: hold creative and targeting stable, move the daily budget up or down by a meaningful increment for two to three weeks, and compare the cohort return of the incremental tranche against the base. The precision is poor, but the question is only whether the last tranche clears the hurdle, and poor precision is enough to answer it most of the time.
Allocate by equalising marginal return. Move budget from channels whose last tranche is below the hurdle to channels whose last tranche is above it, until every channel's marginal return sits near the hurdle or the channel is capped by inventory. This is the only allocation rule that maximises total return for a given spend, and it will frequently produce a total that is nowhere near any round percentage of revenue.
Re-estimate on a schedule tied to change, not to the calendar. Curves shift when a creative concept fatigues, when a competitor launches, when a network exits. Unity's ironSource Ads shutdown at the end of this month is a live example: the demand that was buying on that network moves elsewhere and the curves on neighbouring networks steepen for everyone.
The decision rule that survives a board meeting
The objection to all of this is governance. A board cannot approve "spend until marginal return equals the hurdle" as a budget line. The workable compromise is to present the budget as a range rather than a number, with the range derived from the curve estimates: a floor below which the studio is leaving above-hurdle returns on the table, and a ceiling above which the marginal tranche is expected to fall below the hurdle. The percentage of revenue that range represents is reported as an output, for comparability, not set as an input.
Inside the range, the UA team has authority to move. Outside it, either the curves have changed or the hurdle has, and both are reasons for the conversation to happen again rather than for the team to keep spending to a share.
The uncomfortable consequence is that a marginal-return budget will sometimes tell a growing studio to spend less as revenue grows, because the curves have flattened, and sometimes tell a shrinking studio to spend more, because a competitor's exit has steepened them. A team that finds those recommendations unacceptable has, in effect, decided that the budget is a signal to the market rather than an investment, and should say so.
