D7 ROAS targets punish long-retention games

A single D7 ROAS target, applied across a portfolio or even across channels within one game, is not a neutral discipline. It's a filter, and what it filters for is the speed of monetisation rather than the size of it. Games whose revenue arrives late and long get systematically underfunded by that filter, and those games now cover most of what leads the revenue charts: strategy, 4X, social casino. The discipline feels prudent; it's a bias with a spreadsheet attached.

The counter-argument is real and worth stating plainly. Early signals are the only signals a buyer has when deciding whether to keep a campaign alive tomorrow, and waiting for D90 isn't an option. All true. The error sits in setting D7 as a fixed acceptance threshold that ignores the shape of the curve behind it.

The mechanism: a D7 target is a hidden multiplier

Nobody actually cares about D7 revenue. Finance cares about payback within some horizon, typically twelve to eighteen months for a live-service title. The D7 target exists because someone divided the payback goal by an assumed multiplier: if D7 ROAS of 20% historically became 120% by D365, the multiplier is six, and the D7 gate goes in at 20%.

Two things go wrong. The multiplier is a property of a specific game, at a specific stage, on a specific channel mix. Applying it elsewhere imports that game's curve shape into a game with a different one. And the multiplier for a slow-burn title is both higher and more uncertain. A game that reaches 120% D365 from 12% D7 has a multiplier of ten, and the confidence interval around ten is much wider than around three.

Finance responds to uncertainty the way finance always does: it discounts. A slow-curve game gets a haircut on its multiplier, which raises the D7 target it must clear, which cuts spend. The cut lands on cohorts whose value was going to arrive in months four through twelve. The game's long tail is the entire commercial reason it exists, and the target refuses to pay for it.

An illustrative example

Consider two hypothetical titles, both reaching an illustrative 120% ROAS at D180.

• Game A, a casual puzzle title with heavy early ad monetisation, shows 40% at D7. Its D7 to D180 multiplier is three.
• Game B, a base-building strategy title, shows 15% at D7. Its multiplier is eight.

A portfolio D7 target of 25% keeps every Game A campaign alive and kills most Game B campaigns, despite identical D180 outcomes. Worse, if the buyer on Game B tries to hit 25% at D7, the only campaigns that clear it are the ones finding unusually fast spenders, and those cohorts tend to have shorter tails. The buyer meets the target and the D180 outcome deteriorates. The target has changed the game's player base, not just its reporting.

This is the second-order effect. When you set a D7 event as the optimisation goal in a network's bidding system, the network learns to find users who behave like fast D7 payers. Over a quarter, the cohort mix shifts toward them, the tail flattens, the historical multiplier no longer holds, and the D7 target now looks even more justified because the curve behind it really has got shorter. The target manufactures the evidence for itself.

What the market context adds

The recent industry reports from Sensor Tower and AppsFlyer both describe the same strategic shift: away from new-user volume and toward lifetime-value expansion, with downloads softening while spend per player rises. Both point the same way: growth is coming from extracting more from each player over longer periods, not from acquiring more players cheaply.

A measurement regime built for the hyper-casual era, where D7 really did capture most of a cohort's value, now runs against a market whose leading titles by revenue are long-retention products. As Sensor Tower and others have reported, the month's charts remain dominated by Last War: Survival, Whiteout Survival, Honor of Kings, Royal Match. None of those curves are steep at D7.

A framework that keeps the early signal without the bias

The fix isn't to abandon D7. It's to change what D7 gets to decide.

First, use D7 as a rejection filter rather than an acceptance filter. Set a floor below which a campaign is clearly failing regardless of curve shape, and kill below it. Above the floor, D7 gets no vote on scaling. Scaling decisions move to the earliest horizon at which the curve's shape has stabilised, which for many strategy titles is somewhere around D30 to D45, and which you can identify by finding the day after which cohort rank order stops changing.

Second, set the target per curve shape, not per portfolio. Estimate the multiplier from the game's own mature cohorts, then widen it by the observed dispersion across those cohorts. A game with a consistent multiplier earns a tight target. A game with a noisy multiplier earns a looser target and a smaller maximum daily budget, so that uncertainty shows up as a spend cap rather than as a punitive gate.

Third, watch the multiplier itself as a health metric. If the D7 to D90 ratio on recent cohorts is drifting down while D7 ROAS is stable or rising, the network is finding faster payers with shorter tails. That is the signature of the self-fulfilling target, and it should trigger a review of the optimisation event, not a celebration of the D7 number.

Fourth, for a genuinely new title with no mature cohorts, borrow the multiplier from the closest genre analogue you have, and state the borrowing in the review deck. A borrowed multiplier is a hypothesis. Recording it as one means the team knows to replace it the moment the game's own D60 data exists.

The decision this leaves with the growth lead

A D7 gate is cheap and legible and easy to defend in a finance review, which is why it survives. The cost is invisible because it appears as campaigns that were never scaled and players who were never acquired. For a long-retention title, the honest conversation is that patience is a budget line, and the team has to choose how much of it to fund. A D7 target does not remove that choice. It makes it by default, and it makes it against the game.

Related archive reading

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

• The hidden fixed costs of a paid UA operation
• UA metrics explained: CPI, ROAS, LTV and payback