Payback windows are a cost of capital decision

Studios argue about whether their payback window should be 90, 180 or 365 days as though it were a measurement question, something the analytics team should settle with better cohort curves. It is not. The right payback window is set by one thing: what the studio's cash costs. A team that has never written down its cost of capital has no basis for choosing a window, and the window it has chosen is almost certainly whatever the previous UA lead inherited from a previous employer.

This will strike some measurement leads as finance intruding on their domain. The cohort curve is theirs; the choice of where to draw the line on it is not. The curve tells you when money comes back. The cost of capital tells you how long you can afford to wait for it. Confusing the two is how a studio ends up with a 365-day window it cannot fund and a 90-day window that starves a game with genuinely long-tail revenue.

The mechanism: waiting has a price

Every day between paying for an install and recovering its cost, the studio is lending money to its player base. The price of that loan depends on where the cash came from.

For a studio funded by its own operating cash flow, the price is the opportunity cost of the next best use, which in practice is the next cohort. If a cohort repays in 90 days, the same cash can be spent roughly four times a year. If it repays in 360 days, once. Those are not similar businesses even if their 12-month ROAS is identical.

For a studio running on a recent equity raise, the price is whatever return the investors expect. That is not zero, whatever the board deck implies. Money spent on slow-payback installs is money not spent extending runway, and runway is the thing that determines whether the studio exists to collect the long tail at all.

For a studio using revenue-based financing or a UA credit line, the price is explicit and printed on the term sheet. This is the one case where UA teams usually do know their cost of capital, and, not coincidentally, the case where payback windows tend to be shortest and most disciplined.

What the window actually encodes

A 180-day window says: we are willing to have cash out for six months before it comes back, and we trust our forecast of the curve that far. Both halves matter.

The first half is the funding decision. Multiply monthly UA spend by the window in months and you have the working capital the UA programme consumes at steady state. Illustratively, a studio spending 500,000 pounds a month on a 180-day payback has roughly three million pounds permanently tied up in cohorts that have not yet repaid. Shorten the window to 90 days and half of that comes free. Most studios have never calculated this number, and when they do, the finance lead tends to look unwell.

The second half is the forecast risk. A 365-day window means the campaign decision rests on a prediction of month-twelve revenue made from a few weeks of data. The wider the gap between the data you have and the horizon you are underwriting, the more the model is doing the work rather than the players.

The second-order effect: windows shape the game you build

Here is the trade-off that gets missed. Once a payback window is fixed, it stops being a measurement setting and starts being a product brief.

A short window rewards games that monetise early: aggressive first-session offers, starter packs, early ad density. A long window rewards games that monetise late: deep progression, social systems, seasonal content. Neither is wrong, but the window is chosen by finance and then quietly enforced by UA, and the product team ends up building toward it without anyone having made the decision on purpose.

The reverse also happens. A genre with a long natural monetisation curve, run by a studio with expensive money and therefore a short window, will see UA systematically underbuy its best cohorts. The buyer sees D90 ROAS below target, cuts the campaign, and never learns that month eight would have carried it. The window did not measure the game. It overruled it.

A decision rule

Rather than picking a window from convention, derive it.

Start with the annualised cost of capital. For a self-funded studio, use the annualised return on the fastest-repaying channel you could scale, because that is what the cash would otherwise do. For an externally funded studio, use the return investors are underwriting, or the interest rate on the credit line if there is one.

Then take the cohort curve for the game and find the point where the incremental return from waiting another month falls below the monthly cost of capital. That is the window. Beyond it, the studio is paying more to wait than the wait earns.

An illustrative case. A game's paid cohorts recover 60 percent of cost by day 90, 95 percent by day 180 and 125 percent by day 365. Between day 180 and day 365 the cohort adds 30 points of return over six months, about five points a month. If the studio's money costs more than about five percent a month, which describes a lot of studios on short runway, the marginal wait is not worth it and the window should sit at 180 with a target ROAS near breakeven. If the money costs one percent a month, the window can run to 365 and the buyer should be underwriting the tail.

Three practical consequences follow:

• The window is a per-game figure, not a studio policy, because curves differ by genre.
• It should be revisited whenever the funding situation changes, not when the analytics team gets better data.
• The target ROAS at the window is derived from the window, not chosen alongside it.

Who owns the number

The measurement team owns the curve. Finance owns the cost of capital. The window is the intersection, which means it belongs to neither and needs an explicit owner, usually whoever holds the growth budget. An earlier UA Ledger piece, ROAS targets versus payback windows, made the case that the two settings interact; the point here is that one of them is upstream of the other, and it is not the one most teams start with.

The tell that a studio has this backwards is a payback window that has not moved in two years while the studio has raised money, run low on it, and raised again. Money got more and less expensive across that period. The window that governs how long the studio waits for it should have moved every time.

Related archive reading

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

• ROAS targets versus payback windows
• Automation did not remove judgement, it moved it