The economics of buying your own players back

Re-engagement is the campaign type that most reliably produces a beautiful dashboard. Cost per returning user comes in at a fraction of the cost of a new install, conversion rates are several times higher, and the users who come back already know how to play. The claim in this piece is that most of that beauty is borrowed. A large share of the players a re-engagement campaign "wins back" were on their way back regardless, and the campaign's true economics depend on a number almost nobody measures: the natural return rate of the segment being targeted.

This is not a dismissal of retargeting. Done against the right segment with the right control, it is one of the few places left in mobile UA with genuine unpriced value. It is a dismissal of the way it is usually reported.

Why the cheap conversions are the least valuable ones

The mechanism is selection. Players who lapsed three days ago are far more likely to return on their own than players who lapsed thirty days ago. They are also far cheaper to convert with an ad, because they are still thinking about the game, still have it installed, and still have the muscle memory. A network optimising for the lowest cost per re-engagement will, correctly by its own incentives, concentrate spend on the recent lapsers. The result is a campaign that looks extraordinarily efficient and consists disproportionately of players who did not need to be bought.

Adjust's report in March on gaming app sessions made the broader point that retention has become a core growth strategy, with sessions rising sharply across strategy and casual. The more a studio invests in retention mechanics, live events and notifications, the higher its natural return rate, and the more of any retargeting campaign's reported wins are, in truth, wins of the retention team's making.

The second-order effect that gets missed is what this does to acquisition measurement. Players brought back by retargeting are usually still attributed to their original install source for lifetime value purposes. Their late revenue lifts the LTV of the original acquisition cohort, which makes the network that acquired them look better than it was, which raises the bids the team is willing to pay that network for new users. Retargeting spend, in other words, silently subsidises acquisition bids, and the subsidy is invisible unless someone separates re-engaged revenue from organic-return revenue at the cohort level.

The number that decides everything

The economics reduce to a single comparison. For a given segment, the campaign is worth running only if the cost of reaching the segment is less than the incremental revenue from players who returned because of the ad and would not have otherwise.

An illustrative worked example. Take 100,000 players lapsed between fourteen and thirty days, with a natural thirty-day return rate of 8 percent. Left alone, 8,000 come back. Run a re-engagement campaign against the whole segment, and 12,000 come back at a media cost of 30,000. The dashboard reports 12,000 re-engagements at 2.50 each. The incremental count is 4,000, and the true cost is 7.50 per incremental returning player. If a returning player in this segment is worth 9.00 in expected future net revenue, the campaign clears. If the segment's natural return rate had been 11 percent instead of 8, the incremental count falls to 1,000, the true cost rises to 30.00 and the campaign is deeply underwater while still showing 2.50 in the platform interface.

The lesson is that the natural return rate is not a detail; it is the whole answer. Small changes in it move the true cost by an order of magnitude, and it varies enormously by lapse window, by player spend tier and by what the retention team happened to ship that week.

A decision rule built on a hold-out

The framework has three parts, and the first is non-negotiable.

Hold out. Every re-engagement audience should have a randomly selected slice, on the order of ten to twenty percent, that is eligible for the campaign but suppressed from it. The return rate of the hold-out is the natural return rate. Without it, the incremental figure is a guess, and the guess is systematically optimistic because the network has chosen the audience for exactly the players most likely to return anyway. On iOS, the AdAttributionKit re-engagement postbacks give the measurement a cleaner footing than SKAN ever did; this desk covered the setup in "AdAttributionKit Re-Engagement Testing After iOS 26" last year, and the hold-out logic sits above the attribution layer regardless of platform.

Segment by lapse window and value tier, and bid by incremental cost, not reported cost. Recent lapsers will often show a high natural return and a low incremental lift; they are cheap to hit and frequently not worth hitting. Deep lapsers, sixty days and beyond, have a low natural return and a low conversion rate, so the reported cost looks terrible and the incremental cost can be reasonable. High-value lapsers justify a far higher incremental cost than the segment average. The right shape for most games is a campaign that spends little on the segment the network wants to sell and more on the one it does not.

Set the ceiling from expected net revenue of a returning player in that segment, after platform commission, and treat the incremental cost from the hold-out as the price. If the price exceeds the ceiling, stop or move the window.

What retargeting can hide

The last point is about what re-engagement spend does to a studio's understanding of its own game. A re-engagement campaign that consistently "works" against recent lapsers is a mechanism for masking a retention problem. Players are leaving at a rate the game cannot sustain, and paid media is patching the hole at a cost that appears small because it is measured against the wrong baseline. The retention team sees the day-thirty curve holding and concludes the game is fine.

The hold-out fixes this too. If the suppressed slice returns at almost the same rate as the treated one, the campaign is not the reason players come back, and the studio can spend the budget on the thing that is. If the suppressed slice returns at a much lower rate, the game genuinely needs the reminder, which is useful to know before the network's price for reminders goes up.

Related archive reading

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

• The organic baseline you are not tracking
• Payback windows are a cost of capital decision