Quick answer: A good rewarded UA payout structure pays a small amount for the install and puts most of the money on later in-app events that predict payback. Because you only pay a deep-event reward for the users who reach it, your effective cost per install is a blend of every event payout times the share of installers who complete it – which is the number to set against your target CPI.
To change payouts without hurting retention, model the ladder first: list each event, its payout and its completion rate, and compute the blended cost per install. Then move money from the install toward the event most correlated with your payback window, re-run the model, and test the new ladder on a separate offer before you touch the live one.
One caveat: competition decides how much traffic you actually get. Competing advertisers pay affiliates on a straight CPI basis so the affiliate carries as little risk as possible, and affiliates send volume where their risk is lowest. A down-funnel ladder is not something you can impose on the market. It gives your program the best chance of success only once its CPI component is competitive enough to win traffic in the first place.
Once that CPI component is competitive, paying more for the install is the most expensive way to buy retention. It raises cost on every user, including the ones who leave in the first week. Paying more for a predictive event raises cost only on the users you wanted.
How does a rewarded UA payout structure work?
In rewarded user acquisition, the user earns cash or virtual currency from a publisher's app or offerwall for installing your app and completing actions inside it. That makes it incentivized traffic: the user is paid for the action, so the action you pay for decides who you get. Pay only for the install and you buy a download from someone whose motivation ends there.
An event-based structure splits the payout into a ladder. A typical ladder has three layers:
- Entry event – the install or first open. A small payout that gets the user in the door and confirms the attribution.
- Early engagement events – tutorial complete, a level reached, a feature used within the first few days. These separate people who tried the app from people who started playing it.
- Value events – a milestone reached within a set window, a first purchase, a subscription start. These are the events that look like your paying users.
The publisher shows the user the total reward available across the ladder, so the headline incentive can be large while your average cost per install stays controlled. Every event below the install is paid only for the users who reach it.
How can I model reward payout changes to hit a target CPI without hurting retention?
Build a ladder model: for each event, multiply its payout by the share of installers who complete it, and add the results. That sum is your effective CPI. To protect retention, shift payout from the install toward the event most predictive of payback, keep the blended total at or under your target, and validate on a separate test offer before changing the live campaign.
The formula
Effective CPI = sum of (event payout × completion rate for that event), with the install itself counted for every installer.
You need two inputs per event: the payout you set, and the completion rate your attribution data shows for rewarded installers. Use rewarded-cohort rates, not organic or paid social rates – incentivized users move through a ladder differently.
A worked example (illustrative numbers, not network data)
The table below is a hypothetical ladder built to show the arithmetic. The payouts and completion rates are made up for illustration and do not come from our network data or any advertiser's campaign. Plug in your own.
| Event (illustrative) | Payout per completion | Hypothetical completion rate | Cost per install contributed | What it selects for |
|---|---|---|---|---|
| Install and first open | $0.50 | 100% | $0.50 | Attribution, not quality |
| Tutorial complete | $0.40 | 70% | $0.28 | Users who actually started |
| Level 10 within 3 days | $1.00 | 35% | $0.35 | Early engagement |
| Level 25 within 7 days | $2.50 | 15% | $0.38 | D7-style retention |
| First purchase within 14 days | $8.00 | 4% | $0.32 | Payers |
| Ladder total | $12.40 headline reward | – | $1.83 effective CPI | – |
Compare it with a hypothetical install-only offer at a $3.00 CPI: that offer pays $3.00 for every installer, while the illustrative ladder averages $1.83 per installer, shows the user a much larger total reward and puts more than half of its spend on events only users past the tutorial can reach.
If your target CPI is the install-only price, the ladder leaves headroom on paper. Once the install payout is competitive, spend that headroom where it does the most good – usually by raising the payout on the event most correlated with your payback, not by raising the install payout further.
Competition decides whether the ladder gets traffic
The ladder is the structure that gives a rewarded program the best chance of success, but you cannot impose it unilaterally. Competing advertisers pay on a straight CPI basis so affiliates carry as little risk as possible, and affiliates send volume where their risk is lowest. An affiliate comparing your ladder with a straight CPI offer is comparing a payout that depends on how your users behave after the install with one that does not.
So set the CPI component first, and make it competitive enough to win traffic on its own. The illustrative $0.50 install payout above shows the arithmetic, not a price that would win traffic against a $3.00 straight CPI offer. Once the install payout is competitive, add the down-funnel events on top of it: the ladder's job is to pay more for the users who stay, not to pay less for the install.
What the model does not tell you
Completion rates are not fixed. A larger deep-event reward pulls more users toward that event, raising its completion rate and your effective CPI. Treat the first model as a hypothesis and re-run it on the test cohort's actual rates.
The model also does not tell you which window your payback is judged on. That depends on how your app makes money, and we cover how advertisers choose cohort windows in our guide to rewarded app CPI rates. Set your deepest paid event inside the window your finance team actually measures.
Why is paying more for the install the most expensive way to buy retention?
Because most installers are gone within a week, across all games and not only rewarded ones. In the GameAnalytics 2026 Mobile and PC Gaming Benchmarks, median mobile D1 retention in 2025 was about 22%, while median D7 retention was just under 4%. Those are industry-wide medians across thousands of titles, not rewarded-traffic figures, but the shape holds: the drop between the first day and the first week is steep.
When you raise the install payout, you raise cost on every one of those users, including the large majority who leave before they are worth anything. When you raise the payout on a day-seven milestone, you raise cost only on the minority who reach it – the users who look like your retained base.
Raising a mid-ladder payout also makes the offer more attractive to users willing to play, not users who plan to stop at the install. The trade-off is volume timing. A heavier deep-event payout can lower install volume at first, because some users skip offers whose rewards feel far away. That is why the early engagement events matter – they give users a reachable reward in the first session while keeping most of the money on the events that predict payback.
Which in-app events belong on the payout ladder?
Pick events that pass all of these tests:
- Predictive. Users who complete it pay back at a meaningfully higher rate than users who do not, in your own historical cohorts. If you cannot show the correlation, it is a guess, not a ladder step.
- Reachable inside your payback window. A milestone most retained users only hit after your measurement window closes will not help you hit early ROAS targets.
- Verifiable server-side. Fire it from your backend or a validated SDK event, not a client event that can be spoofed. Rewarded traffic pays users for actions, so the events you pay on need to be the hardest ones to fake.
- Spaced. Each step should filter out a real share of users. Two events almost everyone completes together are one event paying twice.
- Understandable to the user. "Reach level 25 within 7 days" is a goal a player can plan around. "Achieve 40 minutes of engaged session time" is not.
A purchase is the strongest value signal for in-app purchase games, but do not make it the only deep step. Few rewarded users buy, and a ladder whose money all sits on a purchase looks unreachable. Pair it with a progression milestone.
For apps that monetize through ads rather than purchases, progression and return-visit events usually do the work a purchase does for paying games, because ad revenue accrues with sessions over time.
What should you change first when install quality drops?
When a quality review shows rewarded cohorts are not paying back, the reflex is to cut the CPI or pause the budget. Work through this list first, in order:
- Check the data before the payout. Confirm the events are still firing correctly and that attribution windows match what you agreed. A broken event looks exactly like a quality drop.
- Check where the drop sits on the ladder. If completion falls at the tutorial step, the problem is the first-session experience or the offer description, not the payout. If it falls at the deep step, the deep event may be too far away.
- Move money, do not just remove it. Lower the install payout, as far as competition allows, and raise the payout on the step where quality users separate from the rest. This often protects volume better than an across-the-board cut.
- Tighten the window on the deep event. If users are reaching the milestone but too slowly to pay back inside your window, shorten the time limit instead of raising the reward.
- Ask your network for a breakdown by traffic source. Quality issues are often concentrated in one or two sources. Isolating them is cheaper than repricing the whole offer.
- Then, and only then, cut the blended payout. If the ladder is sound and the cohort still does not pay back, the total is too high for your economics.
Which channels drive quality installs with good retention for gaming?
No channel guarantees retention; what you pay for decides most of it. Paid social and other bid-based channels let you optimize delivery toward in-app events, while rewarded lets you pay directly on those events. Both reward the same discipline: define the event that predicts payback, and stop buying installs as the goal.
Adjust's Gaming App Insights 2026 report frames it plainly: acquisition costs are rising, which makes retention a core focus in 2026.
Compare channels on cost per user who reaches the payback event, on the same window. Rewarded tends to fit apps with a clear progression system a user can be rewarded for working through; apps whose value is hard to express as a milestone are harder to price on a ladder.
How do you roll out a new ladder without breaking a live campaign?
Treat a new ladder as a new offer, not an edit to the old one.
- Launch it alongside, not in place of, the current structure. Ask your network to set the new ladder up as a separate offer or tracking path, so its cohorts are measured cleanly and the live campaign keeps running while you learn.
- Compare at the same event and window. Judge both on cost per user who reaches your payback event, on the same cohort window, and only after the deepest events have had time to complete.
- Tell publishers what changed. They present your offer to their users, and a clear description of each step and its time limit helps completion.
Our guide to testing an offer with no track record covers how to size a first test so the read is meaningful.
If you want help modeling a ladder against your target CPI and payback window, talk to our advertiser team. We can set up a test offer alongside your current structure so you can compare cohorts before you commit budget.
FAQ
What is a rewarded UA payout structure? It is how an advertiser splits the money paid for a rewarded install across events – the install itself, early engagement steps like a tutorial or level, and value events like a milestone or first purchase. Users earn rewards for each step, and the advertiser pays each event payout only for the users who complete it.
How can I model reward payout changes to hit a target CPI without hurting retention? Multiply each event's payout by the share of installers who complete it and add the results to get your effective CPI. Keep that total at or under your target, move payout from the install toward the event most predictive of payback, and validate the new ladder on a separate test offer before changing the live campaign.
Should I pay more for the install to get more volume? Not once it is competitive. Competing advertisers pay affiliates a straight CPI, and affiliates send volume where their risk is lowest, so your install payout has to be competitive enough to win traffic before any ladder can work. Past that point, a higher install payout raises cost on every installer, including the many who leave in the first week, and it attracts users who plan to stop at the install. If you need more volume then, make the early engagement steps more reachable or raise a mid-ladder payout instead.
Which events should a rewarded gaming offer pay on? Events that predict payback in your own cohorts, can be reached inside your payback window, fire server-side, filter out a real share of users at each step, and make sense to a player. Progression milestones with a time limit and a first purchase are the most common deep steps for gaming offers.