Rewarded user acquisition has a source problem, and the version of this page I published in May 2026 was part of it.

It claimed walled gardens own about 60% of paid UA while listing components that added to 70%. It gave a rewarded CPI ceiling of $3.50 and then, two sentences later, a rewarded CPI of $10 to $20 for midcore. It welded two vendor case studies measuring different metrics into a single continuous performance range. And it described Mistplay’s audience as the near-inverse of what that platform’s own marketing is built on.

I have rewritten it against what the four cited pages actually say. The channel is real and I recommend it. Most of the numbers attached to it in public are not.

Key Takeaways

  • No source publishes a rewarded-specific CPI band. I looked. What exists is general paid UA pricing from adjoe, and rewarded campaigns priced against it deal by deal. Any article quoting a rewarded CPI range is quoting itself
  • The market-share figure everyone cites is from 2022. Meta and Google each took 30% of paid UA and Apple Search Ads took 10%, per Matej Lancaric, describing 2022. Those three are all walled gardens, so they sum to 70%
  • Price rewarded against the canonical rule, meaning CPI at 30% to 70% of projected LTV. The 3:1 shorthand is the 33% point of that same band
  • Every retention and ROAS uplift figure in this space comes from a party selling the channel, with no published sample size or methodology. They are labelled as such throughout this article
  • The operator edge is calibration and isolation. Reward tied to a survival-correlated event, rewarded cohorts measured separately from walled-garden cohorts, and payback read past the reward window

What Rewarded UA Actually Means in 2026

Rewarded user acquisition is a model where a player voluntarily installs and engages with your game in exchange for a third-party reward. The advertiser pays on a CPI or CPE basis, the platform takes a margin, and the publisher app compensates the user. Do not confuse it with a rewarded video ad running inside your own game, which is monetisation rather than acquisition.

Three formats dominate.

  • Offerwall UA. Users browse a list of game offers inside a publisher app and choose what to try. ironSource Offerwall, Tapjoy and AdGate operate here
  • Recommendation-driven rewarded networks. An engine matches games to user profiles and pays out on install or mission completion. Gamelight is the reference product
  • Playtime-based rewards. Users earn continuously for time spent in your game rather than for the install alone. adjoe’s Playtime is the category-defining format

The model works because it inverts the consent dynamic. Traditional UA pushes an install at someone scrolling past. Rewarded UA gets the user to raise their hand first. Whether that produces a better player is a question your cohort data answers, and the honest state of the public evidence is covered further down.

If you have followed our UA CPI benchmarks for 2026, the framing there applies here without modification: a CPI means nothing until you know the LTV it is being measured against.

The Market Share Figure, With Its Real Date

Almost every rewarded UA article opens by sizing the walled gardens, and almost all of them cite the same sentence. Here it is with its vintage attached.

Matej Lancaric writes: “In 2022, Meta and Google each captured 30% of the paid user acquisition market, while Apple Search Ads contributed 10%.” That is a 2022 figure. I could not find a 2026 update from a source I would cite, and the previous version of this page presented it under a 2026 label, which was wrong.

The second thing to notice is arithmetic. Apple Search Ads is a walled garden. So the three named platforms are 70% of paid UA in 2022, and the “walled gardens own about 60%” line this page used to carry contradicted its own components. That 60% figure does not appear anywhere in the cited source.

None of this changes the strategic point. Two or three platforms took the large majority of paid UA spend four years ago and there is no evidence they have been displaced. Rewarded UA is not competing for that position. It is the layer you stack on top when walled-garden CPIs stop clearing your payback window. For the full allocation framework, see our guide on diversifying mobile game UA channels, and for the other high-leverage diversification play, our TikTok ads playbook.

Rewarded UA vs Walled Gardens: Where Each Wins

DimensionWalled-garden paid UARewarded UA
Scale ceilingVery highMedium, growing
Pricing modelCPI, CPA, value-based biddingCPI or CPE on completed events
Price levelPublished ranges existNegotiated per deal, no published band
User intent at installLow to mediumHigh, the user opts in
Signal after ATTDegraded, SKAN and modelledDeterministic, event-based
Best forReach, lookalike scaling, remarketingInstall efficiency, incremental volume
Main riskCPI inflation, creative fatigueReward-induced churn if miscalibrated

Notice what is missing from the price row. The previous version of this table claimed rewarded runs “20-40% lower on comparable genres”. That claim is not reproducible even from the article’s own figures, which put walled gardens at $1.50 to $5.00 and rewarded at $0.80 to $3.50, a discount of 47% at the bottom and 30% at the top. Neither the range nor the figures behind it had a source. Both are gone.

The same applies to the retention row, which claimed a +15-30% D7 lift. That figure is Gamelight’s, appears on a Gamelight page with no dataset, sample size or methodology attached, and Gamelight sells rewarded UA. It is not a measurement and it does not belong in a comparison table where every other row reads as fact.

The CPI Question, Answered Honestly

Here is what I can source. adjoe publishes general paid UA pricing for 2026 as follows: iOS at “$1.5 – $5+ range globally, with higher costs in Tier 1 markets”, Android at “roughly $1.5 – $4+ depending on region”, and high-value genres at “CPIs ranging from $10–$20+ per install”.

Read those labels carefully, because the previous version of this page did not. The $10 to $20 band is adjoe’s figure for general paid UA in high-value genres. It was republished here as the price of a rewarded install for 4X and midcore RPG, sitting three paragraphs below a rewarded ceiling of $3.50. Both could not be true. The displacement, rather than the invention, is what made it survive a casual fact-check: the number does appear in the source, with a different label.

So the usable position is this. Rewarded pricing is negotiated per platform, per geo and per genre, and the general market figures above are what you are benchmarking a rewarded quote against. If a rewarded platform quotes you meaningfully below the general band for your genre and geo, that is the discount, and it is a number you obtain from a rate card rather than from an article.

For where those general figures sit against primary CPI datasets, and why published CPI benchmarks disagree by an order of magnitude, our CPI benchmarks article covers the methodology. The one-line version: never compare a volume-weighted global blend with a Tier 1 estimate.

Setting the Target: LTV, Payback, and the 3:1 Shorthand

Our planning rule across the blog is that CPI should land at 30% to 70% of projected lifetime value, tighter when you need fast payback or carry little risk. The 3:1 LTV:CPI ratio operators quote is the 33% point of that band, so the two rules agree and you can use either.

What changes on rewarded channels is how you measure both sides.

  1. Cohort past the reward window. Rewarded users front-load engagement to claim their reward, so a D7 read is systematically optimistic. Read D30 as your minimum, and D60 to D90 on anything with a long monetisation ramp
  2. Separate reward-driven return from organic return. A user who only comes back when the offerwall re-prompts them is a rented session, and their retention curve will collapse the day the campaign stops
  3. Put the reward cost in the CPI line. Some platforms expose the reward economics to advertisers and some do not. Where they do, it belongs in your acquisition cost rather than in a footnote

If you are running a 4X or midcore title, the payback discipline in our midcore UA and monetisation guide applies with extra force here, because a long monetisation ramp and a front-loaded engagement curve are the worst possible combination for a hasty read.

Reward Mechanics and the UX Trade-off

This is where most studios get it wrong on the first pass. Too small a reward fails to motivate, and too large a reward recruits people who collect and leave. Gamelight’s own guidance says as much, in exactly those terms: “Balance reward value. Too small won’t motivate, too large distorts behavior.”

The structure I have seen hold up is a tiered funnel.

  • Install-level reward, small. Enough to move someone from browsing an offerwall to installing, no more. I previously published a $0.05 to $0.15 equivalent here. That figure was precise to the cent and sourced nowhere, so it is gone. The right level is what your platform’s own tests show clears the install without inflating volume you cannot retain
  • Mid-funnel reward, medium. Paid at the moment your cohort data shows correlates with day-3 survival. That moment is game-specific and you already have the data to find it
  • Deep-engagement reward, large. Paid at first purchase, or at whatever level your economy treats as the point of no return

The second trap is misalignment between the offerwall promise and your game. If the offerwall advertises a reward for reaching level 5, and level 5 takes three hours behind a paywall, you get refund requests, one-star reviews and deprioritisation by the platform. Set the goal at the actual median session pattern of your existing players. Our retention strategies guide covers the cohort modelling that makes that calibration possible.

Ready to design a rewarded UA strategy that protects unit economics? Book a call to review your UA mix.

Choosing Platforms

There is no universal best platform, and the right stack depends on genre, geo and what mediation you already run. What follows is my operator read, and I have removed the audience characterisations that were in the previous version because I could not source them from a page I would cite.

  • Gamelight. Recommendation-engine driven, prices per install, positioned on casual and hybrid-casual. Useful as an alternative to mediation-led offerwalls
  • adjoe Playtime. The reference product for time-based rewards. Structurally suited to games with real session depth and a progression curve worth spending time on, and structurally poor for games whose loop is over in ninety seconds
  • ironSource Offerwall. The default if you already run LevelPlay. The value is integration and reporting continuity rather than a performance edge
  • Tapjoy, AdGate, Mistplay. Worth testing against your own genre. Each has a distinct publisher base and audience composition, and I would ask each of them directly for their current audience breakdown rather than trusting a third-party description, including the one this article used to carry

The operating rule matters more than the shortlist: one new platform per quarter, four to six weeks of measured spend, keep or kill on D30 retention and payback against your blended baseline. Spreading a $50K test across five platforms in month one is the most common mistake in this channel, and it produces five datasets too small to read.

What the Vendor Case Studies Actually Say

Rewarded UA marketing leans on case studies, and they are worth reading with the labels intact. From adjoe’s own pages, on Playtime:

  • Homa Games: “Up to 130% increase in D7 ROAS”
  • Trailmix Games: “+30% installs”
  • Unico Studio: “+30% ROAS”

Those are three clients measuring three different things, and adjoe presents them individually. The previous version of this page compressed them into “case studies show 30-130% performance uplift”, which turns a set of unrelated single-client results into what reads like a distribution. It also included a fourth case, 4399 growing a title “by 32.7% with Playtime”, where the source never names the metric being grown. A figure carried to one decimal place with no unit attached is worse than no figure, so it has been removed.

Treat all four as existence proofs that the format has worked somewhere, which is genuinely useful when you are deciding whether to run a test, and as evidence of nothing about what it will do for you.

Measurement, Fraud, and the Signal Advantage

Rewarded UA has one underappreciated structural advantage after ATT: the model is deterministic by design. The user opts in, the platform validates the install and the qualifying event, and reporting flows back without depending on IDFA or on SKAdNetwork postbacks. For a studio whose iOS measurement has degraded to modelled conversions, that alone can justify a test.

Three operational rules still apply.

Fraud controls are not optional. adjoe cites an industry estimate that “up to 25% of global mobile ad spend is impacted by fraud”, attributing it to Anura, a vendor selling fraud detection. Both parties in that chain sell against the problem they are sizing, so treat the figure as an argument rather than a measurement. The operational advice survives the caveat: reputable rewarded networks have native anti-fraud, and you layer your own MMP-level rules on top regardless.

Measure rewarded cohorts in isolation. Never blend rewarded and walled-garden installs in the same dashboard. The engagement curves have different shapes, and blending them will misprice your LTV in both directions at once.

Treat rewarded as a line item, not a side experiment. The pattern in our complete mobile game growth strategy playbook puts rewarded alongside paid, owned and earned rather than in a footnote, which is also how it should appear in your budget.

What We Could Not Verify

This article publishes fewer numbers than the version it replaces, and three of its four sources are companies selling rewarded UA. Here is what I looked for and what I removed.

  • “Walled gardens own about 60% of the paid UA market.” Not in the cited source, and contradicted by the source’s own components, which sum to 70%. Removed, and the surviving figure is now labelled with its real year, 2022
  • A rewarded-specific CPI band. The previous “$0.80 to $3.50” had no source, and sat in the same section as a rewarded midcore CPI of $10 to $20 that was actually adjoe’s figure for general paid UA. Removed. No source I found publishes rewarded pricing as a band
  • “20-40% cheaper than walled gardens.” Not reproducible from the article’s own figures, which implied 30% to 47%, and sourced nowhere. Removed
  • “Case studies show 30-130% performance uplift.” Fabricated as a range by joining a D7 ROAS result to an install result across different clients. Replaced with the individual cases, each with its own metric
  • “4399 grew Legend of Mushroom by 32.7%.” The source names no metric. A one-decimal figure with no unit is not usable. Removed
  • “Mistplay skews male, mid-core.” This was wrong, and wrong in the direction that matters, since Mistplay’s commercial positioning is built on the opposite audience. I removed the characterisation rather than replace it, because the platform’s audience data is published on a domain this blog does not cite for figures and I could not verify a breakdown from a primary page. Ask the platform directly
  • “82% of developers say reward-based UA outperformed traditional UA, and 68% saw improved ROAS.” adjoe attributes this to “a 2025 survey” without naming the survey, the organisation or the sample size. It is a vendor citing an unnamed survey about the vendor’s own category. It is also worth knowing that a separate 82% circulates in this space, from Lancaric, about rewarded video as the most popular ad format among developers, which is a monetisation statistic rather than a UA one. Two different 82% claims, two different subjects, routinely conflated. Removed from the headline and the conclusion of this article, where they were load-bearing
  • “+15-30% higher D7 retention on rewarded installs.” Gamelight’s own figure, on a page that cites no dataset for any of its numbers. Removed from the comparison table and labelled as a vendor claim wherever it is mentioned
  • “$0.05 to $0.15 install reward.” Precise to the cent, sourced nowhere. Removed
  • “Up to 25% of mobile ad spend impacted by fraud.” Kept, with its chain of attribution stated, because the operational advice does not depend on the number being exact

None of the four sources cited here is a measurement house. Gamelight and adjoe sell rewarded UA. Lancaric is an independent practitioner writing from experience rather than from a dataset. That is the entire published evidence base for this channel that I was able to verify, and it is thinner than the volume of content about the channel would suggest.

Conclusion

Rewarded user acquisition is a legitimate channel with a real structural advantage: deterministic measurement in a market where everything else has been degraded by ATT, and a user who chose to install rather than being interrupted into it. That advantage does not need inflated numbers to be worth testing.

What the channel lacks is published evidence. Pricing is negotiated rather than listed, performance claims come from the companies selling it, and the case studies are single-client results presented as categories. So run it the way you would run any channel with no reliable benchmark: one platform at a time, four to six weeks of measured spend, cohorts kept separate from your walled-garden traffic, payback read past the reward window, and a written threshold agreed before the money goes out rather than argued about once it is gone.

Treat rewarded UA as neither a magic bullet nor a low-quality channel. It is a high-intent, deterministic install source that demands the same operator rigour as any other, and it pays that rigour back in signal quality your other channels can no longer give you.

Want an outside read on your UA mix before you commit budget? Get in touch or explore mobile game consulting.

Sources