What is User Acquisition Optimization? A Complete Guide for Mobile Game Studios
How modern mobile game studios use predictive LTV, forward ROAS forecasting, and systematic budget allocation to scale UA spend profitably. Estimated reading time: 14 minutes.
What you will learn on this page
This guide covers the full discipline of user acquisition optimization for mobile game studios: what it is, how it works step by step, the most common mistakes that account for the largest share of wasted UA budget, the landscape of tools mature studios use, and a plain-English glossary of the metrics that matter. If you run UA at a mobile studio spending $50K/month or more, this page is written for you.
What is User Acquisition Optimization?
User acquisition optimization (UA optimization) is the discipline of maximising the return on every dollar a mobile game studio spends acquiring new players, by combining predictive lifetime value (LTV) forecasting, real-time ROAS signals, and systematic budget allocation to continuously improve campaign efficiency.
The four dimensions that define modern UA optimization:
Predictive LTV
Knowing a cohort's long-run revenue within days of install, not months, so budget decisions are made on forward signal rather than trailing data.
Campaign ROAS accuracy
Forecasting D90/D180/D365 ROAS before you have waited 90 days to see the outcome.
Budget allocation
Systematically shifting spend toward channels, creatives, and geos that maximise incremental return at portfolio level.
Feedback loop speed
Compressing the cycle from data to insight to action from weeks to hours, so campaign adjustments happen before millions in budget are committed to underperforming cohorts.
The concept is not new: direct response advertisers have used cohort analysis for decades. What changed in mobile gaming is the combination of privacy-first attribution (ATT, SKAdNetwork, Privacy Sandbox), the collapse of user-level targeting, and the availability of ML models trained specifically on mobile gaming revenue curves. These forces made cohort-level prediction the primary competitive lever in UA, and created the demand for purpose-built UA optimization tooling.
How User Acquisition Optimization Works
Most studios still manage UA through a trailing ROAS window. They wait 30, 60, or 90 days to see whether a campaign was profitable, then adjust. By the time that signal arrives, months of budget have already been committed. Modern UA optimization inverts this model.
Cohort-level LTV prediction at Day 1 to 7
Machine learning models trained on a studio's own historical cohort data generate LTV curves within the first 1 to 7 days of install. Well-trained models produce D90, D180, and D365 predictions with median error rates below 10% by Day 7, giving UA managers a forward view of campaign profitability before most budgets have been committed. The key variable is training data depth: models trained on 90+ days of cohort history consistently outperform those relying on industry averages or short windows.
Campaign ROAS forecasting
LTV predictions are mapped to campaign spend to generate forward ROAS curves. A campaign running at $50K/month can be assessed for whether it will return 2x, 3x, or 4x by D365 within the first week of running, rather than after waiting for the revenue to mature. This replaces gut-feel and trailing windows with a probabilistic forecast that improves over time as more cohort data accumulates.
Systematic budget allocation
UA optimization platforms surface budget signals across every active campaign simultaneously, ranking channels, creatives, and sub-geos by predicted efficiency. In Kohort's customer base, studios that adopt forward ROAS reallocation typically shift 15 to 30% of spend toward higher-return cohorts within the first 60 days, without increasing total budget. The gain comes from eliminating the lag between signal and action, not from spending more.
Continuous model improvement and compounding
As more cohort data accumulates, model accuracy compounds. Specialist ML-based cohort forecasting platforms materially outperform simple D30 extrapolation at the D365 horizon. In Kohort's back-testing across $6B+ of UA spend, gradient-boosted models trained on 6+ months of studio-specific cohort data achieve D7-to-D365 prediction accuracy above 90% on cohorts of sufficient size, where simple linear extrapolation from D30 data sits closer to 70%. The gap widens for games with heavy long-tail LTV, where linear extrapolation breaks down structurally. This creates a durable competitive advantage: the longer a studio runs systematic UA optimization, the more accurate its forward signals become relative to competitors still using trailing windows.
Step 1 above is the cohort-level LTV prediction layer. For a deeper walkthrough of how the underlying models are built, trained, and validated, see our companion guide: What is Cohort Forecasting?.
Paid ROAS vs Blended ROAS vs Incremental ROAS vs Predicted ROAS
One of the most common sources of poor UA decisions is optimising for the wrong ROAS metric. The table below clarifies each.
| Metric | What it measures | The blind spot | Best used for |
|---|---|---|---|
| Paid ROAS (trailing) | Revenue attributable to paid channels over a fixed window (D30, D90). | Looks backward; by the time you see it, budget has already been committed. | Channel-by-channel retrospective reporting. |
| Blended ROAS | Total revenue (paid + organic) divided by total paid spend. | Masks which campaigns are actually working; strong organic can hide weak paid. | Portfolio-level health checks, board reporting. |
| Incremental ROAS | Revenue that would not have happened without the ad spend. | Requires holdout tests or geo-based experimentation; harder to measure at scale. | True ROI decisions, channel incrementality testing, scaling decisions. |
| Predicted D365 ROAS | Forward-modelled D365 return using cohort LTV prediction (generated at Day 7). | Requires accurate LTV models and sufficient training data; early predictions carry uncertainty bands. | Real-time budget allocation, campaign scaling, M&A diligence. |
Predicted D365 ROAS, generated within the first week of a campaign, is the metric that most directly informs where to put tomorrow's budget. Leading UA optimization platforms surface this at campaign level as a standard output, not a bespoke analysis.
Common Mistakes in User Acquisition Optimization
The following mistakes consistently account for the largest share of wasted UA budget in mobile gaming.
Optimising on trailing 30-day ROAS as the primary bid signal
A D30 ROAS figure tells you what a cohort returned in its first month. For games with D180+ LTV tails (mid-core, strategy, RPG, hybrid casual) this is a small fraction of total cohort value. Studios that bid against D30 ROAS consistently over-invest in campaigns that look good early and underperform at D180, and under-invest in campaigns with slow early curves that deliver strong long-run return. The fix is introducing predicted D180 or D365 ROAS as a parallel signal alongside trailing data.
Treating organic and paid cohorts as a single pool
Organic installs inflate blended ROAS. When paid and organic users are not segmented at the cohort level, studios systematically overestimate the efficiency of their paid spend. Blended ROAS can materially overstate paid campaign efficiency in apps with strong organic volume. AppsFlyer's published analysis of organic uplift notes that the magnitude varies significantly by app, platform, and country, with one cited example showing organic uplift four times higher on iOS than on Android for the same product. In Kohort's customer base, the average overstatement when paid and organic cohorts are not segmented sits in the 20 to 30% range, but the variance across apps is large enough that the only safe approach is per-app measurement. The fix is cohort-level MMP segmentation from day one.
Running UA without genre-calibrated LTV benchmarks
LTV curves vary dramatically by game genre and monetisation model. Hypercasual games monetise heavily on D1 to D7, predominantly through IAA. Mid-core, strategy, and 4X games sit at the opposite end: monetisation curves with long, heavy IAP-driven tails where a meaningful share of D365 LTV materialises after Day 90, in some sub-genres a majority. Using a single cross-genre benchmark to evaluate either group leads to systematic over- or under-bidding. The fix is training prediction models on the studio's own historical data, or using genre-specific benchmarks.
Ignoring privacy-adjusted measurement
Apple's App Tracking Transparency (ATT), SKAdNetwork (SKAN), and Google's Privacy Sandbox have materially reduced user-level attribution quality. Studios still relying on last-click attribution at the user level are working with a degraded signal. The shift to cohort-level modelling, operating on aggregate revenue curves rather than individual device IDs, is structurally more compatible with the privacy-first era.
Under-investing in creative signal infrastructure
Creative quality is one of the highest-leverage variables in UA performance, and one of the most under-measured. Studios that do not track D7 LTV by creative at cohort level cannot distinguish between creatives that drive high-LTV users and those that drive high-install-volume but low-LTV users. The two are often inversely correlated in hypercasual and casual genres. The fix is tagging creative variants at the MMP level so cohort revenue can be attributed back to individual creative assets.
Delaying budget reallocation until trailing data matures
Waiting for D30 or D90 data to make budget allocation decisions means that a significant portion of a campaign's total spend has already been committed before any corrective action is possible. Studios with predicted D7 ROAS signals can reallocate within the first week of a campaign flight, capturing the full budget cycle rather than the tail of it.
The UA Optimization Tool Landscape
UA optimization draws from several categories of tooling, and most mature studios use a combination.
Mobile Measurement Partners (MMPs)
Ad Networks and DSPs
Creative Analytics Platforms
In-house BI and Data Warehouses
Predictive UA Analytics Platforms
This category, which includes Kohort's Ktrl, as well as tools like GameAnalytics (for smaller studios) and bespoke in-house ML stacks, sits on top of MMP data and adds the forward prediction layer. The key differentiators within this category are model accuracy at early prediction windows (D7 and D14), training data approach (industry averages vs studio-specific vs cross-customer aggregate), and the depth of campaign-level ROAS signal.
Specialist platforms in this category typically deliver forward ROAS predictions in weeks. Equivalent in-house builds usually take 6 to 18 months before producing trustworthy output, and run roughly $1M to $2M per year in fully-loaded data science headcount, depending on team size and seniority. Current US ML engineer total compensation sits at $160K to $260K per role; a production-grade cohort forecasting pipeline typically requires 2 to 3 dedicated engineers plus shared infrastructure. Specialist platforms also aggregate cross-customer training data, which improves early-window prediction accuracy in ways a model trained only on a single studio's history cannot match.
Most studios spending $5M+ per month on UA run a specialist platform for the foundational predictive LTV and ROAS forecasting layer, then augment with in-house tooling for studio-specific extensions (custom cohort segmentations, internal dashboards, integrations into their financial planning stack). The pattern mirrors how the same studios use MMPs like AppsFlyer or Singular rather than building attribution from scratch.
Studios Using UA Optimization: Results
Mobile game studios across hypercasual, casual, mid-core, and hybrid models have implemented systematic UA optimization with measurable outcomes.
52 Entertainment
Global evergreen games leader. 13% ROAS lift in 4 weeks and a 3.5x improvement in campaign efficiency by moving from trailing ROAS windows to forward LTV-based budget allocation.
Lotum (4 Pics 1 Word)
Makers of one of the longest-running casual word games. Scaled UA spend by 55% and increased revenue by 15% by combining long-tail LTV forecasting with disciplined budget reallocation.
Apps Teknoloji
Leading Turkish mobile gaming studio. Reduced campaign analysis time by 30% and saved 5+ hours per week per UA manager by replacing manual LTV prediction workflows with real-time predictive ROAS signals.
See all customer outcomes at /customers.
UA Optimization Glossary
The following terms appear throughout this page and across UA optimization discussions.
- LTV (Lifetime Value)
- The total revenue a user or cohort is expected to generate over a defined period (D30, D90, D180, D365). Usually expressed as ARPU (average revenue per user) at a given day horizon.
- ROAS (Return on Ad Spend)
- Revenue generated by a cohort divided by the ad spend used to acquire that cohort. A ROAS of 3x means every $1 spent returned $3 in cohort revenue.
- CPI (Cost Per Install)
- The average cost to acquire one install of a mobile game from a paid campaign. CPI alone is a poor performance metric: a low CPI from low-quality users can produce a worse ROAS than a high CPI from high-LTV users.
- eCPM (Effective Cost Per Mille)
- The effective revenue earned or cost paid per 1,000 ad impressions. Used primarily in IAA-monetised titles to measure ad revenue yield.
- IPM (Installs Per Mille)
- Installs generated per 1,000 ad impressions. A measure of creative and targeting efficiency at the top of the acquisition funnel.
- MMP (Mobile Measurement Partner)
- A third-party attribution platform (AppsFlyer, Singular, Adjust, Kochava) that tracks which ad campaigns drove which installs and in-app events. The data foundation for all UA optimization.
- SKAN (SKAdNetwork)
- Apple's privacy-preserving attribution framework, introduced with iOS 14.5. SKAN provides campaign-level conversion signals without user-level device data, using a conversion value schema that studios must configure.
- Blended ROAS
- Total revenue (paid + organic) divided by total paid ad spend. Blended ROAS overstates paid campaign efficiency when organic volume is significant, so always segment paid cohorts separately.
- Incremental ROAS
- The revenue generated specifically because of the ad spend, i.e., excluding users who would have installed organically anyway. Measuring incremental ROAS requires holdout experiments or geo-lift tests.
- Payback Period
- The number of days it takes for a cohort's cumulative revenue to equal the ad spend used to acquire it. A payback period of 90 days means a cohort breaks even at D90. Shorter payback periods allow faster budget recycling and compound growth.
Frequently Asked Questions: User Acquisition Optimization
How Kohort's Ktrl Platform Approaches UA Optimization
Kohort built Ktrl specifically for the UA optimization problem described on this page. Ktrl connects directly to your MMP (AppsFlyer, Singular, or Adjust), trains prediction models on your studio's own cohort data, and surfaces D90/D180/D365 ROAS forecasts at the campaign level within the first 7 days of each cohort. No data science team required.
Ktrl is trained on $6B+ of mobile gaming UA spend across studios from hypercasual to mid-core, and is the platform of choice for studios that want forward ROAS signals without building the prediction infrastructure themselves.
Further Reading and External Sources
- AppsFlyer: State of Gaming App Marketing Report: annual benchmarks on mobile gaming UA spend, ROAS, and attribution trends.
- Singular: ROI Index: network-level ROAS benchmarks and media source performance data.
- Adjust: Mobile Gaming Benchmarks Report: retention, ROAS, and session benchmarks by genre.
- Liftoff: Mobile Gaming Apps Report: IPM, CPI, and LTV benchmarks across ad formats and genres.
- Apple: SKAdNetwork Documentation: official documentation for iOS privacy-preserving attribution.
- Google: Privacy Sandbox for Android: official documentation for Android's Attribution Reporting API.
- Apple: App Tracking Transparency: official documentation on user consent for tracking.
Stop hoping your campaigns pay back.
Daily ROAS signals trained on $6bn of UA spend. Free to start, set up in minutes.