Your campaign dashboard is lying to you (a little)
Every mobile game UA manager knows the Monday ritual. You open the dashboard, sort by CPI, pause the expensive stuff, scale the cheap stuff, and feel productive for about eleven minutes. Then someone from the product team asks why D7 retention dropped, and suddenly the cheapest campaign of the week looks a lot less heroic.
This is the core problem with most UA campaign optimization in mobile games. Teams optimize what is easy to see (CPI, CTR, install volume) rather than what actually pays the bills (retention, monetization, LTV and ROAS). The ad platforms happily follow along, because they learn from whatever signal you give them. Ask for installs and you get installs. Whether those installs ever open the game twice is not really their concern.
In this article, we’ll walk through what UA campaign optimization should actually mean for a mobile game, why cheap installs are a misleading target, and a practical workflow for optimizing creatives, conversion signals and store pages around player value.
TL;DR
Good UA campaign optimization for mobile games is not about lowering CPI. It is about teaching ad platforms what a valuable player looks like and then feeding them creatives and store experiences that attract more of those players. In practice, that means three things: measure creative performance beyond the install, send conversion signals that correlate with long-term value without starving the algorithm of volume, and make sure your store page confirms the promise your ads make. Bid tweaks come last, not first.
What is UA campaign optimization for mobile games?
UA campaign optimization for mobile games is the process of improving how paid acquisition campaigns find, convert and retain valuable players. It covers creatives, audiences, conversion signals, store pages, bidding and budget allocation, and its success is measured by downstream outcomes such as retention, LTV and ROAS rather than by CPI alone.
That definition sounds obvious, yet most optimization work still stops at the install. The reason is mostly practical. Install data arrives fast and cleanly, while player value takes days or weeks to show up, and on iOS it arrives through the privacy-shaped keyhole of ATT and SKAdNetwork. So teams optimize the number they can see today and hope the number they care about follows.
A more useful way to think about a campaign is as a chain:
- Creative: the ad makes a promise about the game.
- Acquisition: the network finds someone who responds to that promise.
- Store conversion: the store page confirms or breaks the promise.
- Install and activation: the player starts, finishes the tutorial, hits the first meaningful moment.
- Retention: they come back on D1, D7, D30.
- Monetization: they buy something, watch ads, or both.
- LTV and ROAS: the only numbers your CFO actually remembers.
Optimizing one link in isolation is how you end up with a creative that has fantastic CTR, a great CPI and players who churn before the tutorial ends. Good optimization looks at the whole chain and asks one question: did this campaign acquire valuable players?
Why cheap installs are a trap
A cheap install is not necessarily a valuable install. This is probably the single most important sentence in mobile game UA. Imagine a mid-core strategy game running two creatives on Meta. Creative A is a fast, satisfying “merge the soldiers” hook with a CPI of $1.20. Creative B shows actual base-building and PvP gameplay with a CPI of $2.60. On a CPI dashboard, Creative A is the obvious winner, and it gets the budget.
Now follow those players into the game. Creative A’s players arrive expecting a hyper-casual merge game, find a strategy game instead, and leave. Its D7 retention is a fraction of Creative B’s, and almost nobody from that cohort makes a purchase. Creative B’s players cost twice as much, but they came for the game you actually built. Measured on D30 ROAS, Creative B wins comfortably. Measured on CPI, it gets paused. (The numbers here are illustrative, but if you’ve worked in UA, you know the drill.)
The issue is that CPI measures acquisition efficiency, while the business depends on player quality. Those two things are related, but they are not the same, and they often pull in opposite directions. Broad, clickbaity creatives tend to lower CPI by attracting a wider and less committed audience. Specific, honest creatives often raise CPI because they filter out people who wouldn’t enjoy the game anyway.
None of this means CPI is useless. It means CPI is an input to the equation, not the answer. A campaign is only well optimized when the cost of acquiring a player makes sense relative to what that player is worth.
Optimize the signal before you touch the bid
Much of modern UA campaign optimization is no longer done by you. Meta, TikTok and Google’s app campaigns do the heavy lifting on targeting and bidding. Your real job is to tell those algorithms what success looks like, and they learn exclusively from the signals you send them.
If the only meaningful event you pass back is an install, you are effectively telling the platform: “Find me people who install.” It will do that extremely well. If instead you pass back events connected to install, tutorial completion, progression, retention and revenue, the platform gets a much clearer picture of what a valuable player looks like, and it can go find more of them.
What makes a good conversion signal?
A good conversion signal is an event that happens often enough and early enough to give the ad platform learning data, while correlating strongly enough with long-term player value to separate good players from bad ones.
The catch is that these requirements fight each other. Further downstream does not automatically mean better.
A useful signal balances:
- Value correlation: does the event actually predict retention or revenue?
- Volume: does it happen often enough for the algorithm to learn from it?
- Timeliness: does it happen soon enough after install to be useful for optimization windows?
- Reliability: is it tracked consistently across devices, versions and platforms?
For example, a first purchase is a very strong value signal, but if only a small share of players ever buy, the algorithm may not get enough conversions to learn. On the other end, “opened the game” happens constantly and tells the platform almost nothing. The sweet spot is often something in between. If a puzzle game sees a strong relationship between completing level 10 in the first day and D7 retention, that event might be a far more useful optimization target than either a raw install or a first purchase.
How to find your best signal
In practice, start by pulling your cohort data and looking at which early behaviors best separate your D7 or D30 retainers and payers from everyone else. Typical candidates are tutorial completion, reaching a specific level, a session count threshold within the first 24 to 48 hours, or reaching a certain number of rewarded ad views. Then check how often each candidate fires. A beautiful predictor that fires 12 times a week will not train anything.
Finally, treat signals as hypotheses, not permanent settings. Test a new optimization event on a subset of campaigns, compare downstream cohorts rather than CPI, and iterate. The goal is not to send every possible event. The goal is to send the right one.
Creative optimization is about knowledge, not volume
Once the algorithm is chasing the right players, creatives become your biggest remaining lever. Targeting has largely been automated away, so the creative now does most of the targeting for you. A creative showing cozy decorating attracts different people than one showing competitive leaderboards, even inside the same broad audience.
The common response to this is “make more ads.” That helps up to a point, but a team pumping out 40 variations a week without a hypothesis is mostly generating noise with a nice thumbnail. Creative optimization works better as a learning loop:
- Form a hypothesis. For example: “Players who respond to a narrative hook retain better than players who respond to a satisfying-puzzle hook.”
- Isolate the variable. Change the hook, not the hook, the music, the end card and the character at once.
- Measure beyond the click. Compare cohorts on retention, revenue and ROAS, not only CTR and CPI.
- Extract the principle. What element made the difference? The theme, the gameplay moment, the fantasy, the first three seconds?
- Feed the next iteration. Build the next batch on what you learned, not on what happened to win last week.
The point is that a test should leave you knowing something new about your players, not just holding a winner that will fatigue in three weeks anyway.
The creative-level measurement gap
This loop has one practical problem. Standard UA reporting tends to tell you which network, campaign or ad set performed. It is much harder to see which individual creative acquired the players who went on to retain and pay, especially on iOS, where SKAdNetwork aggregates results and privacy thresholds limit granularity.
Without creative-level downstream data, teams fall back on CTR and CPI to judge creatives, which is exactly how the “merge soldiers” ad from earlier ends up with all the budget. Closing that gap, connecting creative to acquisition to player behavior to value, is what turns creative testing from a popularity contest into an actual optimization process.
Your store page is part of the campaign
UA teams often treat the store page as someone else’s problem, usually the ASO team’s. That’s a mistake, because every paid user passes through it. The ad creates the expectation and the store page confirms or breaks it.
If your ad sells a relaxing farm fantasy and your first screenshot shows a combat UI, a chunk of the people you just paid for will bounce before installing. That lowers your install rate, pushes up effective CPI, and sends the algorithm weaker signals, because fewer of the “right” people are converting.
A few practical checks:
Match the first screenshots to your top creatives. The first two or three screenshots carry most of the weight for people arriving from ads. They should show the same fantasy, gameplay and tone as the ads driving the most traffic.
Test the icon and screenshots with real audiences. Opinions in a meeting room are not data. Test variants and look at conversion behavior, not just preference.
Use custom product pages where it makes sense. If you run several distinct creative concepts, pairing each with a store page that continues the same story can reduce the promise gap.
Watch for positioning drift. If your best-performing ads have slowly moved away from what the store page says the game is, your store page is now advertising a different game.
The broader point is that ASO and UA are two halves of the same conversion funnel. Optimizing one without the other leaves money on the table in both.
When performance drops, don’t blame creative fatigue by default
Creative fatigue is the “have you tried turning it off and on again” of mobile UA. It’s sometimes right, but it’s also the answer people reach for before they’ve looked at anything. When a campaign’s performance declines, there are several possible culprits:
Creative fatigue: the same people have seen the ad too many times. Look at frequency and CTR decay on the creative itself.
Audience saturation: the creative still works, but the platform has run out of the easiest people to reach. CPI rises while CTR holds up.
Concept saturation: the whole market is running the same hook, so your ad no longer stands out even though it’s new.
Platform or signal changes: an algorithm update, a tracking change or a broken event can quietly change what the platform is optimizing for.
Market and seasonal changes: Q4 auction pressure, a big competitor launch or a holiday can move costs regardless of your creative.
Each explanation calls for a different fix. Refreshing a creative won’t help if the event feeding the algorithm broke in your last build. So before replacing the ad, form a hypothesis, check the data that would confirm or reject it, and only then act. It’s slower by about an afternoon and faster by several wasted weeks.
A practical UA campaign optimization workflow
If you want a sequence to follow, here’s the order we’d tackle things in. It moves from the foundations that affect every campaign to the tweaks that affect one.
- Define player value first. Agree internally on what a valuable player is for this game: D7 retention, D30 ROAS, payer rate, ad revenue per user, or a blend. Everything else gets judged against this.
- Find your early predictors. Use cohort data to identify which early in-game events best predict that value, and check that they fire often enough to train on.
- Fix the signal. Send those events to Meta, TikTok and Google as optimization or conversion events, and test them against your current setup on a subset of campaigns.
- Measure creatives downstream. Judge each creative on the quality of the players it brings in, not just its CTR and CPI.
- Close the store page gap. Make sure the store experience continues the story your winning creatives tell, and test assets with real audiences.
- Run hypothesis-driven creative tests. Iterate on the elements that drive player quality, not just clicks.
- Then tune bids and budgets. With the right signals and creatives in place, budget shifts finally start to mean something.
Where Geeklab and Audiencelab help
Two of the steps above tend to be the hardest in practice: seeing creative-level downstream performance on iOS, and testing store experiences quickly. That’s where our tools come in.
Audiencelab is a creative-level framework for attribution and analytics on iOS. It connects individual ad creatives on networks like Meta, TikTok and Google to what those players do after install, such as tutorial completion, progression, retention, IAP and ad revenue. It also lets marketers shape and send better conversion signals back to the networks. Its signal-engineering layer, Pulsar, lets you define rules like D7 retention, level completion or purchase thresholds in the browser, without a new development cycle for every change. Audiencelab complements your MMP rather than replacing it: the MMP keeps doing its job, and Audiencelab adds the creative-to-player-value view on top.
On the store side, Geeklab’s platform lets studios test icons, screenshots, messaging and even whole concepts on look-alike App Store, Google Play and Steam pages with real audiences, without waiting for store approval. That makes it much faster to check whether your store page actually confirms the promise your ads are making.
You don’t need either tool to follow the workflow above. They just make steps 3 to 5 considerably less painful.
Optimize for players, not for dashboards
UA campaign optimization for mobile games has changed. The platforms now handle most of the targeting and bidding, which means the work that matters has moved upstream: defining what a valuable player looks like, sending signals that describe that player, building creatives that attract them, and making sure the store page doesn’t scare them off.
So the next time a creative wins on CPI, ask what its players did on day seven before handing it the budget. Cheap installs make for a nice Monday dashboard. Valuable players make for a nice year.
Measure beyond the install. Optimize for player value. The dashboard will catch up eventually.
