If you run paid acquisition for a mobile game, you will meet two campaign goals sooner or later: ROAS and retention. They sound similar, because both aim at players who stay and pay, but they steer the ad network in different ways. Choosing the wrong one for your game wastes budget, and choosing well is one of the easier ways to improve ROAS for mobile games.

This post explains the difference between ROAS campaigns and retention campaigns, how to set up user acquisition for a game that monetises through both ads and in-app purchases, and what actually moves ROAS up.

What is the difference between ROAS campaigns and retention campaigns?

A ROAS campaign tells the network to find players who are likely to return revenue. You give it a target, usually a return at a set day such as day 7, and the network’s model bids more for installs it predicts will earn enough. The goal is money back, measured against what you spend.

A retention campaign tells the network to find players who are likely to still be playing after a set period. Unity Ads, for example, offers retention campaigns that target users expected to be active seven days after install, and you set a target cost for each such returning user. The goal is engagement, and revenue is not what the model is chasing.

The practical differences come down to four things:

  • What the model predicts. Revenue for ROAS, a returning player for retention.
  • Who it finds. The two goals tend to reach different groups of people. Unity’s own guidance suggests running them side by side for that reason.
  • How you judge them. Retention campaigns should be read on retention and install cost, not on ROAS alone. Judging them on ROAS will often make them look worse than they are.
  • What data you share. A retention goal needs a return-visit signal from your attribution partner, but it does not need purchase values, which matters if you would rather keep revenue data private.

Neither is better in general. They answer different questions, and the right mix depends on how your game earns.

When retention campaigns make sense

Retention goals fit best in a few situations:

  • Your game earns mostly from ads, where every returning day produces income, and purchases are rare or slow to appear.
  • You are early in a launch and have too little purchase data for a revenue model to learn from.
  • You want to widen your reach, because your ROAS campaigns have started to find the same kind of player again and again.
  • You care about long-term engagement for reasons beyond one quarter, such as live events or community.

Retention campaigns are weaker when your cash position demands fast payback. A player who returns on day 7 has not necessarily paid back what you spent to get them, and a retention goal does not check.

When ROAS campaigns make sense

ROAS goals fit when you have enough revenue data for the network to learn from and a payback target you can state in numbers. They are the natural choice for games with a solid purchase flow, and for studios that need to prove each campaign pays for itself.

They need more data to work. If your volume of paying players is low, the model has little to learn from, and results can swing wildly week to week. In that case a different goal, or a longer test, is usually the safer path.

How do we set up UA for a game that earns from ads and in-app purchases?

This is the question we hear most from hybrid games. Most of ours earn from both, and for our titles the mix is about 80 percent ads and 20 percent in-app purchases. Here is the setup we recommend.

1. Measure both revenue streams on every install. Your ROAS target should include ad revenue and purchase revenue together, because that is what the player is worth. A campaign judged only on purchases will undervalue players whose ad income is high, and the reverse is true for a campaign judged only on ad income. We explain how the two combine in our post on LTV for mobile games.

2. Send total revenue back to the network where you can. Where an ad network supports optimising on combined revenue, give it that signal. The more the model sees of what a player is truly worth, the better it targets. If a network only accepts purchase events, use it for what it does well and track the rest yourself.

3. Run ROAS as the main engine and retention as a supporting test. Put most of the budget behind the goal that pays back, and use a smaller retention campaign to explore audiences your ROAS campaigns are missing. Compare them at the same day, and let them run long enough for cohorts to mature.

4. Keep the campaigns separate. Do not mix goals inside one campaign, or you will not know which one drove a result. Give each its own budget, its own targets and its own reporting line.

5. Review blended results, not just campaign results. A retention campaign can improve the quality of players that reach your ROAS campaigns’ audiences indirectly, or it can simply cost money. The blended view over a quarter tells you which is true. Our guide to what a good ROAS looks like covers why we favour quarterly averages.

How to improve ROAS for mobile games

When ROAS is below where you want it, most of the gain comes from a short list of changes. In rough order of impact:

  • Fix monetisation before you buy more traffic. Ad placements, offer design and pacing change what every player is worth. Higher player value raises ROAS on every campaign at once. Our post on monetisation design is a good place to start.
  • Choose the right optimisation event. If the network optimises toward an event that does not reflect real value, it will find the wrong players. Pick the event that best predicts revenue, and revisit it when your game changes.
  • Cut weak sources and countries. Rank cohorts by value per install at the same day, and stop funding the ones that never catch up.
  • Refresh creatives. Tired ads push up the cost of each install and lower quality. We cover this in our post on creative fatigue.
  • Scale in steps. Large jumps in budget reset a network’s learning. Raise spend gradually and watch ROAS at each step. Our guide to scaling Applovin campaigns shows how we do it.
  • Be patient with the data. Judging a campaign after two days usually means judging noise.

A note on measurement windows

ROAS is always measured at a point in time, and ROAS campaigns usually optimise toward an early day such as day 7. Games whose players pay slowly can look poor at that point and strong at day 60. If that describes your game, use a longer window for judging and an earlier one only for steering. Mixing the two up is one of the most common reasons good campaigns get switched off too soon.

Putting it together

For a game that earns from ads and purchases, a good default is a ROAS campaign built on combined revenue as the main engine, a small retention campaign running alongside to find new audiences, and a regular review of blended results. Across 40 months of aggregated results, our blended ROAS has stayed above break-even, and that has come from this kind of steady, measured mixing of goals across Applovin, Google Ads, Mintegral and Unity, not from any single setting.