Google Ads app campaigns are one of the most reliable ways to buy Google Play user acquisition at scale, and also one of the easiest to set up badly. The campaign type gives you very few manual controls. You hand Google a budget, a creative set, a bidding goal and a conversion event, and the system does the rest. That means the quality of your setup decides almost everything.

This post covers how we set up Google Ads app campaigns for mobile games, how we ramp bidding from installs to in-app actions, and when it makes sense to move to target ROAS (tROAS). Google is the second network in our mix after Applovin, and the same discipline applies to both: pick the signal carefully, then give the algorithm room to learn.

How Google app campaigns work for games

An app campaign is not a classic search or display campaign. You supply text, images, video and an app listing, and Google assembles ads across its own inventory. You cannot choose placements or write individual ad groups the way you would elsewhere.

Because you cannot steer placement, your levers are the ones the system listens to:

  • The conversion event you tell it to optimise for.
  • The bidding goal and target attached to that event.
  • The budget, which sets how much room the system has to explore.
  • The creative set, which decides which formats and messages get tested.
  • Location and language, which keep each campaign’s data clean.

Setup: get the signal right first

Before any money is spent, check the data path. The system can only optimise for what it can see.

For a game on Google Play, conversion events normally come in through Firebase or through Google Play codeless conversions. Google’s own guidance for value-based bidding requires events that pass a revenue value, so a tutorial-complete event with no value attached will not support tROAS. Confirm that the events you plan to use are marked as primary goals and that they fire correctly in a test build.

Then decide how to structure campaigns. Our default is one campaign per geo group and one per optimisation event. Mixing a high-income market and a low-income market in a single campaign makes the learning noisy, because the algorithm sees very different players converting at very different rates. Clean campaigns also make it easier to read results later.

Finally, load a real creative set at launch, with videos in several aspect ratios, images and varied text. Our post on creative fatigue explains why you should also plan refreshes from day one, not after performance drops.

Which event should you optimise for?

This is the most important decision in the account, and it connects to a simple question: which in-game action best predicts the player becoming valuable?

There are three broad choices.

Install. The campaign optimises for the cheapest installs it can find. It is useful for a brand new app with no event history, or when you only need volume to seed data. Cheap installs are often low quality, so keep it a short phase.

An in-app action. The campaign looks for players who install and then complete a chosen event. The event needs to be early enough to happen often, and meaningful enough to separate engaged players from tourists. Reaching a certain level, finishing the first few sessions or hitting a first ad impression are typical candidates. If the event is too rare, the system starves for data. If it is too easy, it tells you nothing.

In-app action value. The campaign optimises for the revenue value attached to your events, and you set a target ROAS. This is the goal for games that already have a stable picture of what a player is worth.

For hybrid games, where revenue comes from both ads and purchases, the value you send should reflect both. We cover that setup in more detail in ROAS vs retention campaigns for hybrid games.

The bid ramp: from installs to tROAS

We treat bidding as a ladder, not a switch. Each step earns its place by producing enough data for the next one.

Step 1: install or in-app action bidding

Start with an install goal or a tCPA goal on a simple in-app action. Google’s documentation notes that a very small budget can prevent a maximise-conversions campaign from getting conversions at all, because the bid needed may be higher than what the budget allows. So set a budget that can realistically afford several conversions a day at your target cost.

Google also suggests starting with maximise conversions or tCPA until you have around 30 conversions in 30 days. We treat that as a floor, not a goal. More volume gives a steadier signal.

Step 2: build value history

While the campaign runs, make sure revenue events are reaching the account with correct values. You are collecting evidence on what a player in each geo is worth after a set window, and your own attribution data should confirm it.

Step 3: move to tROAS

Google’s guidance for moving from tCPA to tROAS is to wait for at least three to four weeks of conversion value history. We agree with that and are usually stricter for games with long payback, because early revenue can be a poor guide to later revenue.

For the starting target, Google recommends using the return that a comparable campaign has already achieved. A comparable campaign is the same app, the same optimised event, and the same location and language targeting, measured over the same conversion window. If your tCPA was set well above the actual cost per action, the starting tROAS needs a matching correction, otherwise it will be set too loose.

How to set the tROAS target

A tROAS target is a promise about average return over the conversion window, not about every day. Google states that all app bidding strategies are target strategies that aim for the selected performance across the window, so individual days will land above or below it.

Two ideas help here.

The target controls scale. A higher target narrows the pool of installs the system will buy, so volume falls. A lower target widens it, so volume rises and the average return per unit of spend usually softens. There is no free setting. You are choosing a point on that trade-off.

Your target should come from your break-even line. Work out the return you need on your own payback window, then set the target with some margin above it. Our post on what a good ROAS looks like shows how we calculate that line. Across 40 months of aggregated results, our blended ROAS has stayed above break-even, and the reason is that we set targets from break-even maths, not from hope.

How to adjust once it is live

The most common mistake is changing too much, too often. Each change can push the campaign back into learning.

  • Wait until the campaign has a stable baseline of conversions before touching the target.
  • Make big changes no more than once every one to two weeks.
  • Keep each adjustment modest, in the range of 15 to 20 percent of the historical average.
  • Judge results over 14 to 30 days and ignore the most recent days, because conversion reporting lags.
  • If spend runs higher than you want, raise the target so the campaign becomes more selective.

When to widen, split or pause

Once tROAS is stable, scaling is mostly about adding clean campaigns, not squeezing one. Open a new geo group in its own campaign. Add a second event-based campaign if you need to test a different goal. Refresh creative on a regular schedule so the system always has fresh material to test.

Watch for signs of saturation: rising cost for the same volume, falling return at the same target, or creative that stops winning. Our guide to channel saturation and test design covers how to tell a bad week from a real ceiling.

Where Google fits in a wider mix

Google Ads is strong for reaching Android players at scale, but no single network should carry the whole plan. In our mix, Applovin is the lead network, followed by Google Ads, Mintegral and Unity, and each earns its share by clearing the same return bar. If you are choosing between channels, our guide to the best ad networks for mobile game UA lays out how we compare them.

The short version

Get the events and values right before spending. Start on installs or a simple in-app action, build at least three to four weeks of value history, then move to tROAS with a starting target taken from a comparable campaign. Adjust slowly, judge results over a window, and keep your creative fresh. Do those things and Google app campaigns become a steady source of players, not a black box.