Marketing a new mobile game: Where should studios start?

• Facebook can provide a starting point for testing a new game's market fit using gameplay creatives.
• For tycoon and idle games, AppQuantum's 2025 benchmarks put day 1 retention at 40% or higher and day 30 retention at around 6%.
• Early ROAS tests can help studios build initial forecasts for return on ad spend and lifetime value, with around 50 paying users providing a starting point for estimates.

Ilia Maltsev is CMO at AppQuantum.

Every publisher has their own view of marketing. Some look only at payback metrics and ROAS. Others evaluate a broader set of metrics, test both marketing and product performance, and also look at marketability and other signals.

Over time, we’ve developed our own workflow and internal benchmarks based on a large number of launched projects. Some parts of this process work specifically because of our experience. Others are constantly being improved. But in this article, we can still break down the basics of working with a new project - including principles that may be useful for smaller studios and independent teams.

We’ll start with the general logic, then move into more specific metrics, traffic sources, and approaches in the context of idle games, which are closest to our expertise.

In short, when a new project comes to us for testing, it usually doesn’t have much depth yet. So at first, we look at the acquisition funnel, early retention, and market fit. If those signals are strong, we add monetisation, evaluate performance, and start building out content depth.

But first, let’s choose the right traffic sources for testing and walk through the process step by step to better understand how a publisher typically approaches a new project - and how you can apply the same logic independently, without large volumes of historical data or the experience of a major company.

Traffic sources and creatives

If we’re talking about the first traffic source to consider, I would look at Facebook. It’s a reactive channel with a broad audience, relatively fast and affordable learning, and fewer early-stage mistakes.

Facebook tends to produce fewer false positive signals than some ad networks, where you first need to spend a lot of time training the network before it finds the right audience. With Facebook, you can usually start testing right away.

Next, we launch a test in a large market, such as the United States. For marketing creatives at this stage, we use gameplay creatives specifically, because the goal is to test the market fit of the core mechanic and setting. We’re looking for the right balance between audience size and CPI.

Then we start analysing the first metrics.

Metrics and benchmarks

First, we look at CPI, CTR, IR, and IPM.

Second, we look at retention. Retention benchmarks differ from publisher to publisher. Some teams can only work with a product after it reaches a certain threshold. Others rely on experience with many projects in a specific genre, and so on.

Market averages can be found through analytics platforms, though they’re expensive and often inaccessible to smaller publishers, let alone independent teams. So I’ll include a few numbers here. They can also be found in public materials from some analysts.

Based on averaged data, in 2025, for tycoons and idle games, Day 1 retention should generally be at least 40%, assuming homogeneous traffic from one region, such as the United States. D3 should be around 20–25%, D7 at least 8–12%, though ideally higher, and D30 around 6%.

That said, a new product usually doesn’t have enough depth yet, so we typically look at R1, R3, and R7 first. For CPI, we usually aim to stay below $1.50.

In marketing tests, it’s important to remember that chasing “perfect” statistical significance is not always rational. To validate a hypothesis at a classic confidence level, you may need thousands of installs, which also means thousands of dollars in ad spend.

At the early stage, we treat metrics more as orders of magnitude. We evaluate R1, CPI, and other indicators together with the quality of the team, iteration speed, and our subjective view of the product’s potential - a view that comes from deep analysis and time spent inside the game.

This approach allows us to make decisions faster and cheaper, without waiting for an academically “perfect” data volume. For a primary R1 and CPI test, we usually use 300–500 installs. In most cases, that volume is enough to understand the direction and decide whether it makes sense to scale the test further.

Depth and the first real test

Let’s say we get workable retention for the first seven days. At that point, the first priority is to confirm that the game has enough depth, add content, and build level progression so the product can at least pass the R30 gate.

From there, we work on R30, add monetisation, and bring in users at a larger scale. This can be a real test with a budget of $15,000 or more, but now the goal is different. We are no longer testing installs like before - we are testing payback. At this stage, we look at ROAS while continuing to tune level progression and push retention toward what we consider a healthy benchmark.

If the first tests were run on Facebook, we usually return to the same source and relaunch with more targeted optimisation goals. This allows us to validate ROAS for specific days and build more reliable UA payback predictions.

From there, since we use AppLovin MAX as our mediation layer, the next step is typically to launch on AppLovin. After that, we evaluate whether it makes sense to expand further - for example, to Unity or Google, keeping in mind that Google usually requires a fairly large budget to exit the learning phase.

Acceptable CPI and CPA thresholds are defined by internal benchmarks. They depend on the type of optimisation, but at this stage, product metrics, actual ROAS, and ROAS dynamics become more important for us.

During a ROAS test, one question inevitably comes up: should we stop buying traffic once we have enough data to validate the next update, or does it make sense to continue testing?

There is no universal answer. The decision depends on the results of the previous stage: how convincingly the key metrics improved, whether there are signs of positive dynamics, and how likely further growth seems. Available budget also matters. If the product shows strong signals, additional investment in the test may be justified. If resources are limited, it may be more reasonable to end the test and focus on the next iterations. The main task is to balance getting enough data with using the marketing budget efficiently.

In our case, at this stage we stop the test only if there is a serious problem. First, a product problem: if something broke inside the product itself, that is the main red flag. Second, if the actual ROAS curve turns out to be significantly lower than the progression we based the test on.

In practice, we haven’t had any cases like that, because before reaching this stage, we already run install-optimised tests to validate retention for the target day. Adding monetisation on top of strong engagement, along with proper payment traffic optimisation, rarely breaks performance. That said, metrics still need to be monitored closely so problems can be spotted early.

This is why we recommend testing products step by step, so you don’t end up spending money for nothing at the next stage.

What comes next

The first ROAS tests give you not only user acquisition cost data, but also your first paying players, along with their product and monetisation metrics. Based on this data, you can already build early, rough predictive models for ROAS and LTV over the next several months. The accuracy of these forecasts increases significantly if the team has historical data from similar projects. Their retention, monetisation, and LTV curves can serve as a strong basis for calibrating early models.

In practice, we usually start building the first forecast curves after reaching around 50 paying users. In most cases, that volume is enough to get the first meaningful estimates and make a decision about further investment in the product.

At the same time, during further user acquisition, the marketing team does not rely only on predictions. We continue to monitor actual product and marketing metrics, as well as ROAS dynamics.

Throughout this entire process, the project itself is also being developed. Ideally, by this point, we already have content depth for Day 60. This is important: even if you run a D30 ROAS test, the game should have more content than that - closer to 60 days, not even counting the fact that paying players may consume content much faster.

This is why development depth should start after successful CPI and R1 tests. In idle games, content production takes a lot of time and resources. We usually try to build depth up to D90–D180, because a full-scale launch is not really possible before content production has been scaled up.

But content production often takes three to six months. If the content is not ready yet, traffic can be maintained for some time at around $30,000–50,000 per month. If we see results close to KPI, we do not stop acquisition. There have been cases where we launched a product, ran a test, and received poor results. In those cases, we stopped the test and continued refining the product until we reached payback, while also improving retention and other metrics.

In parallel, we start actively iterating on creatives and preparing new ones. We create several dozen gameplay creatives with stronger visuals and mechanics to maximise conversion at launch.

In the end, if everything goes according to plan, we have product depth, we have completed a real performance test, and we have reached target payback. From there, we expand user acquisition. The product team’s job is to keep increasing depth and improving monetisation at every stage. UA’s job is to bring in more traffic as a multiplier.

But keep in mind that companies calculate money differently. Some focus on volume, some on margin. Some are willing to increase turnover and reach a target ROAS 30 payback with the help of organic traffic. Others may lose a little on organic, but get more predictable profit from paid traffic.

So always adapt this logic to your own experience, resources, and situation.

Scaling up

After that comes the magic of operations, global launch, at least $100,000 for the next month, and off you go.

Of course, there are many more nuances at that stage. But among developers who have not been deeply involved in marketing, there is a common belief that all you need is some money for user acquisition and one UA manager to launch campaigns - and then the project will fly to the top. Unfortunately, that is not how it works. The exceptions are usually survivorship bias.

We hope this article helps developers and smaller studios better navigate the basics of primary marketing.

And one final note: benchmarks, hypotheses, traffic sources, and approaches - everything we discussed here - are constantly changing, updating, and eventually becoming outdated. So you need to keep watching the market. Ideally, you build your own expertise. Alternatively, you can use someone else’s experience. The second option is simply easier and more cost-effective, although it is not a universal rule.