How to verify your AI monetisation is actually working

Here is the thing nobody in monetization tooling wants to say out loud: most of them have no idea whether they're actually helping you.

They can show you conversion rate charts. They can show you revenue trending up. What they can't show you is whether any of it is happening because of them, or whether your game just had a good month.

We wanted to be different. So we built the measurement differently.

The problem with most attribution

Imagine you turn on a new monetization tool in January. By March, revenue is up 18%. The tool's dashboard says "18% uplift". Your CEO is delighted.

But what if your game was featured by a major gaming publication in February? What if your new level pack launched and brought back players who had churned? What if the whole mobile games market had a strong quarter?

The tool has no way to separate its contribution from everything else that happened. You're comparing revenue before and after adding the platform — on the same player base, over a period when many other things also changed. That's not a measurement. It's a correlation that you happen to like.

A studio that can't separate those two things has handed a third-party service a permanent revenue share on their game's organic growth. That is the sleight of hand at the centre of most performance-based pricing in this industry.

How a control group changes this

When you connect a game to gamemantra.ai, we permanently hold back 10% of your players. These players never see an AI offer. They play your game exactly as they would if we weren't there at all.

The 90% of players who do receive AI offers go through the platform's recommendations each session. The 10% who don't are the baseline.

Every month, we compare actual in-game purchase revenue between the two groups. Same time period. Same seasonal conditions. Same content updates. The only difference is whether the AI served that player.

If the AI-served players out-spend the baseline group, that difference is the uplift. If they don't, the answer is zero. This structure is not a gesture of goodwill. It's the only honest way to know whether the AI is contributing anything at all, and without it, neither we nor you could answer that question.

Why this protects you

The control group does three specific things for your studio.

It isolates what the AI did from everything else. Good months happen. Bad months happen. Viral moments happen. The control group is your baseline — whatever the market does, it does to both groups equally. What remains when you compare them is the AI's contribution, stripped of seasonal noise and external factors.

It makes the revenue share honest. We earn a percentage of growth we can demonstrate with evidence, not growth we can claim with a before/after chart. A studio that had a strong organic quarter and a flat AI-served month pays us nothing for that quarter. That is the right outcome, and it only works because the control group exists.

It gives you the ability to verify. At any time, you can look at the same numbers we see. The raw event data that drives the comparison is accessible to you directly. If you think our methodology is wrong — the time window, how we classify a purchase event, the statistical threshold we require — you can tell us. We'd rather have that conversation than have you quietly stop trusting the dashboard.

What we measure

The comparison runs on actual in-game purchase revenue — real money changing hands, not impressions, not clicks, not predicted lifetime value.

We use industry-standard statistical testing to determine whether the difference between the two groups is real or just random variation across samples. A result has to be conclusive at a meaningful sample size before we report it as uplift. We don't select the best two-week window and show you that as the month's result. We show you the full measurement period.

The test also accounts for how many players are in each group and how long the measurement has been running. A smaller game will take longer to reach a statistically meaningful result than a game with hundreds of thousands of monthly active players. We surface this as part of the dashboard — you will always know whether a result has cleared the threshold or whether it needs more time to resolve.

How to check the numbers yourself

The comparison lives in your analytics dashboard under Measurement. You can see daily revenue per player for both groups, the overall uplift figure, and the statistical confidence level behind it.

If you want to export the raw events and run the calculation yourself, you can do that. If you disagree with how we are measuring something — the time window, the categorisation of certain IAP event types, the confidence threshold we require before reporting a result — talk to us. The conversations studios have had about methodology have consistently improved how we run the measurement.

The goal is a number both sides trust independently. A revenue share based on a number that only one party can verify is just an invoice dressed up as a performance model.

---

See how it works → or book a 20-minute demo to walk through the measurement with your game's numbers in mind.