The fraud problem nobody talks about (until it's too late)

Why indie studios stay blind to fraud

Here's the thing about being a small team: you don't have a dedicated fraud analyst. You don't have access to the machine-learning models that Adjust or AppsFlyer use to flag suspicious patterns across millions of apps. You get an install number, a cost, and a hope that your ad partner isn't gaming the metrics.

That studio I mentioned? They'd been paying for real user acquisition, but roughly 30% of their "installs" were never opening the app. The ad network's own dashboard said everything looked normal. Their retention was tanking, but they didn't have the attribution layer to see which ad network was the culprit.

We started getting questions like this every few weeks. Not from enterprise shops with fraud teams. From developers running on tight budgets, trying to grow lean, who needed to know if their ad spend was actually landing real people.

What changed when we looked at the data ourselves

We pulled together patterns from thousands of installs across Attribr's network. Bot behaviour isn't random. It's predictable. Fake installs cluster in specific time windows. They come from suspicious device fingerprints. They often skip the retention check entirely (they're not designed to keep the app open). They don't trigger user actions.

We realised we were already sitting on the data. Attribr captures install source, timing, device signals, and whether a user comes back at day 7, 14, and 30. Fraud patterns emerge from that picture immediately. A studio doesn't need enterprise tooling to spot them. They need their attribution SDK to flag anomalies in real time.

That's when we decided to build fraud signals into the Pro tier. Not as a standalone service (we're not a fraud-detection company). But as a native part of the attribution dashboard, because the two are inseparable. You can't understand where your installs came from without knowing which ones are real.

Then we noticed something about ad networks

The studio with the bot problem was running across five different networks. Their dashboard showed five separate reports, five different metrics, five different versions of the truth. Aggregating them meant copy-pasting numbers into a spreadsheet and hoping they lined up.

We built ad-network roll-up into the same feature. One dashboard. All five networks. Single view of which one was underperforming. Which one was delivering real users versus which one was just inflating numbers.

This matters because it removes friction from a decision that indie studios should be able to make quickly. If Network A is delivering 40% real users and Network B is delivering 85%, you should see that in under a minute. Not in a spreadsheet. Not in a support ticket to your ad partner.

The fraud signals tell you something is wrong. The ad-network roll-up shows you which network it is. Together, they answer a question that used to require either paying for enterprise software or just accepting the loss.

Why this isn't a replacement for everything

I'll be direct: Attribr's fraud detection is built for indie and studio scale (up to 100K monthly installs per app). We're looking for obvious patterns: bot clusters, impossible retention signatures, timing anomalies that indicate automated fake installs. We're not running the kind of cross-app machine-learning models that catch sophisticated fraud rings or adversarial techniques that change every week.

If you're running millions of installs monthly, you probably need a dedicated fraud team or a platform designed purely for fraud detection. But if you're at 5K, 25K, or 50K installs a month and you've been flying blind, this changes things. It means you can actually see which ad networks are trustworthy and which ones are costing you money.

The feature lives in the Pro tier (£99 a month, up to 100K installs). Fraud signals + ad-network roll-up together. We bundled them because they solve the same problem: visibility into which installs matter and where they're coming from.

What that studio learned

They switched off the bad ad network immediately. Their CPI stabilised. They started seeing real retention numbers because they could finally separate signal from noise. They told us it felt like putting on glasses.

That's the story that stuck with us. Not a feature announcement. A developer getting clarity on something that was costing them money. And realising that they didn't need enterprise software to get it. They just needed their attribution SDK to do what it was supposed to do: tell them the truth about their installs.