Bot installs and click spam make campaigns look cheaper than they are. Data analytics expert Harish Kumar walks through the maths, and explains why a fraud check must be in place before any budget is scaled.
One ad network is delivering installs at ₹40 each while the others cost ₹50 or more. The decision feels easy: move more budget to the cheap network before the festive sale. It is also, says Harish Kumar, exactly the moment to stop and check.
Kumar has spent more than eight years in marketing analytics, from consulting at ZS Associates and the startup years at PharmEasy to Disney+ Hotstar, Almosafer in Dubai and SWVL. As Data Analytics Lead at Interesting Times, he now sets up tracking and attribution for large-scale app launches. In every role, he has treated fraud as a data problem first and a finance problem second.
“When a cost per install looks remarkably good, that’s when you check it. Right numbers start with real users.”
The four fakes that look like growth
Bot installs: software, not people, installing your app
Click spamming: floods of fake clicks that steal credit for real installs
Click injection: a fake click slipped in just before a genuine install
Device farms: racks of real phones installing and opening apps at scale
The maths of a ‘cheap’ network
Say a brand spends ₹5,00,000 on the cheap network and receives 12,500 installs at ₹40 each. A fraud check flags 3,750 of them as invalid. The brand has paid 3,750 × ₹40 = ₹1,50,000 for users who do not exist, and its real cost per install is ₹5,00,000 ÷ 8,750 = about ₹57. An honest network charging ₹50 would have delivered 10,000 real installs for the same money.
Follow it to customers. If 2 in every 100 real users buy, the cheap network delivers 175 buyers at about ₹2,857 each; the honest one delivers 200 at ₹2,500. Double the budget on the cheap network and the money lost to fakes doubles too, to ₹3,00,000.
The second cost is harder to see. Ad platforms learn from the signals they receive, so fake installs teach the algorithm to find more of the same. Fraud also distorts every report built on top: retention looks weak because fake users never return, and channel comparisons tilt toward the sources that cheat most. India’s high-volume, thin-margin app market, with budgets that surge around festive sales and the IPL, makes that damage expensive.
Kumar’s fraud mechanism
The first layer is the protection built into the mobile measurement partner, such as Adjust or Branch, which should always be switched on. The second is an independent check: specialist tools such as mFilterIT analyse traffic across the funnel and integrate with whichever MMP a brand already uses, so suspicious installs can be kept out of both the data and the publisher bills. The third is commercial: agree with networks in advance that installs flagged as invalid will not be paid for.
A weekly red-flag checklist
Installs that happen seconds after a click
Unusual activity at odd hours
Sudden spikes from a single source
Many users who install but never do anything else
“I don’t believe in fancy dashboards. A beautiful chart built on fake installs is still wrong.”
The brands that scale safely, in his view, are not the ones that find the cheapest traffic. They are the ones that know how much of their traffic is real.
Through Explain Data, Kumar works with brands on exactly this: no fancy dashboards, just the right numbers and the right tracking in place.
Figures in the examples are illustrative, to explain the method.
Website: https://explaindata.in/
LinkedIn: https://www.linkedin.com/in/iamharishkumar/
Email: explaindataofficial@gmail.com