A developer running a small puzzle app called Dayzle spent CA$220 (approximately $165 USD) on Google Ads over two weeks and discovered that the majority of reported installs were not from real users.
The developer initially set a target cost per install of CA$1.50, but Google couldn’t find installs at that price. After removing the cost target, the campaign spent double the daily budget and reported 21 installs in one day. However, checking the app’s admin panel revealed only 1 actual install—a significant discrepancy.
Upon investigating raw analytics, the developer found that 20 of the 21 reported installs were from devices running an outdated version of the app that the Play Store had stopped serving. Since the old version wasn’t available through official channels, these installations came from elsewhere despite claiming Google Play as the installer. Each of these devices opened the app once, spent zero seconds on any screen, and never returned. The phones represented 28 different models across 19 states.
Over the full two-week campaign, 56 installs were billed: 33 followed the suspicious pattern, 7 came from untargeted countries, and 13 appeared to be from real users. The 13 genuine users completed 92 games combined.
According to the developer’s analysis, the suspicious installs exhibited bot farm characteristics. The farm would watch the shortest available video ad without clicking it, then install the app from a saved copy rather than through the Play Store—a faster method that might avoid detection. This created a feedback loop: Google’s algorithm optimized for the install goal by sending more ads to the farm, which installed the app more frequently, making the campaign appear successful to Google while wasting the ad budget.
The developer filed an invalid-traffic form with Google and changed the campaign goal from “opened the app” to “won a puzzle,” raising the bar for what counts as a conversion. The reasoning: while it requires minimal effort to script an app opening and clicking, solving a Sudoku requires significantly more sophistication, making the app a less attractive target for bot farms.
The developer noted that larger apps likely encounter similar bot farm activity at higher volumes and may not realize the extent of the problem.
Key facts
- Of 56 billed installs, 33 exhibited bot farm behavior patterns
- 13 genuine users completed 92 games, showing real engagement
- Bot farms installed the app from saved copies to avoid Play Store detection
- The developer changed campaign optimization to “puzzle won” instead of “app open” to deter bots
- Bot farm activity created a feedback loop that encouraged Google to send more ad spend to fraudulent installs