I spent $220 on Google app ads and 60% of the installs were robots
A small app developer discovers that over half of his Google ad installs are bots, exposing the pervasive fraud within the ad ecosystem. This story resonates with HN's skepticism of large ad platforms, highlighting how their incentives might be misaligned with advertisers' success. The author's forensic analysis and strategic pivot offer a pragmatic, albeit cynical, path forward for others battling ad fraud.
The Lowdown
The author, creator of the puzzle app Dayzle, embarked on a Google Ads campaign for Android with a daily budget of CA$40. Initially, the campaign barely spent, but after removing the target cost per install, it quickly blew past the daily budget, reporting 21 installs. A deeper dive into his analytics, however, revealed a stark discrepancy between Google's reported numbers and actual human engagement.
- Initial Discrepancy: While Google reported 21 installs, the author's admin panel showed only one new install. Further investigation showed 20 of these 'installs' were using old app versions no longer served by the Play Store, indicating they were sideloaded.
- Bot Behavior: These non-human installs originated from 28 phone models across 19 states, yet all exhibited identical behavior: one open, zero seconds spent in-app, and never returning. The author determined these were bot farm activities.
- Fraudulent Loop: The bots would watch the shortest ad video without clicking, then install the app from a saved file (faster, less detectable). Google's algorithm, seeing a 'view followed by install' as a conversion, then directed more ad spend to these bot-generating placements, creating a self-perpetuating waste of advertising budget.
- Mitigation Strategy: To combat this, the author changed the campaign goal from 'installs' to 'won a puzzle,' making it significantly harder and more expensive for bots to fake engagement, thereby pricing out the bot farms.
- Warning to Advertisers: The author concludes with a public service announcement, urging other advertisers to scrutinize Google's reported install counts and delve into their own analytics, as ad fraud can affect even small budgets.
The author is currently awaiting a response from Google regarding invalid traffic and will report back on any refund, emphasizing that such detailed investigation is crucial when relying on ad platform metrics.
The Gossip
Ad Fraud Frustration and Skepticism
Many commenters shared the author's frustration, recounting their own experiences with ad fraud on Google and Meta platforms. A prevailing sentiment was that these large ad networks have a vested interest in not fully eradicating fraud, as it inflates their reported metrics and revenue. Users pointed out the increasing sophistication of bots and the lack of reliable reporting or appeal processes from Google, suggesting a fundamental misalignment of incentives. Some outright called Google and Meta ads a 'con,' while others noted that big corporations have 'dumb money' to spend, creating an inefficient system.
The Ad-Tech Ecosystem: Who Benefits?
The discussion delved into the mechanics of how ad fraud profits various actors. Commenters explained that bot farms get paid by Google for providing 'ad space' and faking conversions, which in turn leads Google to funnel more advertiser money into these fraudulent placements. The system creates an incentive loop where both the bot farm and Google (indirectly) benefit from the advertiser's wasted spend. The concept of 'residential proxies' was brought up as a method bot farms use to evade detection, often by compromising user devices for 'free' services.
Strategic Safeguards and Alternatives
Amidst the skepticism, some users offered practical advice and alternative strategies for managing digital advertising. Suggestions included manually excluding IP ranges from bot networks, implementing extremely stringent Google Ads campaign settings (e.g., exact match keywords only, disabling AI Max and automations), and setting specific in-app conversion goals beyond simple installs. The discussion also touched upon content marketing (like the author's blog post, which effectively drove app installs) and the general Y Combinator advice to avoid online ads, especially for early-stage companies, due to their inherent inefficiencies.