Google Ads, a 60% Bot Claim, and the Trouble with Counting Installs

Key takeaways

• Evaluating the 60% bot claim involving Dayzle requires knowing how bots were identified and which installs were counted.
• An install followed by no activity does not establish that the user was fake.
• If 60% of installs were invalid, the cost per valid install would be 2.5 times the reported cost, assuming no refund.
• Campaign performance needs to connect spending with meaningful app use, retention, and purchases.

An app-install chart can climb while the business goes nowhere. A claim involving Dayzle—that 60% of installs from Google Ads were bots—raises a question anyone buying growth should care about: what did that ad spend actually buy?

Cheap installs can get expensive quickly

Start with a hypothetical campaign. Spend ₩1 million and get 1,000 installs. Your cost per install is ₩1,000.

Now suppose 600 of those installs are invalid, leaving 400 valid ones. Without a refund, each valid install effectively costs ₩2,500. That is 2.5 times the number on the dashboard.

These are illustrative figures, not Dayzle’s reported spending or losses. They show how much the economics change when the install count includes activity that should not count toward the campaign’s results.

Even those 400 valid installs are only a starting point. Some people may never return. Others may use the app without paying. If the business needs subscribers, the relevant calculation eventually becomes spending divided by subscribers acquired.

An install is a useful milestone. It is an early one.

Before accepting 60%, check what was counted

The Dayzle-related figure should be treated as an allegation. Its meaning depends on both the denominator and the method used to identify bots.

Did the calculation cover every install attributed to advertising? Only installs during a particular period? A subset already selected because it looked suspicious?

Those are different populations. A high rate within suspicious traffic cannot establish the rate across an entire campaign. A finding from one campaign cannot establish the rate across Google Ads.

The classification method matters just as much. Labeling every inactive install a bot is a much weaker basis than identifying repeated behavior that would be difficult to explain as human activity.

A developer’s observations can reveal a problem worth investigating. Establishing its cause takes more than an alarming percentage.

People abandon apps for perfectly human reasons

Someone sees an ad, installs the app, and leaves. Perhaps the promised feature is missing. Perhaps registration asks too much. Perhaps curiosity lasted exactly as long as the download.

None of that is good news for the advertiser. None of it, by itself, proves fraud.

The reverse also holds: an install followed by an app launch does not guarantee a real person was involved. Software can imitate those actions.

The useful distinction is between low-quality acquisition and fabricated activity. The first may point to poor targeting, a misleading ad, or a frustrating first-use experience. The second may provide grounds to suspect fraud. They require different responses.

“Invalid traffic” and “bot installs” also need care. An ad impression, an ad click, an installation, and a first launch record different events. Evidence of something abnormal at one stage needs to be connected to the others before it can explain the whole journey.

Follow what happens after the download

For an app operator facing a similar pattern, the first step is to make the suspicion specific.

Do installs arrive in unusual bursts? Is the time between installation and first launch suspiciously uniform? Does the same sequence of actions repeat inside the app?

Each pattern can justify a closer look. None is conclusive alone. Measurement errors and legitimate usage can also produce strange-looking records. Several signals appearing together within the same group of installs make a stronger case.

The same discipline improves ordinary campaign analysis. Connect installs to the first meaningful use of the product, then to return visits and purchases. Decide which behavior the campaign is supposed to produce before declaring the install count a success.

A complaint to the platform also needs that specificity: the campaign, the dates, the observed patterns, and the basis for separating suspicious activity from ordinary users.

The Dayzle allegation cannot be settled by repeating 60%. But it does put a useful question in front of every advertiser: does the biggest number on the dashboard tell you whether your customer base is growing?