What ad click fraud is
Contents
Share
Click fraud, which Russian advertisers call скликивание, is a click you pay for even though the person who clicked was not going to buy. If Yandex Direct already marked the click invalid and did not charge you, it is not part of the loss. It never reached the spend line.
The rest of this article is the boundary between those two sets, and a check you can run in your own account in one evening. How the word click fraud differs from скликивание is a separate piece. Here the only question is which click was removed and which click you paid for.
A charged click that was not a buyer
You pay for a click the ad system counts. Once that click sits in spend and there was no intention to buy, it is click fraud in the practical sense.
Yandex describes a wider basket. In the Yandex Direct antifraud guide, invalid traffic means clicks and conversions that are not tied to a real interest in the advertiser. The basket includes deliberate clicks from bots and from people paid to complete a task, and accidental repeated clicks. When the system accepts that classification, it does not charge and it removes the event from Direct reports.
Advertisers use a narrower word. On the help page «Что делать, если меня скликивают», Yandex answers that clicks the system treats as erroneous or invalid do not enter the reports and are not charged. A complaint about скликивание is about the clicks that stayed in the report.
What Direct has already taken off the bill
The filter looks at traffic before an impression, at the moment of the click, and after the visit. If the system is confident it is facing a bot or a fraudster, the ad is not shown. A suspicious click can be stopped with a captcha. What still reaches the site and is later classed as fraud is removed from the reports and is not used to train the bidding strategy.
The removed events sit in the “Invalid traffic” report in the report library. The number is not frozen: some events are reclassified later. The same guide says spend on invalid traffic is returned to the shared account balance. The return covers what the system accepted, not every session that looks odd in Webvisor.
Webvisor shows every visit, including visits already filtered. You are not charged for the filtered ones. The “Direct” report group in Yandex Metrica is built from clicks that passed the antifraud check. Other Metrica slices do not apply that filter. The same session can look alarming in Webvisor and be absent from spend.
| Where to look | What is inside | Money |
|---|---|---|
| Webvisor | Every visit, including filtered ones | Filtered visits are not charged |
| “Direct” reports in Metrica | Clicks that passed antifraud | A counted click is paid |
| “Invalid traffic” report | What antifraud removed | Not charged; spend goes back to the balance |
| Campaign spend | Clicks left after the filter | Charged |
Where the money goes when the click is empty
Suppose a campaign received 1,000 counted clicks in a week at 40 ₽ each. Spend is 40,000 ₽. That is arithmetic, not a market average.
If some of those clicks were not buyers, the cost per click is already gone and no lead came with it. On search, the empty click eats the daily budget you would otherwise have spent on a person who searched. On the ad network, the click has a second side: the publisher can earn from the visit. Google classes both automated clicks and manual clicks that raise an advertiser’s costs or a publisher’s revenue as invalid traffic, including clicks from competitors (Google Ads on invalid traffic).
Who clicks, and why, is covered in who clicks your ads. For the bill, the motive is secondary. An empty click that stayed in spend costs its price.
What is not click fraud
A real person who disliked the price, or who was comparing offers, is not fraud under Direct’s antifraud rules. The guide lists these visits explicitly as traffic the filter does not remove. No lead is not yet click fraud. How to separate a poor-fit person, a broken goal and an empty click is the subject of clicks without leads.
A search crawler that walks the site to index it is not an ad click either. It can raise the robot share in Metrica and still never appear in Direct spend. The two systems are different, and mixing them up is expensive.
How to compare the reports for one period
Pick one interval, a single day is enough, and read two numbers.
First the “Invalid traffic” report: how many clicks antifraud removed. Then campaign spend, clicks and conversions: what you were charged for. If Webvisor shows repeated visits or visits that last seconds, Direct’s help tells you to open that day by the minute and compare it with the “Direct” report group. A session that appears there passed the filter. A session that is missing from that group is already out of spend.
From there the useful evidence is a trace, not the feeling that “someone is draining us”. Those traces are collected in signs of click fraud. How to turn the remaining spend into an estimate in rubles, rather than a guess, is in the budget-loss calculation.
The comparison only sees what has already happened. It does not see the visit at the second the page opens, and it cannot split a run of empty visits from one real person when both passed Direct’s filter. On the site that is a separate check. A script looks at behavior, the browser environment, the network and the ad-click parameters. The decision is one of three: let the visitor through, ask the browser to solve a short computational task, or refuse the visit. If the check does not answer, the visitor is let through, so a real person is not locked out by an outage. Binetro is not an ad platform and does not return money Direct has already charged. The refund is the platform’s decision.