How to find suspicious placements in the Yandex Advertising Network

A suspicious placement in the Yandex Advertising Network is a row in a report, not a verdict. The Placements report already carries the name, the spend, the clicks and the conversions. The guide itself treats a conclusion drawn from one or two conversions as unreliable.

A high share of one placement among paid clicks is a signal to look at the visits, not proof of an order. That is covered in signs of click fraud. Here the question is which report to open, and what a row does not say.

Where the list sits

The report library describes a ready-made Placements report. It already filters placement type to Networks and campaign status to Active and Stopped. The table groupings are placement type, placement name, campaign number and campaign name. The metrics are spend, impressions, clicks, conversions, CR and CPA. If those are not enough, Groupings and metrics adds others, income and the share of advertising expense for example.

The path into the wizard, if you build the report yourself: in the Direct menu, Statistics → Report wizard. To see where a click came from, the wizard guide suggests the groupings ad number, Placements, impression-condition type and click location, and the metric Clicks. You set the period.

The library guide gives an example of a column filter: keep placements with spend over 1,000 rubles for the reporting period. That is an example of how to narrow the table, not a threshold that says the placement is bad. A row cheaper than that sum does not vanish as honest, and a row that costs more does not become click fraud.

The same report already includes the statuses Active and Stopped. A row from a stopped campaign does not disappear from the table. The campaign number sits on the same row as the placement name. Before you read it as today’s attack, look at that name: the spend may be left from the days when the campaign was still running. The report does not label that separately. It simply does not drop the stopped ones.

What a row has not proved yet

The statistics guide says outright that going through placements by hand and turning off impressions on ineffective ones is not a sound use of time. It recommends the Maximize conversions strategy instead, and for other strategies it recommends setting the key goals. The bid then follows the conversion forecast: the less convertible the traffic from a placement, the cheaper the clicks.

The same page warns about small numbers. If a row has one or two conversions, their price in the report can differ sharply from the price you would see with a great many conversions. When you compare two slices, you allow for error. Suppose placement A shows 2 conversions in the row and placement B shows 40. The second row is steadier for a conversation about price, even if the first row’s CPA looks more frightening. The numbers are round and they are not from an account: they show why the guide refuses a conclusion from one or two conversions. They are not a threshold of “bad”. Comparing two slices, the guide asks you to allow for error: the figures you got can deviate from the true value. A CPA gap between placements, without that margin, is not yet an order to turn the higher one off.

CPA in this table is a field of the report. How cost per lead is calculated from your own spend and your own leads, and where it diverges from the account field, is covered in how to calculate CPA. The formula is not repeated here.

Zero conversions next to clicks is not the name of a harmful placement either. The goal may not be recording. That order is in clicks without leads. Until the goal is checked, a row of zeros stays a row.

What a row is not

A placement name does not say whether a person clicked there or a program did. A kind of program is not assigned from a site name. Which purposes the guides actually name is collected in kinds of bots. A click that stayed in spend and a click Direct has already removed are different events. The line is set in what ad click fraud is.

This article cannot turn a placement off. The guide calls manual disabling unsound, and the click that excludes a placement is a separate step, which is not here. First the row has to survive two checks. It has more conversions than one or two, otherwise the price jumps. And the empty leads are not explained by a broken goal. Even after that, the row is a reason to look at visits, not a refund.

A check of the visit on the site does not replace the Placements report. It looks at behavior, the browser environment, the network and the ad-click parameters, and it decides to let the visit through, to give the browser a short task, or to keep it out. If the check does not answer, the visit is let through. It does not print a placement name. Binetro does not turn a placement off in Direct and does not return the spend on a row.

A short pass:

1. Open the Placements report in the library. The Networks filter is already on.

2. Narrow the table with a column filter if there are too many rows. The guide’s example is spend over 1,000 rubles for the reporting period. That is not a quality score.

3. Set aside rows with one or two conversions: the guide treats their price as unreliable.

4. For rows with zero conversions, check the goal first.

5. Do not turn a placement off from a single evening. The guide recommends a strategy with goals, not a hand-made list.