• Возможности
  • Тарифы
  • Вопросы
  • Блог
ВойтиНачать бесплатно
0%
Binetro

Защита

  • Защита от ботов
  • Защита от скликивания

Аналитика

  • Аналитика трафика
  • Отчёты

Подключение

  • Тильда
  • WordPress
  • Любой сайт

Начало работы

  • Как подключить
  • Тарифы
  • Блог

Документы

  • Политика конфиденциальности
  • Договор публичной оферты
  • Согласие на обработку данных
  • Политика использования cookie
  1. Blog
  2. Analytics and solutions

How to prove to a client or a manager that ads are being click-fraudded

Published October 25, 2026Binetro editorial teamAnalytics and solutions11 min read—
Contents
  1. Sheet one: a report that can still arrive
  2. Sheet two: the visit is visible, the charge is not
  3. Sheet three: counted clicks and yclid
  4. Sheet four: a call that did not land on that day
  5. The folder, in order
  6. A meeting as an example
  7. Folder checklist
  8. How to say each sheet out loud
  9. When a call stays out of the folder
  10. What to refresh if the meeting moves

Contents

  1. Sheet one: a report that can still arrive
  2. Sheet two: the visit is visible, the charge is not
  3. Sheet three: counted clicks and yclid
  4. Sheet four: a call that did not land on that day
  5. The folder, in order
  6. A meeting as an example
  7. Folder checklist
  8. How to say each sheet out loud
  9. When a call stays out of the folder
  10. What to refresh if the meeting moves

Click fraud looks like traffic growth

Binetro checks ad traffic and stops click fraud before it causes further budget losses.

Protect your ads

Share

You can prove click fraud to a client only with what the account can already show. Direct’s report library has an “Invalid traffic” report, and the number in it can change if filtering happens later. Webvisor shows visits that are already filtered, and those visits are not charged. Metrica’s Direct report group counts only counted clicks. The {yclid} tag ties a click to a visit, and a call becomes a goal only when the link to a visit succeeds.

A whiteboard for the review: the invalid-traffic report, the click id and the call date.
Proof is a folder of reports, not a screenshot of one chart. · Photo: CC0, Steen919

Click fraud in this conversation is not a feeling about a chart. It is a folder. Below is what to put in it, in which order to show it, and which sentence on each sheet is honest. Click fraud and sklikivanie stay the definitions. This is the assembly of the proof.

Sheet one: a report that can still arrive

Open “Invalid traffic” in Direct’s report library and write down the period you are defending. Next to it, write the export date. The number can change if filtering runs later. For a manager this is not a courtesy caveat. It is a property of the report: today’s figure is a snapshot, not a verdict.

Show two snapshots if the conversation lasts more than a day. The gap between them is later filtering, not “the report is lying”. Invalid clicks are exactly about why not every suspicious click is already in this report.

Do not replace this sheet with a screenshot of spend. Spend and the invalid-traffic column answer different questions. The client should see the report’s name.

Sheet two: the visit is visible, the charge is not

Webvisor shows visits that are already filtered. Money is not charged for those visits. A recording on the table proves the visit can be watched. It does not prove the budget was spent on it.

The sentence for the meeting: “this visit is in Webvisor, it is not in the charge, because it is already filtered”. The reverse sentence — “the recording exists, so the money left” — contradicts the help text.

If the client brought a stack of recordings, split them into filtered and the rest. Only the rest go into the folder as evidence of a charge, and only together with the sheet of counted clicks. The first group stays an appendix: “visible, not paid”.

Sheet three: counted clicks and yclid

Metrica’s tag help says Direct sets yclid. A sample link looks like https://example.com/?yclid=123. That is a format, not your live link.

Ad visits are read in the Direct group: summary, placements and costs. Statistics in these reports count only clicks that click-fraud protection did not filter. For the proof “here is the spend”, take this group, not the whole counter.

In Direct’s table of dynamic parameters, {yclid} is the identifier of a click on the ad. The tag is there to trace the visit source, and it is passed as a number. Beside it, {source_type} distinguishes search (search) and networks (context). {source} passes none on Yandex search, and the placement domain on networks. On YAN search placements either a domain or none can arrive.

{source}=none on search is a normal value, not an empty placement and not evidence. Write that in the folder, or the client will read none as “they hid the placement”.

Sheet four: a call that did not land on that day

Metrica does not link a call to a visitor and a visit if the identifier you passed is not in the database, if the visit was after the call, or if the visit was earlier than 21 days before the data was sent. “Call” goal statistics appear when the link succeeds.

The date in “Calls, in detail” is the moment of the call. The goal achievement sits on the visit, which may have been earlier. In the help-text example the visit is 11 March and the call is 13 March: on 13 March the detailed report shows the call, and “Conversions” for that day has no achievement.

For a manager this is a frequent “hole in the proof”. The call on the 13th is there, the conversion on the 13th is not — and that is the help-text example, not a lost lead. Put both reports in the folder and one line: visit on 11 March, call on 13 March, the goal sits on the visit.

Twenty-one days is the call-linking window. A call on a visit older than that window will not become a goal, even if the conversation was real.

The folder, in order

  1. Choose the period and the export date of “Invalid traffic”. Note that the number can still arrive.

  2. Separate Webvisor visits that are already filtered: they are visible, and they are not charged.

  3. Take spend and clicks from the Direct group — counted clicks only.

  4. Attach yclid to disputed visits as the numeric click id. Next to it, {source_type}: search or context.

  5. Read {source} by place: none on Yandex search is normal, a domain is a network placement. On YAN search both values are possible.

  6. Match calls to a visit. No identifier in the database, a visit after the call, or a visit older than 21 days before the send — there will be no link, and “Call” goal statistics will not appear.

  7. Do not add “Calls, in detail” and “Conversions” into one day without a check. The help-text example: visit on 11 March, call on 13 March, no achievement in “Conversions” on the 13th.

  8. The closing slide is three numbers for the period: invalid traffic on the export date, counted clicks in the Direct group, calls that linked. A fourth number, “all Webvisor recordings”, is not spend.

Who clicks ads, and why does not go on the slide instead of these three numbers. A motive without a report is a hypothesis. Whether Yandex refunds click fraud is a separate conversation after the folder shows traffic that has already been recognised, not instead of the folder.

A meeting as an example

A teaching example, not a client case.

The period is 9–15 March. On 16 March “Invalid traffic” shows one figure. On 18 March it is higher: filtering ran later. Both snapshots are on the table.

Webvisor has a visit on 12 March that is already filtered. It is shown with the sentence “money was not charged”. It is not put into spend.

The Direct group for the week is counted clicks. A disputed visit has a numeric yclid, {source_type}=context, and {source} is a placement domain. That is a network, not search. A neighbouring visit with {source_type}=search and {source}=none is labelled as Yandex search, not as an “empty placement”.

The call on 13 March is in “Calls, in detail”. “Conversions” on 13 March has no achievement. The visit was on 11 March, the link succeeded, the goal sat on the visit. The client is shown both reports, not one “missing lead”.

If the click identifier had been absent from Metrica’s database, this call would not have become a goal. The folder then honestly says “no link”, not “click fraud ate the call”.

Folder checklist

  • The report is named “Invalid traffic”, with a period and an export date.

  • It is said that the number can change if filtering runs later.

  • Filtered Webvisor visits are labelled: visible, not paid.

  • Spend is taken from the Direct group, that is, from counted clicks.

  • Disputed clicks have yclid as a number.

  • {source_type} separates search and networks.

  • {source}=none on Yandex search is not called a hole.

  • A network has a domain, and YAN search is allowed both a domain and none.

  • The call is younger than 21 days before the send, not earlier than the visit, and the identifier is in the database — otherwise do not expect the goal.

  • The detailed-report date and the conversion date do not have to match: the example is 11 and 13 March.

While any item is empty, “we proved click fraud” is early. You can already say: “here is a snapshot of invalid traffic, it can still arrive” and “here are visits that were not charged”.

How to say each sheet out loud

A manager hears short sentences. A long retelling is not required, but swapping one sentence for another breaks the folder.

The invalid-traffic sheet sounds like this: “on the export date this much was filtered, and the figure can grow if filtering arrives later”. It does not sound like: “we counted this much by eye”. The eye is not in this report.

The Webvisor sheet sounds like this: “the recording exists, the visit is already filtered, there is no charge”. It does not sound like: “look, a bot walked the site, so the budget burned”. The help text separates visibility and payment.

The Direct-group sheet sounds like this: “these are counted clicks, protection did not filter them, read spend here”. The whole counter is not substituted into that sentence.

The tags sheet sounds like this: “yclid is the click number, Direct sets it; search or context says whether this is search or a network; none on Yandex search is a normal value”. A domain in {source} names a network placement. On YAN search, be ready for both values and do not build a conclusion on one none.

The call sheet sounds like this: “the goal appears only after a link; the detailed date is the moment of the call; the achievement can sit on an earlier visit”. Say the 11 and 13 March example out loud, or an “empty conversion day” becomes an accusation.

When a call stays out of the folder

A second teaching example.

The click is on 1 March. The call export names an identifier, but it is not in Metrica’s database. The call is on 20 March. Even if the conversation is long, there will be no link: an identifier missing from the database is not enough to tie the call to a visit. “Call” goal statistics will not appear. The folder line is “a call without a visit”, and it is not added to invalid traffic.

Another branch of the same rule: the visit is on 1 April, the call is on 28 March, the data was sent on 2 April. The visit was after the call. There is no link. That is not a “late lead”. It is the help text’s rule.

A third branch: the visit is on 1 February, the call data is sent on 1 March. More than 21 days passed between the visit and the send. There is no link, even if {yclid} was once on the visit. Count the window to the send, not to the conversation in a manager’s head.

In all three branches the honest sentence is one: the goal did not appear because the link failed. Click fraud does not follow from that and is not refuted by it. For click fraud, the sheets that remain are invalid traffic, filtered Webvisor and counted clicks.

What to refresh if the meeting moves

The folder ages with the report. Before the next meeting, export “Invalid traffic” again and write the new date. If the number grew, show both snapshots and say filtering ran later. An old screenshot without a date looks like a trick, although it is ordinary report behaviour.

yclid and {source} do not “arrive later”: they were on the click link. Recheck them only if new visits were added to the folder. Recheck calls always: the link depends on whether the data was sent within 21 days and whether the identifier was found.

After the meeting moves, rebuild the closing slide. Three numbers: a snapshot of invalid traffic on the new date, counted clicks in the Direct group, and calls with a successful link. Webvisor recordings still sit beside them with the note “not paid” if the visit is already filtered.

Click fraud looks like traffic growth

Binetro checks ad traffic and stops click fraud before it causes further budget losses.

Protect your ads

Related

  • How to clean analytics of bots so the reports stay honest

    Which Metrica reports already count clicks after antifraud, why the CRM still shows more, and how a broken form goal is not the same thing as bots.

    12 min read
  • How to choose and roll out click-fraud protection

    What to check before you turn protection on: one request per visit, three decisions, an observation mode, and what the docs do not promise about page speed or spend before and after.

    12 min read
  • Why manual click-fraud protection does not scale

    Why Yandex Direct’s help calls manual placement review irrational, advises against blocking impressions by IP, and how geotargeting looks beyond the person’s current point.

    6 min read
  • Does bot protection affect SEO and real visitors

    How robots.txt, Google crawlers, and Yandex robots relate to site access, which Core Web Vitals thresholds Google sets, and what the Binetro script does when a check does not answer.

    6 min read
All articles