Clicks come in, leads do not

Clicks with no leads usually mean the goal was never recorded, a person arrived and did not buy, or the click was counted before the site opened. An empty click that stayed in spend is a fourth possibility, and it is the one to check last. Start there, and it is easy to “find fraud” on a broken form.

Suppose a day brought 80 clicks and zero leads at 50 ₽ a click. Spend is 4,000 ₽. That is an example, so there is something to divide. Until you know the goal is alive, there is nothing to divide. A zero in the denominator does not say who clicked.

The goal may not have been recorded

Direct’s recommendations describe the check in plain steps: complete a conversion yourself and confirm that Metrica recorded it. Until that manual lead shows up in the report, the campaign says nothing about demand. Zero leads beside living clicks are then a property of the tag, not of the traffic.

The second mix-up is the date. In the description of the “Maximum conversions” strategy, Yandex gives its own example: a click on 3 July, a lead on 5 July, a charge on 5 July. In the report wizard the lead belongs to 3 July, because it is tied to the click. In the pay-per-conversion report it belongs to 5 July, the day it happened. A week of “clicks came in, leads did not” is sometimes two different calendars.

Direct also warns that a strategy learning from a goal wants on the order of 10 completions a week. Below that, the algorithm is short of examples. That is not a fraud diagnosis. It is a limit of the strategy, and it is fixed with a goal higher in the funnel, with budget, or with the conversion price, not with a filter on visits.

A wrong email or phone in the form, a review typed into a lead form, and a form filled in for a promised prize are not invalid traffic. Antifraud will not remove them. The form has to be fixed.

A person arrived and did not become a buyer

Direct does not filter real people for whom the price was wrong, the service was wrong, or who were only looking over the market. The clicks are genuine, and the absence of leads is genuine too. The ad may have been clicked because it was easy to notice: click-through rate rose, purchases did not. How that number works, and why it has no single norm, is covered in what CTR is.

The cost of a lead rises in that situation with no software involved. There is spend, and there are few leads. How to calculate it, and how not to confuse it with the price you set on the strategy, is in how to calculate cost per lead. Until the page and the offer have been checked against real interest, there is nothing from which to subtract a “share of bots”.

The click was counted, and the site may not have opened

Google counts a click even if the person never reaches the site, for example because the site was unavailable (Google Ads definition of a click). The account shows a click, the visit and the lead do not exist, and there is no fraud in that.

For media placements, Direct counts the click in its own system before the redirect to the advertiser’s site. That is described in the IAB compliance note, and it applies to media ads, not to every search ad. The practical point is narrow. A click in the report and an open page are not the same second.

If the site was down, Direct has an availability monitor: it pauses impressions until the site answers, so the budget is not spent on nothing. That is not a refund of clicks that already happened. It is a reason to check whether the page answered on the day the leads went to zero.

What remains after the three checks

A manual lead appears in Metrica. You have aligned the click calendar and the lead calendar. The page opens, and the offer does not explain the zero. The clicks are in spend, not only in Webvisor among visits already removed.

Now it is a candidate for an empty click. How to read the traces without treating a spike as proof is covered in signs of click fraud. A robot share by source can sit beside that check, but it does not replace it. A search crawler was never supposed to leave an ad lead. How to read the share without confusing it with the bill is in robotness in Yandex Metrica.

Yandex also writes that it can check conversions for fraud only for the last 30 days. Older detail is technically unavailable. If the gap is older than that, the account will no longer split it by conversion.

A filter on site visits fixes none of the first three causes. A broken goal, an offer people refuse, and a page that does not answer produce zero leads even when every visitor is real. The check belongs on the remainder: behavior, the browser environment, the network and the ad-click parameters, with a decision to let the visitor through, to give the browser a short task, or to refuse the visit. Turned on over a broken form, it may shrink spend on empty visits, and it will not create leads out of nothing. If the check does not answer, the visitor is let through.