How much ad spend research says fraud takes

The two figures usually offered as “how much fraud eats” do not measure a Yandex Direct account. The ANA and White Ops forecast is 5.8 billion dollars globally for 2019, for video, display and other CPM formats. The TAG benchmark for the United States in the first half of 2023 is 0.82% IVT in certified channels and 1.19% in uncertified ones. Search was not in the first study. The second is not about search either, and not about spend in a Russian account.

Below are the sample, the year, and why the percentage does not transfer to a campaign. What Direct has already returned to the balance is not added to these billions: that is spend on traffic it classed as invalid, and the boundary is set in does Yandex return money for click fraud. The letters IVT in the TAG benchmark are GIVT plus SIVT at the measurers. The classes are unpacked in GIVT and SIVT.

5.8 billion is a forecast on impressions, and search is out

The fourth Bot Baseline report, from the ANA and the company White Ops, measured advertising of 50 ANA member companies from 1 August to 30 September 2018: 2,400 campaigns and 27 billion impressions. Search and paid social were not in the study. Losses from fraud that got through and was paid, the authors forecast at 5.8 billion dollars globally for 2019. The formats are video, display and other CPM placements on desktop and mobile.

They estimate fraud attempts at 20–35% of impressions over the year, and the paid remainder as smaller. These are White Ops vendor figures. The ANA press release is dated 1 May 2019. The forecast for 2019 is built from measurements in August and September 2018.

The figure is not Yandex Direct statistics and not the share of Russian search. The 20–35% is the share of attempts among impressions in that study, not the share of your spend and not “one click in three on search”.

0.82% and 1.19% are a US benchmark for half a year

The TAG US Fraud Benchmark, for 1 January to 30 June 2023, was assembled by The 614 Group from agency holding companies and their MRC-accredited measurers. There was no sample: every impression they were given was counted, 871.9 billion. The channels are display and video on desktop, in the mobile web, in apps and in CTV, and the geography is the United States.

The IVT share in channels where the whole TAG chain is certified is 0.82%. In uncertified channels it is 1.19%. IVT here is GIVT plus SIVT under the classification of the measurers DoubleVerify, IAS and Moat. These are figures from an industry program and from measurement vendors.

The 0.82% belongs to the certified channels of that agency transmission, not to “all internet advertising”. Direct search is not in the report.

Those percentages cannot be multiplied by search spend

Suppose search in Direct spent 100,000 rubles in a month. Multiplying that sum by 0.82% or by 20% is not supported. The first percentage is taken from display and video in the United States. The second is the share of attempts among impressions of another year and another format, and the authors call the paid remainder smaller. The example shows the limit of the transfer, not your campaign’s loss.

What Direct returns to the shared account balance is spend on traffic it has already classed as invalid. That is not an estimate of what stayed in spend, and not a conversion of the forecast’s billions into the cabinet’s rubles. What stayed has to be counted from clicks that are still on the bill after the filter. The method is in the loss calculation. A click that stayed in spend does not belong to the returned column. That boundary is set in what ad click fraud is.

A study with a stated method of paid fraud specifically in Direct, or in Russian search, is not among the sources checked. A foreign percentage cannot be dropped into that gap.

An on-site visit check does not count money and does not run the cabinet. The check’s documents do not set a way to measure a protection effect on ad spend before and after. The ANA forecast and the benchmark row do not become your campaign’s loss because a script is on the page.