View-through conversions credit people who never clicked

View-through conversions credit people who never clicked

A case study says a paid social campaign drove a few hundred conversions. It gives the spend, the headline count and a confident sentence about impact. What it does not say is how those conversions were credited, and that missing detail decides whether you are reading about people the ad moved or people the ad merely met.

The conversion that has no click behind it

A view-through conversion is a purchase or sign-up credited to an ad because the person was shown it within a set window beforehand, without clicking it. A click-through conversion is the stricter cousin: the person interacted with the ad, then converted. Both can sit inside one number labelled “conversions”, and the label does not tell you the mix.

This is not a scandal. It is a counting rule, and a counting rule only becomes interpretable once it is disclosed. A figure of 100 credited conversions means one thing if 80 followed a click and something quite different if 70 followed nothing more than an impression. It is also worth knowing that the word itself can hide several stages, a point covered in the word conversion hides several funnel stages.

What the platforms say they count

Each platform states its own rule. None of these is an industry norm.

  • Meta. [EVIDENCE NEEDED: a readable copy of Meta’s attribution settings documentation showing how it defines view-through and which window it names]
  • Google Ads. Its page on conversion windows describes a separate view-through conversion window, and says that if you do not customise it when creating a new conversion, the default window is 1 day. It also says the advertiser can choose a custom number of days.

Two cautions follow. The advertiser also chooses the attribution model, per Google’s models page. Windows and defaults can be changed by the advertiser and by the platform over time, so a case study may not have used the default. A documented default tells you nothing about what a given agency chose. The length of the window is its own lever, discussed in the attribution window is a choice. This post is about whether a click was required at all.

A hypothetical split of 100 credited conversions

The numbers in this section are hypothetical and belong to no real campaign. Picture two campaigns that each report 100 credited conversions on a hypothetical spend of 5,000 in any currency.

Table, hypothetical example only: the same headline of 100 conversions under two invented splits.

Hypothetical campaign Click-attributed View-attributed Headline conversions Cost per conversion (headline) Cost per click-attributed conversion
A 30 70 100 50.00 166.67
B 80 20 100 50.00 62.50

Neither split is typical, and no real-world proportion is implied. The headline efficiency is identical: 5,000 divided by 100 is 50 per conversion in both. But the story about what the ad did is very different. In campaign A, most of the credit rests on exposure alone, and each click-attributed conversion costs more than three times the headline figure.

When view-through is a fair signal

There is an honest case for it. In formats where clicking is not the natural behaviour, such as video or awareness placements, it is plausible that exposure matters even when nobody taps. That is reasoning, not a measured fact.

It becomes more credible when reported separately and beside a click figure, so you can discount it to whatever degree you find sensible. A view-through number blended into one total leaves you no such choice.

What would actually settle the question is a test with a holdout: a comparable group deliberately not shown the ad. That comparison speaks to incrementality, and it is the subject of holdout tests. Without one, view-through tells you the ad preceded the purchase, not that it caused it.

When it turns into an inflation mechanism

The core problem is that someone who was already going to buy, and who happened to be served an ad, is counted the same as someone the ad persuaded. Exposure-based credit cannot tell them apart.

Overlap adds to this. Each platform runs its own attribution and credits itself, so if a buyer saw an impression on two platforms, both can claim the one purchase, and the totals can exceed the orders that happened. That is reasoning about independent systems, not a measured rate.

Retargeting sharpens the effect. Those audiences are made up of people who already visited or added something to a basket, so they are close to buying by construction, and exposure-based credit there is especially generous. A longer view window widens the net further. A related caution applies to counts that are estimated rather than observed, as in modeled conversions reported as measured.

Three questions to ask a case study

  1. What share of the reported conversions were click-attributed and what share were view-attributed? Good: two numbers that sum to the headline. Evasive: the headline repeated.
  2. What view window was used, and was it the platform default or a chosen setting? Good: the number of days and the source. Evasive: “standard attribution”.
  3. Was any of the result tested against a group that was not shown the ad, or is it all platform-attributed? Good: a described holdout and its size. Evasive: a description of the dashboard.

What to do when the split is not given

Treat the headline as an upper bound on ad-driven conversions, not as a measurement of them. The true click-driven figure is somewhere below it, and you do not know how far.

Then look for clues. Does the case study mention video, awareness or retargeting objectives? Was the campaign reach-heavy? Any of these makes it more likely that exposure did a lot of the crediting.

In the library, credited numbers stay attributed to the agency that reported them, so you always know whose claim you are reading. Browse the case study library, or submit your own.

FAQ

Are view-through conversions fake?

No. They are a documented counting rule that credits exposure. Whether the exposure caused the purchase is a separate question, and only a holdout test can address it.

Does a case study have to report the click and view split?

No rule requires it, but without it you cannot tell which claim you are reading. Treat it as something worth asking for, not an obligation.

Sources

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *