What is View-through conversion?

A view-through conversion is credited when somebody was served an ad, did not click it, and converted within the view window.

The event chain is real: the impression happened and the purchase happened. What is absent is any evidence connecting them. Nobody clicked, nobody interacted, and the ad may have appeared below the fold on a page that was closed a second later.

The category exists because advertising without a click genuinely does influence people. A video watched in a feed, a display unit seen in passing, a brand encountered repeatedly — these contribute to purchases that arrive through search or direct traffic weeks later, and a measurement system that counted only clicks would credit none of it.

The difficulty is that the same counting rule that captures real influence also captures pure coincidence, and the report cannot tell them apart.

The pool of people eligible for a view-through credit is the whole audience served, which for display and video campaigns is enormous relative to the number who clicked.

A campaign serving a million impressions has a million opportunities to be credited. If the audience includes existing customers, people already researching the category, or a retargeting pool of recent site visitors, a substantial number of them will convert in the following days regardless of the advertising.

Three factors inflate the count and none of them is about effectiveness.

Audience size. More impressions, more eligible conversions.

Window length. A longer view window catches more of whatever was going to happen anyway.

Audience warmth. A campaign serving impressions to people already close to buying will be credited with their purchases.

That last one is the sharpest. A retargeting campaign counting view-through conversions is being credited for the purchases of people who had a full cart, and the resulting cost per acquisition can look extraordinary.

The practical handling is to report them separately.

A campaign's cost per acquisition including view-through conversions is not comparable with one counting clicks only, and most search campaigns count clicks only. Blending them into a single account-level figure makes the whole report incoherent while looking tidy.

Where they are reported, three things make the figure more honest.

A short view window. One day is defensible; a week is generous; a month is close to meaningless.

Viewable impressions rather than served ones, where the platform supports the distinction. An impression that never entered the viewport cannot have influenced anybody.

Exclusions applied. Removing existing customers and recent converters from the audience removes the conversions most likely to be coincidental.

Nothing in the reporting can settle it. The question is causal and needs a causal design.

A holdout is the direct answer: withhold the campaign from a random, comparable portion of the audience, and compare conversion rates between the held-out group and the served one. The difference is what the advertising caused; everything above it in the view-through count is coincidence.

Geographic splits work similarly where audience-level holdouts are unavailable, at the cost of more noise.

Accounts that run this test on a view-through-heavy campaign are usually surprised, and consistently in the same direction. That consistency is itself informative: it means the inflation is systematic rather than occasional.

Despite all of the above, there is a legitimate case and it is narrow.

Upper-funnel video and display, where the format produces few clicks by design and the influence is genuinely real. Counting only clicks would report almost nothing and would justify cutting activity that works.

Long consideration cycles, where the ad's contribution is early and the purchase arrives through another route entirely.

Connected television and audio, where clicking is not available and some non-click measurement is the only option.

In each of those, the right posture is to count them, label them, keep the window short, and validate periodically with a test rather than treating the figure as evidence on its own.

Report three numbers rather than one.

Click-through conversions, which carry the strongest evidence.

View-through conversions, separately, with the window stated.

The incremental figure from whatever test the account last ran, with its date.

An account that reports all three, and that knows roughly how the third compares with the sum of the first two, is in a position to argue about budget from evidence. An account reporting a single blended number that includes view-through credit is going to win that argument and lose the money.

One practical hazard deserves its own note, because it is easy to create and hard to see.

Where several platforms each count view-through conversions, and each is serving impressions to overlapping audiences, the same purchase can be credited as a view-through by more than one of them. Add a click-through credit somewhere else in the stack and a single order has been claimed three times.

Attributed figures were never meant to be summed, and view-through credit makes the over-claiming considerably worse, because the eligible pool is every impression rather than every click.

The blended ratio is the check. If the platforms' combined claims move further above total revenue after a view-through-heavy campaign launches, the campaign has added claims rather than sales, and the difference is measurable rather than a matter of opinion.

For an account without the appetite to run tests regularly, a conservative default does most of the work.

Count view-through conversions with a short window. Report them in their own column. Exclude existing customers and recent converters from the audiences that generate them. Keep the headline cost per acquisition on click-through credit only.

That arrangement will understate genuine upper-funnel contribution somewhat, which is the error worth making: it is visible, it is conservative, and it does not quietly justify a budget the business cannot support.