What is Viewability?

A display campaign records 1,000,000 impressions, of which 940,000 could be measured and 592,000 met the viewable standard. The viewability rate is 592,000 ÷ 940,000 = 63%, because the standard rate is computed over the impressions that could be measured. The price is not: every one of the 1,000,000 was paid for. At a A$6.00 cost per thousand impressions, the cost per thousand viewable impressions is 6.00 × (1,000,000 ÷ 592,000) = A$10.14. Of the measured impressions, 37% did not meet the standard; of the impressions paid for, 40.8% did not.

An impression is recorded when an ad is delivered. That count makes no claim about whether the ad appeared anywhere a person could have seen it, and in display advertising a substantial share of delivered impressions never do.

The ad below the fold on a page nobody scrolled. The unit at the bottom of a long article. The placement in a background tab. The ad that began rendering as the reader hit back. All counted, all paid for, all unseen.

Viewability is the separate measurement built to address this. It asks a narrow question with a testable answer: was enough of the ad on screen for long enough that it could have been seen.

The widely adopted standard is deliberately modest.

As the standard states it: for display, at least half the ad's pixels in the viewport for at least one continuous second; for video, at least half the pixels for at least two continuous seconds. Large-format display units carry a longer dwell requirement, on the reasoning that a bigger unit takes longer to register.

Every one of those numbers is a floor rather than an aspiration. One second with half the pixels showing is a weak claim about human attention, and it is meant to be: the standard measures opportunity, not engagement. Attention measurement is a separate and much harder discipline.

Some buyers apply stricter internal standards — full pixels, longer dwell — and report against both. That is reasonable, and it makes cross-supplier comparisons harder unless the standard is stated.

A viewability figure has two numbers behind it and quoting only one is misleading.

The first is the measurable share: the proportion of impressions where the measurement script could actually run and observe the ad's position. Cross-domain restrictions, certain in-app environments, blocked scripts and some video players all prevent measurement. Those impressions are neither viewable nor non-viewable; they are unknown.

The second is the viewable rate among the measurable impressions.

A campaign reporting 70% viewability over a measurable base of 40% has told you about a minority of its delivery. One reporting 60% over a measurable base of 95% has told you far more. The pair belongs in the report together, and a supplier quoting only the rate is quoting the more flattering of two numbers.

The most practical use of the metric is arithmetic, and it changes supplier rankings regularly.

Take the cost per thousand impressions and divide by the share of the impressions you paid for that were viewable. The result is the cost per thousand viewable impressions, which is the price of inventory that had a chance of working.

The share matters more than it looks. A rate quoted over measurable impressions has to have the unmeasured ones put back in first, because they were bought too: 592,000 viewable out of 940,000 measured is 63%, and out of 1,000,000 paid for it is 59.2%. A A$6.00 placement at 59.2% costs A$10.14 per thousand viewable. A A$8.00 placement at 92% costs A$8.70. The second is the cheaper buy and the rate card says otherwise.

This conversion is not complicated and it is skipped constantly, which is how cheap inventory with poor visibility keeps getting bought. Any comparison of impression prices across suppliers, placements or formats should be done in viewable terms or not at all.

Position on the page explains most of it. Units above the fold are viewable far more often than units below it, for the obvious reason.

Page type explains much of the rest. Articles people read to the end deliver their lower units; listing pages and landing pages people bounce from do not. Infinite-scroll feeds sit somewhere in between and vary with how fast people scroll.

Format matters as well. Units that occupy the screen — full-screen interstitials, in-feed video, sticky units — are structurally more viewable than banners competing with content for space.

And device matters: phone screens hold fewer units at once, which raises the viewable share of the ones that are shown and lowers how many can be shown at all.

Because the variation is structural rather than random, it is predictable and therefore negotiable before a buy rather than discovered afterwards.

Where a supplier offers to sell viewable impressions rather than delivered ones, the risk of non-viewable delivery moves to them. That is worth something, and they will price it accordingly.

The comparison to make is not between the two headline rates. It is between the viewable-impression price and the ordinary impression price divided by its expected viewable share, which puts both offers in the same units. Frequently the guaranteed option is competitive once that arithmetic is done, and occasionally it is a substantial premium for a modest improvement.

Either way, the conversation is better than the one where nobody mentions viewability and the cheaper rate card wins.

It does not measure attention. An ad on screen for a second while somebody scrolled past it met the standard.

It does not measure outcomes. A perfectly viewable campaign against the wrong audience with the wrong message will produce nothing, and the viewability report will look excellent.

It does not apply evenly across environments. In-app and connected-television measurement differ from web measurement in what can be observed, and figures from different environments should not be pooled without saying so.

And it does not replace the rest of the panel. It is a hygiene check on the inventory: it tells you whether the thing you bought was capable of working, and leaves the question of whether it worked to everything downstream of it.

An impression is counted when an ad is delivered; a viewable impression is counted when enough of it was on screen for long enough to have had a chance of being seen. The gap between the two counts is not small, and it is not evenly distributed: it is wide on inventory below the fold, on pages people abandon, on long articles where most ad slots are never scrolled to, and narrow on formats that occupy the screen. Because the auction charges on delivery rather than on visibility, two placements at an identical cost per thousand impressions can differ by half in what they actually cost per thousand seen. The practical instruction is to convert every impression price into a viewable price before comparing suppliers, and to treat a cheap impression price with a poor viewable share as what it is — a discount on something that was not delivered.

• Convert every impression price into a viewable-impression price before comparing inventory. A cheap placement with a poor viewable share is rarely the cheap option it appears to be.
• Report the measured rate and the measurable share together. A high rate over a small measurable base says less than a moderate rate over a large one.
• Read viewability by placement and by fold position rather than as a campaign average. Position on the page explains most of the variation.
• Treat the standard as a floor, not a goal. An ad that met the minimum threshold was on screen for a second, which is a weak claim about attention.
• Ask what a placement's viewable share was before accepting a rate card. It is a negotiable fact about the inventory rather than a property of the campaign.