Viewability: what counts, who decides, and what to ask

Most viewability conversations in India go one of two ways. Either nobody defines it, or someone reports 94% and the room treats that as a result.

It is neither. Viewability is a pixel-and-time rule — a mechanical test that an ad unit either passes or fails. It belongs in the same box as the other delivery metrics in our performance marketing guide: useful, measurable, and easy to misread as proof of something it never claimed.

One boundary first. Viewability is not fraud. Whether a human was even there is a separate question, handled on our page on ad fraud and invalid traffic. Viewability assumes the impression was real and asks only whether it was on screen.

What counts as a viewable impression

The standard is written by the Media Rating Council (MRC) with the IAB. It is short, and it is specific.

For desktop display, the MRC Viewable Ad Impression Measurement Guidelines v2.0 (18 August 2015) require that "greater than or equal to 50% of the pixels in the advertisement were on an in-focus browser tab", and that the time the pixel requirement is met was "greater than or equal to one continuous second, post ad render".

Two details carry most of the misunderstanding. The clock starts after render, not after the ad call. And the tab must be in focus — an ad sitting perfectly in a background tab is not viewable.

The MRC thresholds for display, video and audio differ by format. Here is the full set, and what actually trips each one.

The large-format 30% rule is desktop only. The Mobile Viewable Advertising Impression Measurement Guidelines (28 June 2016) state plainly that "these guidelines do not stipulate a similar exception for mobile viewability". If someone applies a 30% threshold to a full-screen mobile interstitial, they are off-standard.

Audio is the honest outlier. The Digital Audio Measurement Standards v1.0 (January 2018) call audibility "an imperfect analogy to the 'opportunity to see' proxy" and say the two-second threshold "requires further study". Treat audio viewability claims with that caveat attached.

Pixels and time, nothing more.

Who decides, and why "MRC-accredited" is a narrower claim than it sounds

MRC writes the guidelines and audits the vendors. The IAB co-authors them. The vendors — Integral Ad Science, DoubleVerify and others — build the measurement and submit to audit.

The catch is scope. Accreditation is granted for specific products in specific environments, not as a blanket badge. IAS, for example, was accredited for impression, viewability and invalid traffic reporting on Amazon properties as a distinct grant in November 2025. An accredited vendor can still be unaccredited for the exact placement you are buying.

The vendor roster also moves. Moat is still named in briefs out of habit, but Oracle wound down its advertising business — the unit Moat sat inside — after revenue fell to about $300 million in its 2024 financial year, reported by Marketing Dive on 13 June 2024. If a 2026 media plan still lists Moat as the verification partner, that plan has not been read.

Why "viewable" is not "seen", and not "attended to"

This is the sentence worth memorising, and it is the industry's own. The IAB and MRC Attention Measurement Guidelines v1.0 (November 2025) describe viewability as "representing the opportunity to see the ad", applying "regardless of whether or not the ad was seen, viewed or heard by a person".

So a viewable impression means the ad was physically capable of being seen. Nothing about eyes, and nothing about memory.

That is what attention measurement sets out to add. The same document defines attention as whether the ad "was seen or heard by a person and to what depth", built from inputs including:

• Impression-level signals — time-in-view, scroll depth, media plays, interaction triggers
• Visual tracking — eye and gaze tracking, facial coding, presence monitoring
• Physiological observation — heart rate, EEG, skin conductance
• Panel and survey work — brand health studies, ad effectiveness surveys
• Predictive models that scale the above across impressions

Two things to hold onto. First, the guidelines say attention measures need "a base of compliant viewable or audible" impressions, because that base "enforces a minimum time threshold (one second for display and two continuous seconds for video)". Viewability is the floor attention is built on — which is exactly why hitting the floor is not an achievement.

Second, the same document concedes that "different approaches and methods within each approach may lead to different conclusions". Attention is not a settled currency either. It is a newer lens, not a replacement, and it matters most where the goal is memory rather than a click — the territory covered in brand awareness measurement.

Why two vendors measuring the same campaign disagree

They disagree because the guidelines leave implementation choices open, and vendors resolve them differently. MRC has said so in writing. Its Update on Viewable Impression Reconciliation Guidance (13 April 2015) names the causes directly:

• Multi-unit rules applied inconsistently — some vendors require every ad in a multi-unit buy to be viewable before any is counted
• Page Visibility API treated as proof — it often cannot definitively say whether a page is visible, yet gets used as an input
• Throttle indicator inconsistency — "we currently see inconsistency in when and how this indicator is applied"
• Different processing order — MRC warns that deviating from the prescribed order "could result in discrepancies across vendors"
• Count-on-decision served impressions — counting impressions that never rendered, which distorts the denominator

The denominator is the quiet one. MRC directs vendors to calculate performance rates using rendered served impressions, and to count unresolved cases as unmeasurable. A vendor that divides by a smaller or cleaner base will report a higher viewability rate on identical delivery. No one has cheated. The two reports are simply answering slightly different questions.

What a buyer should do about it is unglamorous and effective. Name one vendor as the measurement of record in the insertion order before the campaign starts. Take the second vendor as diagnostic, not as a rival truth. Reconcile at total campaign level, not line item. And ask for the measured rate alongside the viewable rate, because a 90% viewability rate on 40% of impressions is a weaker result than it looks. These are the same disciplines that make the rest of media buying auditable.

Viewability as a buying lever

Viewability is worth more as a control than as a score. You have three real levers.

Pre-bid targeting. In programmatic buying, a verification vendor scores an impression's predicted viewability before the bid, and the bidder skips what falls below your floor. You are buying a forecast, not a guarantee — predicted viewability and measured viewability will not match.

Make-goods. The IAB's State of Viewability Transaction 2015 (16 December 2014) proposed a 70% threshold on measured impressions, with publishers delivering additional viewable impressions until the threshold was met, applied to total campaign impressions rather than individual line items. Read that as the structure of a make-good clause, not as a 2026 benchmark — IAB framed it as a transition-year number.

Accepting the cost. Insisting on very high viewability is not free. The same IAB paper quotes MRC's position that "100 percent is currently unreasonable", citing variance across ad units, browsers, placements, vendors and methodologies. In practice, a high viewability floor shrinks eligible inventory, pushes you toward above-the-fold and large-format placements that cost more, and raises your CPM. If a 95% floor doubles your CPM to buy a few points of viewability you were never going to convert on, you have optimised the report, not the business.

One format note. In connected TV, the on-screen question is largely moot — the ad occupies the full screen. The open question there is whether anyone is in the room, which viewability was never built to answer.

What we ask for before we accept a viewability report

This is DH's own view, from running verification on display and video campaigns for clients. A viewability number with none of the following attached is not yet information:

1. The vendor and the accreditation scope — accredited for which product, in which environment, measuring the ad or the container
2. The measured rate — what share of served impressions the vendor could measure at all, and how unmeasurable was treated
3. The denominator, named — rendered served impressions, or something else
4. Format-level splits — desktop display, mobile display, video and large-format reported separately, never pooled into one headline figure
5. Which threshold was applied to large format — 30% desktop, or 50% as the mobile guidelines require
6. Whether video used ad pixels or player pixels — MRC requires prominent disclosure if the player is the basis
7. Time-in-view distribution, not just pass or fail — one second and nine seconds both score as viewable, and they are not the same media

We also refuse one comparison. The mobile guidelines say measurers "should not extrapolate desktop-based viewability rates to mobile-based ads". So we do not benchmark a mobile campaign against a desktop figure, and we do not let a vendor blend the two into a single rate.

On tolerance: MRC typically applies a materiality threshold of 5% when assessing whether a measurement limitation matters. That is a reasonable starting point for how much vendor-to-vendor drift to argue about, and how much to accept.

What a better viewability question looks like

Stop asking what your viewability rate is. Ask what share of impressions could be measured, what the time-in-view curve looks like, and whether the formats were reported apart.

Those three questions change the answer you get. They also move viewability back to where it belongs on a dashboard — a delivery check sitting alongside impression share and the rest of your marketing KPIs, rather than a headline. The outcomes still live downstream, in whatever your display advertising was actually meant to do.