Viewability, Attention Metrics and Brand Safety Explained: Definitions, Standards and Vendors

Each metric answers a different question, comes from a different document and is audited separately. The table sets them side by side, with accreditation status as the MRC listed it on 9 November 2026. Invalid traffic sits in the last row because every other metric depends on removing it first.

Viewability is the oldest of the three checks. The MRC wrote the rules with the IAB, and almost every vendor and platform measures against them. The core definition sits in the desktop guidelines, version 2.0, dated 18 August 2015 (MRC, 2015).

For a display ad, at least 50% of the ad's pixels must sit in the visible part of an in-focus browser tab. The ad must stay there for at least one continuous second after it renders. For a video ad, the same pixel rule applies while two continuous seconds of the ad play. Those two seconds need not be the first two. Large display ads get a lower bar: a unit of 242,500 pixels or more, the size of a 970x250 billboard, counts at 30% of pixels (MRC, 2015).

Mobile got its own document on 28 June 2016. It applies the same 50% and one-second test to mobile web and apps and replaced the interim guidance of May 2015 (MRC mobile guidelines, 2016). Google uses the same thresholds in its Active View reporting for Google Ads and YouTube (Google Ads Help, accessed 9 November 2026).

Has anything changed since? Not in the core rule. The MRC's standards list on 9 November 2026 still showed the 2015 and 2016 documents as the current viewability guidance (MRC, accessed 9 November 2026). Two newer documents use stricter versions for other purposes. The Out-of-Home Measurement Standards of December 2025 require 100% of the ad on screen (MRC OOH Standards, December 2025). The attention guidelines of November 2025 build on viewability rather than replacing it, as the attention section below explains.

The MRC asks vendors to sort every rendered impression into three buckets: viewable, not viewable, and undetermined. Undetermined means the tag could not tell, for example because it never loaded. From those buckets come three numbers that should always be read together (MRC, 2015).

The trap is in the last row. A seller can quote a 60% viewable rate while half of its impressions were never measured. The MRC says vendors should not call the 30% figure the viewable rate, because that assumes every unmeasured impression failed (MRC, 2015). Buyers should ask for both rates in every report.

Unmeasured supply is a real cost. In the Association of National Advertisers (ANA) benchmark for Q2 2026, lower-performing advertisers lost 22.7% of programmatic spend to non-measurable inventory, against 10.4% for higher performers. Spend on non-viewable impressions fell from 13.3% to 10.1% across the benchmark in one quarter (ANA via Fiducia, 6 August 2026). The working media benchmark tracks the series.

What is a good viewability rate? No standard body sets a target. Benchmarks come from vendors. IAS reported 67.9% for display and 79.7% for video globally in July 2026, based on its own measurement (company data) (IAS, 9 July 2026). Many buyers write a viewability floor into deals, or pay only for viewable impressions on a vCPM basis.

Here is what that means in money, using the MRC's example rates. A buyer pays a $5.00 CPM for 1,000,000 impressions, or $5,000. With a 50% measured rate and a 60% viewable rate, 300,000 impressions are confirmed viewable. The cost per thousand confirmed viewable impressions is $5,000 ÷ 300, or $16.67. Our CPM and vCPM converter runs the same sum on your own figures. The $5.00 price is illustrative; the rates come from the MRC example.

On a web page, a vendor measures with a script that loads with the ad. The script checks where the ad sits in the browser window, how much of it shows and for how long, then reports events to the vendor's servers. The verification vendor, not the publisher, decides whether the impression met the rule.

Apps are harder, because a script cannot see the screen the way a browser lets it. The IAB Tech Lab released the Open Measurement SDK (OM SDK) on 10 April 2018 to solve this for mobile apps. App owners build in one software kit, and many verification vendors read its signals, instead of each vendor shipping its own kit (IAB Tech Lab, 10 April 2018). A web video version followed in December 2020, and later releases added CTV platforms, including Samsung and LG TVs in version 1.5 (IAB Tech Lab, page updated 4 November 2025).

Platforms are moving to the same kit. Google said it would switch web video viewability measurement in Active View to OM SDK from June 2023 (Google Ads Help, accessed 9 November 2026). In connected TV (CTV), coverage is still partial. The MRC's accreditation of DoubleVerify's CTV video viewability applies to certified applications only, and IAS's to certified traffic only (MRC accredited digital services). Our page on CTV measurement companies covers the TV side in depth.

MRC accreditation means an auditor checked a vendor's method against the guidelines. Each accreditation names the metric and the screens it covers, so one vendor can be accredited on desktop but not on CTV. On 9 November 2026, the MRC's display viewability list included six verification vendors: Adloox, DoubleVerify, HUMAN, IAS, Pixalate and Protected Media. It also named ad servers and buying tools from Adform and Google, and platforms such as Amazon DSP, Instacart and Snap (MRC accredited digital services).

Scope differs in ways that matter. DoubleVerify's display viewability accreditation includes pre-bid data. IAS holds a separate accreditation for pre-bid viewability segments, used to target likely viewable inventory before a bid (MRC accredited digital services). Retail media is catching up. Walmart Connect's onsite display audit, covering viewable impressions, had been in process since November 2025, and Kevel's pre-audit was reviewed in September 2026 (MRC services under review). Our guide to retail media measurement standards covers that side.

Viewability is a test of the ad's condition, not of the person. A fully viewable ad can sit on a screen while the user reads a text message. The IAB and MRC guidelines draw the line clearly. Viewability measures the opportunity to see. Attention measures whether the ad was seen or heard, and to what depth (IAB and MRC, November 2025).

Before 2025 each vendor defined attention its own way. An IAB task force of more than 200 members drafted shared guidelines, released for comment on 12 May 2025 (IAB, 12 May 2025). Version 1.0 was published in November 2025. It groups attention measurement into four methods (IAB and MRC, November 2025):

1. Data signals. Impression-level signals such as time in view, share of screen, scrolling, playback and interaction.
2. Visual tracking. Eye tracking and facial coding, usually with consenting panel members.
3. Physiological and neurological observation. Heart rate, brain activity and similar responses.
4. Panel or survey methods. Asking people what they noticed and remembered.

The guidelines set three conditions that matter to buyers. First, the MRC treats attention as a measure of people, so accredited attention needs a base of viewable impressions, full SIVT filtering and some sign that a user was present. Second, attention measured before or without viewability may be reported, but only as a separate, labelled diagnostic. Third, attention is not an outcome. The document says it should not be used on its own to judge campaign results such as sales or brand lift (IAB and MRC, November 2025).

The four names people search for take different routes to the same question. Two model attention from signals and train against outcomes or eye tracking. One started as an eye-tracking research firm. The table uses each vendor's own description and the MRC's lists, checked on 9 November 2026.

Adelaide. Its AU score predicts how likely a placement is to capture attention and drive results, rather than measuring one viewer's gaze. Adelaide said on 4 November 2026 that its partner programme had reached 100 members, adding video curators such as Cadent, Pixability, Sightly and Silverpush (company claim) (Adelaide, 4 November 2026). A concrete use: Duolingo began selling in-app ads with a contractual minimum AU score in September 2026. An earlier Adelaide audit found Duolingo's in-app video inventory scored 34% above Adelaide's benchmark (company claim) (Adelaide, 15 September 2026).

Lumen, IAS and DoubleVerify. Lumen's roots are panel eye tracking, and IAS uses Lumen data as one input to its score (IAS, accessed 9 November 2026). DoubleVerify builds its score from more than 50 exposure and engagement data points and is the only one of the four with MRC accreditation for attention on 9 November 2026 (MRC accredited digital services). Detailed profiles of DoubleVerify and Integral Ad Science are planned.

How should a buyer use these scores? Treat them as a ranking tool, then test. A brand might move 20% of a video budget toward placements with high attention scores and hold out a matched group. If sales or brand lift do not follow, the score is not helping that brand. The guidelines themselves recommend reporting attentive time relative to viewable time, which keeps the two metrics connected (IAB and MRC, November 2025).

The MRC defines brand safety as the practices and tools that keep an ad from appearing next to, or in, a context that can damage the advertiser's brand. Its September 2018 supplement added guidance for judging single pieces of content, such as one video on a platform, rather than whole sites (MRC supplement, September 2018). The same document notes that different buyers have different tolerance for risk. That is the idea behind suitability: a floor everyone avoids, and a band above it that each brand sets for itself.

GARM, 2019 to 2024. The WFA set up the Global Alliance for Responsible Media in 2019, after the Christchurch mosque attack was livestreamed on Facebook. Its tools were the Brand Safety Floor and the Adjacency Standards Framework. WFA said they helped cut ads beside harmful content on social media from 6.1% in 2020 to 1.7% in 2023 (WFA claim) (WFA, 9 August 2024).

The closure and the lawsuit. X Corp. sued the WFA in federal court in Texas on 6 August 2024. It alleged that GARM coordinated an advertiser boycott (Claims Journal via Bloomberg, 7 August 2024). Three days later, on 9 August 2024, the WFA discontinued GARM. It said allegations had drained the small non-profit's resources (WFA, 9 August 2024). On 29 July 2026 the WFA and X said they were putting the litigation behind them. The WFA added that it would not form or restart GARM or a similar initiative (WFA, 29 July 2026).

The taxonomy that remains. The IAB Tech Lab Content Taxonomy is the shared list of content categories, and the Tech Lab names contextual targeting and brand safety as its typical uses. Version 3.0, released in June 2022, grew the list from about 400 to more than 1,500 categories, and version 3.1 followed in December 2024. In February 2026 the Tech Lab published a free tool, donated by Mixpeek, that maps old 2.x codes to 3.0 (IAB Tech Lab, updated 11 February 2026).

One detail trips up many integrations. In an OpenRTB 2.6 bid request, the cattax field names the taxonomy behind a site's or app's categories. If a seller leaves it out, the default is Content Taxonomy 1.0 (OpenRTB 2.6). AdCOM marks that version as deprecated and lists value 7 for version 3.0 and 9 for version 3.1 (AdCOM 1.0). A buyer's suitability rules can misfire if the two sides read the codes from different versions. OpenRTB 2.6 explained shows where the fields sit.

A brand safety solution in programmatic buying works at three points. Each point catches different problems.

1. Before the bid. The buyer's DSP applies block lists, allow lists and pre-bid segments from a verification vendor, so it never bids on pages or apps that break its rules. DoubleVerify's MRC accreditation for property-level verification includes pre-bid data on desktop, mobile web, in-app and CTV (MRC accredited digital services).
2. At the seller. Publishers and SSPs block advertiser categories they do not want, using the OpenRTB bcat field (OpenRTB 2.6). This protects the publisher's audience rather than the advertiser's brand.
3. After the ad serves. A tag classifies the page where the ad ran. It can block the creative from rendering, log an incident and feed the report that supports a refund or a makegood.

Accreditation is thinner here than for viewability. Property-level verification, which judges a site or app as a whole, was accredited for DoubleVerify, IAS and Protected Media on 9 November 2026. Protected Media's scope covered English, Spanish and Portuguese only (MRC accredited digital services). Content-level accreditation covered only YouTube, through Google and Zefr. IAS's YouTube audit had been in process since April 2023, and Meta's third-party suitability data pipeline had its audit reviewed in August 2026 (MRC services under review).

Where is the risk concentrated? IAS's July 2026 report found mobile web display carried 45.1% of the impressions it measured but 54.9% of brand suitability violations (company data) (IAS, 9 July 2026). The same environment held most made-for-advertising (MFA) site impressions, a quality problem covered in our guide to made-for-advertising sites.

The common failure is over-blocking. A keyword list that bans "shoot" or "attack" also removes football reports and cybersecurity news. A news publisher then loses revenue on safe pages, and the advertiser loses reach among readers it wanted. Suitability tiers and content-level classification exist to replace blunt keyword lists with graded categories.

Each of the three numbers is useful, and each is easy to over-read. The list below sets out the limits a buyer should keep in mind before acting on a report.

• Viewability proves an opportunity to see, on measured impressions. It does not prove a person was present, looked at the ad or remembered it. It says nothing about the unmeasured share, which is why the measured rate comes first (MRC, 2015).
• Attention is an estimate of human exposure and engagement. Most products model it from signals and calibrate on panels. It is not a sales measure, and the IAB and MRC guidelines say so directly (IAB and MRC, November 2025).
• Brand safety proves classification against a rule set at a point in time. It does not prove how readers felt about the adjacency, and a page can change after it is classified. Accuracy depends on the taxonomy version and the classifier.
• None of them proves the traffic was human. That is the job of invalid traffic filtering, and accredited viewability and attention both require it. Our page on ad fraud detection tools compares the vendors.

A simple test keeps the metrics honest. Pick one campaign outcome you trust, such as incremental sales from a holdout test. Then check whether moving budget toward higher viewability, higher attention or stricter suitability changes that outcome. If it does not, the metric is a hygiene check for that campaign, not a lever.

1. Remove invalid traffic first. Use a vendor accredited for SIVT on every screen type you buy.
2. Ask for measured rate and viewable rate together, by seller and by environment, never a single blended percentage.
3. Set viewability terms in the deal, either a floor or payment on a vCPM basis, and agree whose numbers count.
4. Check accreditation scope, metric by metric and screen by screen, in each vendor's accreditation letter.
5. Treat attention as a test variable. Run it against a holdout before shifting large budgets.
6. Write suitability tiers, not keyword lists, and confirm which Content Taxonomy version your sellers send in cattax.
7. Keep impression-level logs so you can rebuild any vendor's rate from your own data.

1. Media Rating Council, "Viewable Ad Impression Measurement Guidelines (Desktop)", version 2.0 (18 August 2015). https://mediaratingcouncil.org/sites/default/files/Standards/081815%20Viewable%20Ad%20Impression%20Guideline_v2.0_Final.pdf
2. Media Rating Council, "Mobile Viewable Ad Impression Measurement Guidelines", final (28 June 2016). https://mediaratingcouncil.org/sites/default/files/Standards/062816%20Mobile%20Viewable%20Guidelines%20Final.pdf
3. IAB and Media Rating Council, "Attention Measurement Guidelines", final version 1.0 (November 2025). https://mediaratingcouncil.org/sites/default/files/Standards/IAB_MRC_Attention_Measurement_Guidelines_November_2025.pdf
4. IAB, "IAB and MRC Release Attention Measurement Guidelines for Public Comment" (12 May 2025). https://www.iab.com/news/attention-measurement-guidelines-public-comment/
5. Media Rating Council, Standards and Guidelines list (accessed 9 November 2026). https://www.mediaratingcouncil.org/standards-and-guidelines
6. Media Rating Council, accredited digital services (accessed 9 November 2026). https://mediaratingcouncil.org/accreditation/digital
7. Media Rating Council, digital services under review (accessed 9 November 2026). https://mediaratingcouncil.org/accreditation/services-under-review/digital
8. Media Rating Council, accredited out-of-home services (accessed 9 November 2026). https://mediaratingcouncil.org/accreditation/out-of-home
9. Media Rating Council, "Supplement to IAB Guidelines for the Conduct of Ad Verification: Enhanced Content Level Context and Brand Safety" (September 2018). https://mediaratingcouncil.org/sites/default/files/Standards/MRC%20Ad%20Verification%20Supplement-%20Enhanced%20Content%20Level%20Context%20and%20Brand%20Safety%20%28Final%29.pdf
10. Media Rating Council, "Out-of-Home Measurement Standards, Phase 1 and 2 Combined", final (December 2025). https://mediaratingcouncil.org/sites/default/files/Standards/MRC%20OOH%20Standards%20Combined_FINAL.pdf