Cumulative attention is the total amount of attention a person pays to a brand's advertising across every exposure in a campaign, usually expressed in seconds. Instead of asking how long a single ad was looked at, it asks how much looking has piled up over five, ten or twenty impressions. The idea exists because most individual digital ads receive very little attention, often well under a second, while brand-building research suggests memory forms only above a certain amount of attention. If many brief glances can add up to that amount, an ad that fails on its own may still contribute to a campaign that works.

The term is sometimes used interchangeably with aggregate attention, the phrase Havas Media Network and Lumen Research chose in their 2024 study. It is a planning concept, not a standardised metric, and no industry body has defined it.

How it is calculated

The building block is the per-impression estimate that attention metrics vendors already produce. Lumen Research, the UK eye-tracking company, describes attention with two numbers for each impression: the probability that the ad was looked at, and the expected viewing time if it was. Multiplying the two gives an expected number of attentive seconds for that impression.

Cumulative attention sums those expected seconds per person. For each respondent, the Havas study combined every viewable impression's probability of being viewed with its expected view time, and also calculated the overall chance that the respondent looked at the campaign at all. The result is a person-level total, which can then be set against brand survey answers from the same respondent.

A worked example shows why the distinction matters. In an October 18, 2019 white paper, Lumen estimated that 1,000 desktop MPU impressions (the 300x250 medium rectangle) produced about 110 attentive seconds: 57% were viewable, around 8% were looked at, and those that were looked at held the eye for 1.3 seconds on average. Spread across a person who saw ten such ads, that works out at roughly 1.1 seconds of cumulative attention, below the thresholds discussed below. The same frequency on a format that earns 2 seconds per impression would deliver several times as much.

Lumen and similar companies train predictive models on eye-tracking panels and then apply them to campaign impressions using signals such as ad size, time in view and screen position. TVision, measures television differently, with a passive panel of 5,000 U.S. households fitted with computer vision sensors that record each second whether eyes are on the screen. Adelaide's AU score, a 0-to-100 placement rating, predicts attention quality but is not itself a seconds count, so it does not sum in the same way.

On the buy side, cumulative attention is used by agency planning and effectiveness teams. On the sell side, publishers and ad tech companies increasingly price inventory in additive units. InMobi's Attention Per Mille (APM), for instance, multiplies the share of impressions viewed by average view time and normalises to 1,000 impressions, which makes it a seconds count that can be added across a campaign.

Origin and evolution

The question cumulative attention tries to answer is older than digital advertising. Herbert Krugman's 1972 paper "Why Three Exposures May Be Enough", published in the Journal of Advertising Research, argued that a viewer's response develops across the first few exposures. In 1979 Michael Naples's book Effective Frequency, published by the Association of National Advertisers, turned that into the planning convention of "3+ frequency". In the 1990s Erwin Ephron's recency planning and John Philip Jones's single-exposure research pushed back, arguing that one well-timed exposure can be enough.

All of those debates counted exposures, an opportunity to see. Attention research replaced that unit with seconds. Dr Karen Nelson-Field, founder of Amplified Intelligence, published work showing that an ad needs roughly 2.5 seconds of active attention to have a lasting effect on brand memory, and that around 85% of paid ads fail to reach that level, according to Amplified. That finding was measured per ad, leaving open whether short glances could substitute for one long look.

Several studies then moved towards the cumulative view. Quantcast filed a patent in March 2016, granted in March 2022, that measures "viewable attention" as screen area multiplied by time, adds it up for each recipient and caps further exposures once enough has accumulated. A Teads and Lumen brand lift meta-analysis covering 16 campaigns from 14 advertisers, published in 2023, found that online ad recall lift was 40% higher at 5 seconds of attention than at 1 second, but measured single exposures only. Its authors suggested that "short dwell times can accumulate over time" and flagged frequency as unresearched.

The most cited evidence arrived on September 16, 2024, when Havas Media Network published "What do 9,000 brand lift studies teach us about attention and memory?" with Lumen and Brand Metrics. The analysis covered 9,089 display campaigns, about 1.88 million respondents and 5.6 million viewable impressions, mostly in the UK. Its third finding was that "results are driven by aggregate attention". The strongest results came from several impressions of at least one second each, summing to roughly 2 to 5 seconds; around 10 display exposures produced close to 7 seconds of aggregate attention. Awareness and consideration responded best to multiple "good enough" impressions, while preference and action intent could also be driven by one high-attention ad.

Why it matters for marketers

Cumulative attention changes how a buyer values cheap, low-attention inventory. If attention only counts per ad, a placement that averages half a second looks close to worthless. If attention accumulates, the same placement can contribute, provided the person sees it enough times. That is the argument Grace Kite of Analytic Partners made at Cannes in 2025, saying that "seeing lots of little ads is a perfectly good substitute for lack of attention on any one ad", according to Mi-3. Tom Roach of Jellyfish backed aggregating "lots of little hits".

At IAB Europe's Programmatic Day in July 2025, Lucy Wallace of Publicis Next cited Meta and Playground xyz research finding that multiple short bursts of interaction were 21% more effective than one long one.

The concept also links attention back to reach, the currency most plans are still built on. Nielsen and Adelaide combined attention scores with reach data in October 2025 so that placements could be compared on both at once. Cumulative attention extends that logic to frequency, setting a target in seconds per person rather than in impressions. It recasts reach and frequency planning as a question of how many seconds each person needs.

Limitations and disputes

The evidence base is narrow. The Havas study covered display advertising only, used no control group and relied on descriptive analysis rather than multivariate modelling, according to the paper itself. Its attention figures were modelled predictions, not observed eye movements of the respondents surveyed.

Nor is it settled that seconds are interchangeable: ten one-second glances and one ten-second view are not obviously equivalent. Spreading attention across many tiny impressions was not effective in that study, which complicates the simplest version of the argument.

The wider attention case is also contested. A Kantar analysis of 873 campaigns with more than $3.2 billion in media spend, published in July 2024, found that channel attention levels did not directly track brand contribution or cost-effectiveness. Meta's Peter Buckley wrote that "some advertisers are buying too much attention".

Standards are a further gap. The MRC and IAB attention measurement guidelines, finalised on November 12, 2025, require attention to build on MRC-compliant viewable impressions and state that attention should not be used as a measure of campaign outcomes. They do not define how attention should be summed across exposures, so two vendors' cumulative totals are not comparable. Person-level totals also need individual frequency data, which is harder to obtain as cookies and device identifiers erode.

Not the same as

Effective frequency counts exposures needed for an effect, typically three or more. Cumulative attention counts the seconds those exposures actually earned.

Attentive reach or attention-adjusted reach is the share of an audience that paid any attention, a breadth measure. Cumulative attention measures depth per person.

Cume, short for cumulative audience, is a radio and audio measure of the unduplicated people reached over a period. It counts listeners, not attention.

Cumulative attention in machine learning is unrelated. In speech synthesis models such as Google's Tacotron 2, described in a December 2017 paper, a "location-sensitive" attention mechanism feeds the cumulative attention weights from previous decoder steps back in as a feature, helping the model move steadily through the input text. The term there describes neural network arithmetic, not human attention.

Recent developments

Sellers have started packaging inventory in additive attention units. InMobi launched FRAME Attention on July 24, 2026 with an 8,000 APM floor across mobile in-app video and connected TV (CTV), with Lumen validating the inventory in testing commissioned by InMobi. Lumen announced a partnership with Cint on April 20, 2026 to match attention measurements against brand lift results.

Television measurement is moving the same way. Viant completed its $40 million acquisition of TVision on May 5, 2026, and TVision's Ad Scoreboard 2.0, launched on August 25, 2026, reported that viewers are attentive to only 9 of roughly 200 daily viewing minutes that carry ads. Buying tools still operate mostly per impression: Adelaide's AU became available for pre-bid targeting in Amazon DSP from June 10, 2026 and in DV360 custom bidding from mid-July 2026. Stellantis and IAS reported on June 4, 2026 on more than 13 billion impressions measured for attention. As of October 2026, none of these products reports a standard, audited cumulative attention figure per person.

Timeline

  • December 1972 - Herbert Krugman publishes "Why Three Exposures May Be Enough" in the Journal of Advertising Research.
  • 1979 - Michael Naples's Effective Frequency, published by the ANA, popularises 3+ frequency planning.
  • March 28, 2016 - Quantcast's priority date for a patent that accumulates "viewable attention" per recipient to control exposure.
  • October 18, 2019 - Lumen publishes "Attentive seconds: One currency to rule them all?", proposing attentive seconds as a cross-media unit.
  • 2022 - Amplified Intelligence's work with OMD sets out the 2.5-second attention-memory threshold.
  • 2023 - Teads and Lumen publish a brand lift meta-analysis suggesting short dwell times may accumulate across exposures.
  • January 4, 2023 - DoubleVerify receives MRC accreditation for Authentic Attention.
  • July 2024 - Kantar's 873-campaign analysis questions the link between channel attention and brand contribution.
  • September 16, 2024 - Havas Media Network, Lumen and Brand Metrics publish the 9,000-study analysis concluding that results are driven by aggregate attention.
  • June 2025 - Grace Kite argues at Cannes for targeting "lots of little bits of attention".
  • October 7, 2025 - Nielsen and Adelaide integrate attention metrics with reach data.
  • November 12, 2025 - MRC and IAB finalise attention measurement guidelines.
  • April 20, 2026 - Lumen and Cint announce an attention and brand lift partnership.
  • May 5, 2026 - Viant completes its acquisition of TVision.
  • July 24, 2026 - InMobi launches FRAME Attention with an 8,000 APM floor.
  • August 25, 2026 - TVision launches Ad Scoreboard 2.0.

Summary

Who: Attention measurement firms such as Lumen Research and Amplified Intelligence, agency groups including Havas Media Network, and advertisers and publishers that plan or sell media on attention.

What: Cumulative attention is the total seconds of attention one person gives a campaign's ads across all exposures, calculated by summing each impression's probability of being viewed multiplied by its expected viewing time.

When: It builds on effective frequency research from 1972 and 1979; the attention-based version gained its main evidence on September 16, 2024, with the Havas, Lumen and Brand Metrics study.

Where: It is applied in campaign planning and post-campaign analysis across display, video, CTV and television, mostly through predictive models trained on eye-tracking panels.

Why: Most individual digital ads earn too little attention to form a memory, so marketers want to know whether many brief exposures can add up to enough; the evidence suggests they can for awareness, while standards for measuring the total do not yet exist.