Why social lowered tolerance for ads, not attention

Social did not lower attention, it lowered tolerance, and the difference changes how an account should be planned, briefed and read. The attention researchers are right about the average. Karen Nelson-Field's work at Amplified Intelligence, a company that sells attention measurement, puts the threshold for durable memory at about 2.5 seconds of active attention and finds more than 80% of digital ads never get there, and every account I have looked at confirms the shape of that finding. Most ads on a feed are gone in under two seconds. The mistake is in what the industry did with the number. It read a low mean and concluded the medium was a low-attention medium, then built a decade of media planning, creative adaptation and frequency theory on top of that conclusion, and the conclusion does not survive contact with the distribution.

Read the distribution, not the mean

Take every ad in one account under one media plan and lay out the share of impressions that turned into a watched view, fifteen seconds or completion on Meta's ThruPlay definition. The picture is never a bell curve around a low average. It is a long tail of ads that almost nobody stayed with and a smaller, separate tail of ads that a meaningful share of people watched to the end, on the same budget, the same audience settings, the same week. In the accounts we have measured, the gap between the least watched ad and the most watched under one media plan is a multiple, not a margin, and the brand's existing creative usually sits in the long tail with the ads built for the feed running beside it in the short one. Average those together and you get a low number that describes none of them.

Meta's own attention deck concedes the point about its own reporting, noting that the average view time per impression in Ads Manager does not represent how much of the content a person actually watched, because the distribution across people and exposures is what matters. That is a platform, which sells the media, saying its headline mean hides the thing you need to see. A distribution with a fat left tail and a thin right tail reads as low attention only if you refuse to look at the right tail. The right tail is what the medium is capable of. The left tail is what most advertising does with it.

Nobody left the room

The other half of the low-attention story is that people are supposedly not paying attention to their phones, and this is simply untrue. GWI's 2024 data, as carried in Meta's attention material, has European adults spending twice the weekly hours on short form video that they spend on television, and Gen Z five times. People are on these platforms for hours a day and they are watching intently, which is why the platforms are worth what they are worth. What changed is not the quantity of attention available. What changed is who controls it.

Television forced the watch. The ad break was captive enough that planners could treat exposure and attention as the same thing, and for decades that was a safe assumption. Skippable formats handed control to the viewer, and the viewer applies a test the captive audience never got to apply. Is this worth more of my attention than the next thing, in a feed already tuned to what I like? The medium did not lower attention, it lowered tolerance. Two seconds is not the attention span of a person on Instagram, it is the time they give an ad to prove it deserves more, and most ads fail the test because they were never built to take it.

Nelson-Field's own data, read this way, becomes evidence for the argument rather than against it. Her Meta-commissioned Australian work found that Meta placements drive active attention, eyes on the ad, at higher rates than passive attention formats, because the phone is held close with little distraction. The attention is there. She also argues that each platform has its own in-built ceiling on paid ad attention, and the ceiling is real, but the right tail of every account sits a long way above where the mean would put it, so whatever the ceiling is, most advertising stops well short of it. Her numbers describe what happens when television-shaped advertising meets a viewer who can leave, not what the medium is capable of when the advertising was built to be chosen.

What the two seconds do

The two seconds are not nothing, and it matters to be precise about what they are. A viewer who sees the brand's distinctive assets in the first second and swipes has had a memory refreshed, if the memory was there to refresh. That is the job Byron Sharp says most advertising does most of the time, keeping existing associations fresh and linked to the moments people buy in, and a branded first second does it. A skip can remind. What a skip cannot do is build, because an association that is not in a person's memory cannot be refreshed into existence however many times the brand's assets pass by, and the buyers a brand grows on are the light and non-buyers who have the least to refresh. So the tolerance test is not a nuisance between the ad and its audience. It is the gate between the two jobs the ad has to do, and the share of people who pass through it decides whether the brand is being maintained or grown.

The test is relevance

So what makes a viewer stay past second two? Not identity data. The decade-long pursuit of the segment of one, dynamic creative assembling itself per person, the right message to the right person at the right time, aimed at the wrong variable. It assumed relevance is a property of the person, and relevance is mostly a property of the situation. The same woman is a commuter at seven, a manager at eleven, a parent at six and exhausted at ten, and what earns her attention at any of those hours is not that the ad knows her name, it is that the ad meets the moment she is in recognisably.

Jenni Romaniuk's work on category entry points gives this a briefable form. The situations, occasions and feelings that bring a person into a category, the long shift ending, the guests arriving in an hour, the three o'clock slump, are the cues buying memory hangs on, and an ad built on one of them is relevant to everyone in that situation simultaneously with no data at all. The viewer does not need to be known. They need to see their moment, their kind of person and their kind of environment in the frame and conclude, in under two seconds, this is for me.

Ipsos's US database of 3,122 ads, from a vendor that sells creative testing, found that ads taking an unconventional approach to their category held viewers 40% longer on skippable formats. Dentsu's attention research, from a media group that trades on attention metrics, concludes creative is the number one driver of attention across platforms. Different interests, same direction. Viewing time on a skippable medium is a choice the viewer makes about the creative, and the creative can move it.

Relevance is not entertainment

There is a way this argument goes wrong, and it is worth heading off. If the viewer's test is relevance, the tempting conclusion is that the ad's job is to be entertaining enough to survive, and that leads straight to work the feed loves and the brand cannot use. Relevance does not collapse into entertainment. The distinctive assets stay in frame from the first second, the brand arrives early, the message has to land, and everything Sharp and Romaniuk established about distinctive assets, penetration and mental availability still applies, which is the argument of How to apply brand science to social media. Attention without branding is entertainment, branding without attention is invisible, and the relevance bar is the entry fee, not the outcome.

No single ad clears it

The consequence follows directly. If the bar is "this is for me", judged in two seconds by people in different situations with different lives, then no single execution can clear it for everyone a brand needs to reach. The ad that reads as "for me" to the commuter reads as background to the parent. The ad that stops the twenty-five-year-old renter slides past the fifty-year-old homeowner. This is not a failure of the ad, it is the arithmetic of relevance, and it is the premise of Why you need a creative portfolio, not more content. A brand that needs to be chosen across many situations needs many ads, each built to be chosen by the people in one of them, consistent in its assets and its idea, varied in its execution.

It also changes what you measure, because on a medium where continuing to watch is a choice, the share of impressions that became watched views is a statement about active attention that television never had, an argument carried in How to measure attention on social with data you already have and the standard in How to read paid social results.

What to do with this

The practical move is to stop reading the average and start reading the spread. Pull every ad from the last quarter, rank them by watched rate on one platform and one threshold, and look at the gap between the top and the bottom. The gap is the medium's capacity. The bottom is the tolerance test being failed, and it usually fails on the same things, an opening built for a captive viewer, production that reads as ad before the idea arrives, a situation nobody in the audience recognises as theirs. The top tells you what relevance looked like for the people who stayed, and it is the brief for the next round.

Low attention was a story the industry told itself because it explained the average and excused the advertising. The viewer has been paying attention the whole time. Media creates the opportunity to see your ads. Creative determines whether they work. What the two seconds and the twenty seconds each do for the brand, and why the ad has to be built for both, is the subject of Why every social ad has two jobs.

Questions

If most people skip in two seconds, isn't social a low-attention channel?

The average is low because most ads are built for a captive viewer and meet one who can leave. Lay out every ad in an account under one media plan and the shape is a long tail of ads almost nobody stayed with and a shorter tail of ads a meaningful share of people watched to the end, on the same budget and the same audience. The mean describes neither tail. The right tail is the channel's capacity, and it is reached by work built to be chosen, not by more spend.

Is a skipped impression wasted money?

Not entirely. If the brand's distinctive assets are on screen in the first second, a skip can refresh a memory that already exists, which is the job most advertising does most of the time. What it cannot do is build an association that was not there, and the buyers a brand grows on have the least to refresh. So a skipped impression maintains, a watched one can grow, and the share of impressions that become watched views is the number that says which of the two the account is mostly doing.

We already personalise with dynamic creative, isn't that relevance?

Personalisation treats relevance as a property of the person, and most of what earns attention in a feed is a property of the situation. The same person is in a dozen different situations a week and an ad that recognisably meets one of them is relevant to everyone in it, with no data at all. Category entry points, the situations, occasions and feelings that bring people into a category, are the briefable form of this, and a portfolio built across them reaches more people than a message assembled around a profile.

Sources

• Amplified Intelligence, "Why does the attention-memory threshold matter?", 27 June 2022. The 2.5-second attention-memory threshold and the finding that over 80% of ads do not reach it. Amplified sells attention measurement.
• Karen Nelson-Field, Amplified Intelligence, Facebook mobile study, Australia 2021, N=1,160, as presented in Meta, "Demystifying attention", 2025, and her platform-ceiling argument as presented by Tom Roach (Jellyfish), "The 3rd age of effectiveness", circa 2023. Active, passive and non-attention, Meta placements driving active attention. Commissioned by Meta, which sells the media; Jellyfish is an agency that buys platform media.
• Meta, "Demystifying attention", 2025. Meta's statement that average view time per impression does not represent how much people actually watched, and the GWI 2024 to Q1 2025 hours data for short form video against television in Europe. Platform sales material, Meta sells the media.
• Dentsu, "Unlocking the Currency of Attention", 2021, as cited in Meta, "Demystifying attention", 2025. Creative as the number one driver of attention. Dentsu trades media on attention metrics.
• Ipsos, Rachel Rodgers, "How to stop wasting money on 'efficient' digital advertising", 11 February 2022. Unconventional ads hold viewers 40% longer on skippable formats, 3,122 ads, 2016 to 2021. Ipsos sells creative testing.
• Jenni Romaniuk, Building Distinctive Brand Assets (2018) and Better Brand Health (2023). Category entry points as the cues buying memory hangs on. Ehrenberg-Bass Institute, sponsored by large advertisers.
• Byron Sharp, How Brands Grow (2010). Penetration, mental availability, distinctive assets, advertising as memory refreshment. Ehrenberg-Bass Institute, sponsored by large advertisers.
• One measured account, anonymised. The shape of the watched rate distribution by ad under one budget and one media plan, from the account's own platform reporting.