Why every social ad has two jobs

Every social ad has two jobs, reminding the people who already know the brand and building a memory in the people who do not, and on a skippable medium the viewer decides which of the two it gets to do. Advertising has always done both and for most of its history never had to separate them. A thirty-second spot in a captive ad break refreshed the memory of people who already knew the brand and, in the same thirty seconds, built a memory in people who did not, and because everyone in the room saw all thirty seconds, nobody planned for the two jobs separately or measured them apart. Skippable formats took that away, and the consequence for how a social ad is built and read is larger than the industry has admitted.

Reminding

The first job is refreshing a memory that already exists. A buyer who knows the brand, has bought it, has the associations sitting in memory, needs those associations kept fresh and kept linked to the moments they buy in. This is the job the brand science says most advertising is doing most of the time, and it works on a glance. The distinctive assets in frame for a second or two, the colour, the pack, the character, the sound, and the existing structure gets a small refresh. Cumulative brief exposure does this well, which is where Byron Sharp's argument about advertising as memory refreshment and Karen Nelson-Field's data on brief attention are both right. Amplified Intelligence's work, from a company that sells attention measurement, finds durable new impact needs more than 2.5 seconds of active attention, but the same research tradition has always accepted that a glance at a well-branded ad maintains what is already there. A skip can remind.

The platform evidence points the same way. Meta's Realeyes lab study of 16,835 video exposures, commissioned by Meta so read with that interest in mind, found feed placements produced 2.4 times the brand uplift per attentive second of in-stream placements, which is another way of saying short exposures to ads people already half-know are efficient at the reminding job. Meta's 2024 analysis with Playground XYZ, using modelled rather than measured attention, put 59% of offline purchase conversions inside the first six seconds of attention in reach campaigns, and 46% of online conversions inside the first two. Those buyers were not learning anything. They were being reminded of something they already knew, and a couple of seconds was enough. The same deck finds the reminding compounds, with people who accumulated five or fewer attentive seconds across a week converting at rates 13% to 21% higher than people who got a single impression of the same brief attention, again on modelled attention and again from the platform that sells the media.

Building

The second job is creating a memory that was not there. A new association, a product the buyer did not know existed, a category entry point the brand has never been linked to in that buyer's head, the situation on the walk home or the guests arriving in an hour. This is the job that grows a brand, because the buyers a brand grows on are the light and non-buyers Sharp's own penetration arithmetic points at, and for them there is nothing to refresh. Building needs encoding, and encoding needs the viewer to stay past the point where a glance stops and a message starts. A skip cannot build.

The Havas, Lumen and Brand Metrics analysis of 9,089 campaigns, from three companies that sell attention and lift measurement, found the strongest brand outcomes came not from one long view but from several exposures of a second or more adding up to two to five attentive seconds in aggregate. That is the reminding job compounding. It says nothing against the building job, because a memory that does not exist cannot be refreshed into existence however many times the assets flash past, and the Advertising Research Foundation's 2026 validation study made the surrounding point, that attention translates into lift only when media weight, reach and frequency are present, and should not be read as a stand-alone proxy for outcomes. Attention is the condition under which building can happen. It is not the building.

Why nobody planned for two jobs

Broadcast measurement could not separate the two because broadcast exposure did not vary. Everyone got the full spot, so the ad that reminded was the ad that built, and one measure of recall or lift covered both. Les Binet and Peter Field's work in the IPA Databank, agency-submitted award cases so read as such, separated the long and the short of it at the level of campaigns, roughly 60% brand building to 40% activation, but that split is between campaigns with different jobs, not between two jobs inside one ad.

Skippable formats force the separation because the same ad now gets two seconds from one viewer and twenty from another. The two-second viewer has received the reminding job and nothing else. The twenty-second viewer has received both. This is not two audiences you can target separately, it is the same audience deciding in real time how much of your ad to take, and the distribution is in every account's reporting, a long tail of brief views and a smaller tail of full ones, the shape described in Why social lowered tolerance for ads, not attention.

Build every ad to do both

The structural implication is that every social ad is built to do both jobs, in sequence, in the order the viewer decides. Brand and distinctive assets in the opening second, so the viewer who leaves at second two has at least been reminded. Message in the body, so the viewer who stays learns the thing the ad exists to teach. And the hook, the thing that converts one viewer into the other, sits at the join. A hook is not an attention trick, it is the mechanism by which a reminded viewer becomes a viewer who can be built.

This reverses the narrative structure most advertising was trained on. The classic arc, setup, build, reveal, held the brand back for the payoff because the captive viewer was going to be there for it. On a skippable medium the reveal plays to the people who stayed, and the people who left carry nothing new away because the assets were saved for the end, so the ad did neither job for them. What high quality creative means on paid social takes that argument into production.

What each number reads

The measurement implication sits inside the standard in How to read paid social results, and it is worth being precise about what each number says and does not say. Stop rate, the share of impressions that became a three-second play on Meta, tells you the reminding job was done and the building job was made possible, and TikTok's two-second view is the nearest native unit but a lower bar, so the two are never blended. The viewer saw the assets and did not leave immediately. Watched rate, the share of impressions that became a ThruPlay on Meta, fifteen seconds or completion, or on TikTok a focused view, six seconds or completion, read separately per platform, tells you the building job was made probable. The viewer stayed long enough for a message to encode. Neither says the building happened. Only a lift study, exposed against holdout, says that a memory was created that was not there before, which is why effectiveness is read first and performance metrics are read as the leading indicators of it, never as substitutes.

Hold rate says whether the body of the ad kept the people the opening earned, cost per watched impression says what the attention cost, and the platforms report a good deal more, all of which has a place. The point is which job each metric is reading.

KitKat

The worked example is a brand everyone already knew. KitKat has one of the great category entry points in confectionery, the break, and decades of reminding behind it. A portfolio built for the feed, real people in real break moments, distinctive assets in frame from the first second, message in the body, delivered two times the purchase intent of the brand's existing creative in brand lift testing, same brief, same media. On a brand that well known, the reminding job was already being done by every exposure, so the gain came from the building job, from viewers who stayed long enough to attach the brand to a break they recognised as theirs, the shift ending, the summit pause, the three o'clock desk, and the lift study is the only reason we can say that with confidence rather than infer it from a watched rate.

Sharp's position, and where the medium changes it

The strongest version of the counter-argument is Sharp's, and it deserves stating fairly. Advertising's main effect is to refresh and reinforce memory structures, most exposure is low attention and that is fine, distinctive assets do the work, and the industry's obsession with engagement and message comprehension misreads how advertising works. Most of that is right and this essay depends on it. We agree on the goal, penetration over loyalty, reach over targeting, continuity over bursts, distinctive assets doing the heavy lifting, and on the scepticism of engagement counts that measure a minority of fans.

Where the medium changes the mechanics is the buyer who has no structure to refresh, which for a growing brand is most of the market, and a format that lets that buyer leave before anything is encoded. On television the refresh and the build came bundled in the same thirty seconds whether anyone planned for it or not, so a theory of advertising as refreshment could carry the building job along silently. On a skip medium an unwatched impression refreshes what is already there and cannot add what is not, so refreshment is the job for the buyers you have, building is the job for the buyers you need, and the ad has to be built for both because the viewer, not the planner, decides which one it does. That is not a departure from the brand science, it is what the brand science requires once the viewer can leave. Media creates the opportunity to see your ads. Creative determines whether they work.

What to do with this

Take the three ads with the most spend behind them last quarter and watch each with the sound off. Pause at second one. If the brand is not identifiable from the frame, the ad is not doing the reminding job for the majority who skip, and that is fixed in the edit, not the media plan. Then watch the body and write down the one thing a viewer who stayed would now know or feel about the brand that they did not before, and if there is no answer the ad is not built for the building job either. Read stop rate and watched rate for each ad separately, on one platform and one threshold, and treat the gap between them as the hook's work. Then check whether a lift study is in the plan, because without one the account can only ever say the building was made probable, never that it happened. Why the second job is worth paying for, in media terms, is the subject of Why creative is your largest working media lever.

Questions

If a glance at a well-branded ad is enough, why do we need people to watch?

A glance is enough for the buyers who already hold the brand in memory, because a branded first second refreshes what is there. It is not enough for the light and non-buyers a brand grows on, because there is nothing in their memory to refresh and a new association needs the viewer to stay long enough for it to encode. Both are real jobs, and the share of viewers who stay is what decides whether an account is maintaining the brand or growing it.

Won't putting the brand in the first second make people skip?

People skip ads that read as ads before the idea arrives, and that is a production problem, not a branding problem. Distinctive assets can be in frame from the first second through the product in hand, the colour in the room, the pack on the bench, the character or the sound, without the ad opening on a logo card. The reminding job needs the brand present early, the building job needs the opening to earn the next second, and the ads that do both are the ones built for the feed rather than cut down for it.

How do we know whether the building job actually happened?

Only a lift study says it did, by polling an exposed group against a holdout on awareness, association or purchase intent. Stop rate says the reminding job was done, watched rate says the building was made probable, and both are read weekly as leading indicators. Effectiveness is read first and on its own clock, and a watched rate is never reported as if it were a lift.

Sources

• Amplified Intelligence, "Why does the attention-memory threshold matter?", 27 June 2022. The 2.5-second threshold for enduring impact. Amplified sells attention measurement.
• Realeyes lab study for Meta, 16,835 video ad exposures, 6,076 US mobile users, December 2022, as presented in Meta, "Demystifying attention", 2025. Feed placements at 2.4 times the uplift per attentive second of in-stream. Commissioned by Meta, which sells the media.
• Meta internal analysis with Playground XYZ modelled attention, 6,003 video ads, 1.33 billion impressions, March to May 2024, and 0.75 billion user-adset pairs on cumulative attentive seconds, as presented in Meta, "Demystifying attention", 2025. Share of offline and online conversions within the first six and two seconds of attention, and higher conversion among people with repeated brief exposure. Modelled attention, platform interest.
• Havas Media Network, Lumen Research and Brand Metrics, "What Do 9,000 Brand Lift Studies Teach Us About Attention and Memory?", September 2024. Multiple exposures adding to two to five attentive seconds in aggregate perform best. All three sell attention or lift measurement.
• Advertising Research Foundation, Attention Measurement Validation Initiative Phase 3, press release, 9 June 2026. Attention should be read alongside media weight, reach and frequency, not as a stand-alone proxy. Member-funded industry body.
• Les Binet and Peter Field, The Long and the Short of It, IPA, 2013. Long-term brand building and short-term activation as distinct effects, roughly 60/40. IPA Databank of agency-submitted cases.
• Byron Sharp, How Brands Grow, 2010. Advertising as memory refreshment, penetration, light buyers. Ehrenberg-Bass Institute, sponsored by large advertisers.
• KitKat, brand lift testing, purchase intent at two times the brand's existing creative, same brief, same media. Cleared proof point from our own work.