What high quality creative means on paid social

A brand team that says its creative is already good is usually right by the standard it was trained on. Well shot, well graded, a considered edit, a score that swells in the right place, that is production value, and for most of advertising's history nobody needed to separate it from quality. On paid social they have come apart. Quality is whatever gets encoded. An ad is high quality if it built or refreshed a memory in the person it reached, and everything else about it is an input that either helped or did not. Quality is a test, not a style.

Argued backward from the memory

Effectiveness, in the order set out in How to read paid social results, is a memory that exists after the ad that did not before, read in a lift study. Work backward and the conditions fall out in sequence. For a memory to encode, the message has to be held long enough to land. For it to be held, attention has to be earned in the first seconds, because the viewer can leave. For attention to count, the ad has to clear a floor, present as the brand's, legal, built for the frame. The floor is the checkable part, and a pretest before media attaches checks exactly that. What pretesting can tell you, and what only the feed can sets out where that check stops. Every gate past it is read in the feed, weekly, against the brand's own previous work.

Attention earned

The evidence on what earns the first seconds is consistent across sources with different interests. Ipsos, which sells creative testing, found across 3,122 ads that work taking an unconventional approach held viewers 40% longer on skippable formats. Kantar's LIFT ROI studies across 56,984 creatives, commissioned by Meta and modelled rather than measured, found ads with a single human making eye contact indexed 81% higher on effectiveness than ads without, visual dynamism 74% higher, a distinctive atmosphere 67% higher. TikTok's creative guidance puts the window at six seconds regardless of format. A person, in motion, in a world that is recognisably somebody's, doing something the viewer has not seen a thousand times. That is what social-first means. A 9:16 crop of a television commercial is a television commercial that fits the frame.

Message held

Attention earned is not attention kept, and the structure decides the second. The television form, setup, build, reveal, the brand arriving at second twenty-six like a curtain call, was airtight for a captive viewer. On a skippable medium the setup is a wager that a stranger will donate several seconds to an unidentified advertiser on faith, and the skip lands mid-setup. YouTube's Creative Works analysis found traditional story arcs twice as likely to be bottom performers on YouTube as emerging digital arcs, and TikTok Marketing Science's 2025 work with System1, platform funded, found early branding produced 57% more happiness and 19% less attention decay. The structure that survives is hook first, brand early, payoff in small frequent instalments, the claim shown rather than stated, because nobody deliberates down a feed. Why every social ad has two jobs sets out why the opening reminds and the body builds.

Casting is a reach decision

The casting gate, the shortlist filtered with "not on-brand", is where production value gets defended without anyone meaning to, and it made sense when one face carried one hero ad for a year. On social the portfolio carries the brand and the face carries the segment, because people stop for people they recognise as their own and the delivery system reads the stopping, so a brand that only approves aspirational faces has filtered its actual buyers out of delivery. In personal care a result on flawless skin is casting and a result on skin that looks like mine is evidence. Vaseline's portfolio, much of whose casting the old gate would have declined, delivered three times the brand lift of the brand's existing creative on the same media. Keep the codes governed. Let the faces do their job.

Production value as proxy

The counter-position is the industry's default and it is not foolish. Production value is visible and controllable, and for seventy years on television it correlated with results, because a crew, a grade and a score were how you earned attention from a captive audience. It fails as a proxy because it is an input, and the medium now tests inputs against a viewer who can leave. Meta's own Reels split tests found 9:16 video with sound in the safe zone delivered a 34% lower cost per result than a still image, production value in the medium's terms rather than the studio's. The test that matters is whether the person reached remembers the brand and the message and links them to a moment they will buy in. Quality is the name for creative that did. Media creates the opportunity to see your ads. Creative determines whether they work.

What to do with this

Pull stop rate, hold rate and watched rate for every ad that ran last quarter, on one platform with one threshold, and rank them. The spread between best and worst is the quality conversation in the brand's own numbers. Look at what the top few have in common, who is on screen, how early the brand arrives, whether the claim is shown or stated, and brief the next round from that. Hold the floor as a written standard every asset is scored against before media attaches, and cast across the segments the brand needs to reach, not only the faces that match the brand book. What the pre-launch gate can and cannot tell you is the subject of What pretesting can tell you, and what only the feed can.

Questions

Is production value wasted on paid social?

No, it is tested rather than assumed. Light, sound, edit and casting all still matter, but the feed judges them by whether a person who could leave chose to stay and whether the brand was on screen while they did. A large production budget that arrives as a 9:16 crop of a television commercial is paying for inputs the medium does not reward, while a modestly produced ad with a real person, an early brand and an opening built for the first second often beats it on the brand's own stop and watched rates.

What does social-first actually mean?

Social-first is whether the message survives contact with a feed, and it is not a size or an aspect ratio. It means the opening does its work in the first second or two, the brand is recognisable before the skip can land, the message is legible with the sound off, and the people and situations on screen are ones the viewer recognises as their own. Evidence from Ipsos, Kantar and the platforms points the same way, a person in motion in a recognisable world, doing something the viewer has not seen before.

How do we know whether our creative is good before we spend?

Partly. A pre-flight check scores the floor before media attaches, brand codes present early, claims inside their substantiated bounds, platform grammar respected, and a forced-exposure pretest predicts what the ad will do with the attention it is given. What no pretest can tell you is whether the ad earns that attention in a feed where the viewer can leave, so the answer arrives in two stages, the floor before launch and the stop, hold and watched rates in the first week after it.

Sources

• 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. Ipsos sells creative testing.
• Kantar LIFT ROI studies, US, Canada, Germany, France and Thailand, 2021 to 2023, 56,984 creatives, 1,295 campaigns, as presented in Meta, "Demystifying attention", 2025. Effectiveness indices for single human with eye contact, visual dynamism, distinctive atmosphere. Commissioned by Meta, which sells the media, modelled.
• Meta, "Creative and Media Excellence", circa 2023. Reels split tests, 9:16 video with sound in the safe zone against a still image. Platform test data, Meta sells the media.
• YouTube Creative Works, traditional story arcs against emerging digital arcs, as cited by Tom Roach, "The 3rd age of effectiveness", circa 2023. Platform funded.
• TikTok, "Creative Guide: Driving Brand Equity", 2020. The first six seconds as vital regardless of format. Platform guidance, TikTok sells the media.
• TikTok Marketing Science with System1, "Creativity pays dividends", 2025. Early branding, happiness and attention decay. Platform funded, System1 sells the test.
• Vaseline, brand lift testing, three times the brand lift of the brand's existing creative on the same media. Cleared proof point from our own work.