Marketing in 3 · Issue 22 · Week of 8 June 2026
Two things, and they pull against each other. Across 22,958 real thumbnails and two experiments, pictures with a moderate amount going on beat both the sparse and the crowded ones by a wide margin. Colour harmony then behaved the other way round: fully matched worked, completely clashing worked, and the compromise between them scored 3.41 out of 7, the worst result anywhere in the study. Aim for the middle on how much you put in the frame, and avoid the middle on how well it hangs together.
A thumbnail is the only part of a video most people ever see. Jiaming Fang, Qiqi Zheng, Miyan Liao, Lintong Han and Lu Liu, working across three Chinese universities, set out to find what separates the ones that get clicked from the ones that do not.
They scraped 22,958 videos from Bilibili, the Chinese platform closest to YouTube, across eight categories from cookery to science, and tracked each one daily for a month. Then they turned the thumbnails over to two trained models. One scored how much was going on in each picture: the number of objects, the colours, the edges, the clutter. The other scored how well the colours and textures in a picture fitted together.
Popularity combined views, comments, likes, shares and saves. To check that the correlations meant something, they then built nine thumbnails of their own, varying the same two qualities, and put them in front of 423 people browsing a mocked-up page.
The first rule is the one people half expect. More going on in a thumbnail earns more clicks, right up until it does not. The curve rises and then falls, and the top of it sits in the middle.
In the experiment, the middling version was clicked far more readily than either extreme, and the margin was wide. A nearly empty thumbnail gives the eye nothing to catch on. A packed one asks for more work than anyone browsing is willing to do, so they scroll on.
This holds in the real data too, across all 22,958 videos, and it holds whether a human chose the frame or the platform picked one automatically. That last point is worth sitting with. The same rule applies to the default frame your platform grabs, which is the thumbnail most small creators and most brands are quietly running with.
| Group | Value (out of 7) |
|---|---|
| Almost nothing in it | 4.11 |
| A middling amount | 5.48 |
| Packed | 4.49 |
Experiment with 423 people shown nine constructed thumbnails on a simulated browsing page.
The second rule is the surprise. Colour harmony does not simply help. Thumbnails where everything matched did well, and so did thumbnails where nothing matched at all. The ones that lost were the half-hearted middle.
Two different things are going on. A harmonious picture is easy and pleasant to take in, so it reads fast. A clashing one is jarring enough to be interesting, which stops the scroll for a different reason. A picture that is neither does neither job.
The authors tested whether the two rules interfere with each other and found that they do not. You can set the amount of detail and the degree of harmony separately, which makes this one of the rare design findings you can act on without a trade-off.
| Group | Value (out of 7) |
|---|---|
| Nothing matches | 4.89 |
| Somewhere in between | 3.41 |
| Everything matches | 5.3 |
The same experiment: the lowest score belongs to the middle, not to either extreme.
The effects were sharpest for two kinds of video. Useful ones, where the viewer wants to learn or decide something, mattered more than entertainment. So did typical ones, meaning videos that look like everything else in their category, where the thumbnail is the only thing doing the separating.
Viewer type mattered too. People who were already interested in the subject reacted more strongly to how busy the picture was. For them a crowded thumbnail is genuinely more work, because they are actually reading it rather than glancing past.
Interestingly, that pattern did not repeat for harmony. Whether a picture hangs together seems to register the same way for everyone, interested or not, which fits the idea that harmony is processed almost instantly while detail has to be worked through.
The study is about video, but the situation it describes is everywhere: a grid of competing images, a second of attention each, one of them has to win. Product listings on a marketplace, creative variants in an ad carousel, app icons on a crowded store page, the hero image at the top of an email.
Going past what was actually tested, the two rules translate neatly. Give the eye three or four things to look at rather than one or ten. Then decide whether your picture is going to be the calm one or the loud one in the grid, and commit, because the version that hedges is the version that gets scrolled past.
The caveat worth carrying is cultural. This is Chinese data, and interfaces there run busier than Western ones by habit. The shape of the curves should travel. The exact place the peak sits probably does not.
Marketing in 3 steps
A moderate amount of detail and a decisive choice on colour. In a browsing test, pictures holding a moderate amount were clicked a good deal more often than either sparse or crowded ones, and the same shape showed up across nearly 23,000 live videos (Fang et al., 2026, Journal of the Academy of Marketing Science).
Neither extreme. Too little and there is nothing to catch the eye. Too much and the viewer has to work, so they scroll past. Three or four clear elements is the practical version of the finding.
Either fully or not at all. Both extremes were clicked more than the compromise, which came last by a wide margin. Harmony reads as pleasant, clashing reads as interesting, and halfway reads as nothing at all.
The same rules applied to automatically selected frames as to ones creators chose, which means a default frame is usually an unmanaged risk rather than a neutral choice.
Practical, information-led videos, and videos that sit in a crowded category where everything looks alike. In both cases the picture is doing most of the work of standing out.
Journal of Marketing Research · 2026
Togawa, T., & Onzo, N.
Logos drawn with depth, through shading and perspective, raised purchase likelihood by reading as more real and so more trustworthy. The gain only appeared in visually simple surroundings and for physical products. Put the same logo in a busy layout or on something abstract and the advantage disappears.
Psychology & Marketing · 2026
Alhouti, S., Pierre, L., & Liu, X.
On a swapping platform, people with realistic profile pictures traded more, and the effect was strongest where the listing itself gave little away. A third study swapped in AI-made portraits and found they persuaded less once other information was plentiful. When buyers cannot check the facts, they read the face.
Fang, J., Zheng, Q., Liao, M., Han, L., & Liu, L. (2026). Thumbnail power: Visual cues and video popularity on UGV platforms. Journal of the Academy of Marketing Science, 54, 815-837. https://doi.org/10.1007/s11747-026-01152-6
This brief is written for practitioners from the published article. Numbers are as reported by the authors; read the article for full methods.
More issues on UGC · Creative · Social media
Posts carrying TikTok's AI label collected 7% to 8% fewer likes at the same reach. Not because people distrust AI. Because the label says the creator did not try.
Judges called 72.3% of the photos in positive Amazon reviews well composed, against 39.7% of those in negative ones. Nobody stages a complaint.
Handing customers a ready-made post lifted sharing from 0.4% to 2.1% in a real gym campaign, and quietly stopped unhappy customers posting at all. How to use it, and where it gets uncomfortable.
Bruno Schivinski is Associate Professor of Marketing at Gdańsk University of Technology. He studies how consumers engage with brands in digital and social media.
