Marketing in 3 · Issue 40 · Week of 12 October 2026
Put your own spin on it. Across 85,000 TikTok dance-trend videos, those that differed more from the rest of the trend earned more likes per view, even for the same creator. Experiments confirmed that the difference was the cause, and a test with 38 real creators agreed. The edge appears once viewers have seen the trend enough to spot the twist, and it came from deviations later in the video.
Following a trend is one of the most common ways brands and creators try to get seen on short-video platforms. The question is how to do it. Copy the dance move for move and ride the wave, or add your own twist? Content creators asked by the authors were split. About half said to stay close to the trend, and about a quarter said to deviate.
The team, from Swiss universities and Wharton, needed a way to measure how different one video is from others in a trend. They built a video-analysis tool, MUVID, that turns each video, with its sound and movement, into a set of numbers. A video's atypicality is then how far those numbers sit from earlier videos in the same trend.
They applied it to more than 85,000 TikTok dance videos grouped into around 1,500 trends, then ran three experiments, one of them a field test with real creators, to check whether the pattern was cause and not coincidence.
In the large dataset, the more a video differed from its trend, the more likes it earned per view. Videos more than one standard deviation more atypical than average earned about 112 likes per 1,000 views. Average ones earned about 101, and the most typical earned about 92.
The link held when the authors compared a creator with themselves, which rules out simply being a talented or popular creator. It also held after allowing for video and dance quality, and across other checks such as matching typical and atypical videos on their traits.
The effect per video is small. One standard deviation of extra atypicality went with about 2.5% higher odds that a viewer liked the video. The authors point out that on a platform with huge volume it adds up. For a creator with 15,000 views per video, shifting from a typical to an atypical video meant roughly 69 extra likes.
| Group | Value () |
|---|---|
| Least atypical | 91.5 |
| Average | 100.8 |
| Most atypical | 112.4 |
Study 1, model-free comparison. Least and most atypical are more than 1 SD below and above the average within a trend.
You can only see a difference if you know what the normal version looks like. The authors predicted that atypicality would pay off more when viewers had seen lots of videos from the trend. In the data it did. The more videos a trend had already collected, the stronger the link with engagement.
An experiment tested this directly. Participants saw five videos from a trend, or five unrelated videos, and then a focal video that either copied the dance or used a different one. After the trend videos, 34% chose to engage with the atypical version against 18% for the typical one. With no trend exposure, the two were level at about 17%.
A follow-up asked why. Viewers did not tire of the standard version. They simply judged the twisted version as more unique, and its creator as more creative.
| Group | Value (%) |
|---|---|
| Typical, trend not seen | 17 |
| Atypical, trend not seen | 17 |
| Typical, trend seen | 18 |
| Atypical, trend seen | 34 |
Study 3, 494 online participants. Without trend exposure 17.4% versus 16.8%; with exposure 17.6% versus 34.2%.
The authors split each video into its first and last five seconds. Atypicality in the last part was linked to more engagement. Atypicality in the opening seconds was not.
Their reading is that viewers first need to recognise the trend, and a surprise only lands after the pattern has been established. The authors describe this as suggestive and worth a dedicated experiment.
In the first experiment, 201 online participants watched five faithful trend videos and one focal video. Among those who saw an atypical version of the focal video, 24% chose to engage, against about 13% for the typical version.
The final test went to the platform itself. Thirty-eight creators each made two videos for two trends, one in the typical style and one with a twist, posting them about three days apart. With creator effects held constant, the atypical videos had about 10% higher odds of a like per view. This was the weakest result in the paper, and it came from small videos, averaging 783 views.
Marketing in 3 steps
Not in this evidence. Videos that put their own spin on a trend earned more likes per view than close copies, and experiments suggest the twist caused the difference (Bravin, Clegg, Hofstetter, Pouly & Berger, 2026, Journal of Marketing).
The paper found the benefit even for moderate deviations. The twist needs to be noticeable to viewers who know the trend, and the authors did not test how far is too far.
Later. Deviations in the final seconds were linked to more engagement and deviations in the opening seconds were not, though the authors call this suggestive.
Less well. The advantage grew with the number of videos in a trend, and in an experiment it disappeared when viewers had not seen the trend first.
An open-source video analysis tool the authors built. It turns videos into numbers that capture picture, movement and sound, so you can measure how similar two videos are. A web app lets you upload videos and see how typical each one is.
International Journal of Research in Marketing · 2026
Adam & Yang
Using machine learning on 60,669 apparel images and 4,087 car images, plus an experiment, this study finds that visually atypical brand images spread further for low-heritage brands because they entertain, and spread less for high-heritage brands because they clash with expectations. Check your brand's age before you break the mould.
Journal of Advertising Research · 2026
Walter, Stocchi, Ward, & Romaniuk
Two online experiments on in-feed ads find that image-based creative earns higher ad recall than text-based creative, and that text-based creative lowers correct brand recall. Pairing creative type against the type of brand element, for example text creative with a visual brand asset, recovers some of the loss, especially among non-buyers and for larger brands. Lead with an image.
Bravin, M., Clegg, M., Hofstetter, R., Pouly, M., & Berger, J. (2026). How Closely Should You Follow a Trend? Atypicality and Engagement on Social Media. Journal of Marketing, 90(5), 187-204. https://doi.org/10.1177/00222429261466668
This brief is written for practitioners from the published article. Open access (CC BY-NC 4.0): free to read at the publisher. Numbers are as reported by the authors; read the article for full methods.
More issues on Social media · Creative
Thumbnails with a middling amount going on won the clicks at 5.48 out of 7. Middling colour harmony lost them at 3.41. Two dials, pulling opposite ways.
Telling viewers a video was live made them more likely to keep watching (5.19 vs 3.59 out of 7) and to follow the creator. When going live pays off.
A meta-analysis of 210 experiments finds virtual influencers win on novelty, lose on credibility, and that how human they look decides which one matters.
Bruno Schivinski is Associate Professor of Marketing at Gdańsk University of Technology. He studies how consumers engage with brands in digital and social media.
If this issue saved you a meeting or sharpened a campaign, buy me a coffee. It keeps Marketing in 3 free, weekly and ad-free.
