Creators who show their face early sell at above-average rates on TikTok Shop

1. Introduction

Brands and creators on TikTok Shop must decide whether a video should feature a person talking to the camera or rely on product footage and voice-over, and how quickly the face should appear. The choice affects casting, production cost and which creators a brand works with. We ask how the presence and timing of the creator's face are associated with sales per view among top-selling videos.

Prior work on short-video popularity finds that attention is long-tailed and that early engagement predicts later popularity [1]. Recent benchmark work suggests that creator-related features carry much of the predictive signal [2], and that multimodal content features add to it [3]. Evidence on sales rather than popularity is thinner, and the returns to advertising-like content are hard to measure from observational data [4]. We therefore report associations and flag where they may reflect other factors.

2. Data

Our analysis covers TikTok Shop from 2 September to 1 October 2026. In that window we tracked 13,304 videos across 29 categories, and our AI analyst watched and analysed 3,342 of them in depth. The analysed count is the number of videos reviewed, not a count of views.

For each analysed video we recorded whether the creator appears on camera and, if so, whether the face is in the first second or later, plus the video's format. Shares of sales refer to the watched videos in this analysis.

Daily rankings of top-selling TikTok Shop videos in the US were collected from several independent sources and merged into one record per video. Where sources overlap, their sales estimates are cross-checked against each other, and videos whose estimates disagree by more than half are flagged. An AI analyst watched each video and recorded its opening, format, angle, production style, use of AI, hook source, length, pacing and what appears in the first second, without seeing how the video sold. Sales per view is a video group’s revenue per view divided by the market’s, shrunk toward the average for small groups, with 90% ranges. This paper covers 2026-09-02 to 2026-10-01: 13,304 videos tracked, 3,342 of them watched and analysed. Predictions come from a logistic model retrained daily and tested on recent weeks it had not seen.

3. Methods

Sales per view compares a group's revenue per view with the market's, so 1 means the market average. Estimates for small groups are shrunk toward the average by empirical Bayes [5], and we report 90% intervals on the log scale. We describe a group as clearly above or below average only when its range excludes 1.

The predictive model is a logistic model of top-selling videos, retrained daily and tested on recent weeks it had not seen (a time split). We report AUC as a ranking measure [6] with a bootstrap range [7]. It was trained on 2,667 rows.

4. Results: creator on camera

Videos with the creator's face in the first second sell at 1.06× the market average (90% range 1.02–1.10) and account for 57.6% of watched-video sales, from 1,560 videos. Videos where the face appears later sell at 1.03× (0.95–1.11) and account for 14.6% of sales, from 395 videos; this range includes 1, so we cannot separate it from average.

Videos where the creator is not on camera sell at 0.89× (0.84–0.93) and account for 27.8% of sales, from 944 videos. Their range sits entirely below 1 and does not overlap the range for an early face. The gap is modest in size: even the best group is only about 6% above average. Confidence is high for all three groups.

5. Results: format

Formats differ more than on-camera status does. Unboxing sells at 1.37× (1.20–1.57), voice-over B-roll at 1.29× (1.18–1.41), review or testimonial at 1.27× (1.17–1.37) and talking head at 1.18× (1.09–1.28). Demo or tutorial is the largest format, at 33.8% of sales, but sells at 0.92× (0.88–0.96). Skits sell at 0.86× (0.77–0.97).

Several formats rest on little data. Green screen (20 videos), before and after (28), live clips (21) and day in the life (8, low confidence) have ranges that include 1 and should not be read as findings. Our evidence does not cross-tabulate format with on-camera status, so we cannot say which formats make up the not-on-camera group.

6. Discussion

For brands and creators, the evidence suggests that showing a face early is associated with selling slightly above average, and that having no creator on camera is associated with selling below average. The effect of a later face is unclear. Format appears to matter more: unboxing, review and talking-head styles sell above average, while demos and skits sell below it. This fits prior findings that creator-related signals carry weight in short-video performance [2].

These are associations, not causal effects. Creators who appear on camera early may also be more established, have larger audiences or promote different products, and boosting with ads is hard to separate from organic performance [4]. The model gives a modest signal (AUC 0.64, 90% range 0.57–0.70). It was not restricted to analyses that never saw sales (cleanOnly is false), so its performance may be flattered. Its strongest drivers were filmed live action (odds ratio 2.54) and the combination of a visual pattern-break with a review or testimonial (1.92), while a product in the first second was associated with lower odds (0.62).

7. Conclusion

Among top-selling TikTok Shop videos in this window, a creator's face in the first second is associated with sales per view of 1.06× the market average, while videos without the creator on camera sell at 0.89×. The differences are small compared with those between formats, where unboxing (1.37×) and demo or tutorial (0.92×) sit at opposite ends among the well-sampled formats.

Brands choosing what to make should treat an early face as a modest advantage to test, not a guarantee, and pair it with formats that sell above average. Longer windows and category-level cuts would show whether the pattern holds.