Quick answer: Across 120 eligible posts from six retained public-brand samples, Video/Reel posts produced the highest pooled mean of visible interactions, while carousel posts produced the highest pooled median. The Video/Reel mean was 433,299 visible interactions, compared with 221,515 for carousels. The carousel median was 89,331, compared with 50,524 for Video/Reel posts.
That difference matters because a small number of unusually large posts can pull a mean upward. It does not prove that carousels or Reels generally perform better. The six brands have different audiences, observation dates, returned date windows, and format mixes. Among the four samples that contain both Video/Reel and carousel posts, each format has the higher median in two accounts.
Visible interactions in this benchmark means observed likes plus observed comments. All 120 returned posts had both fields available and were eligible for this calculation.
What was observed
Each account has its own observation time. The 20 returned posts for an account can cover a different period from the 20 returned posts for another account.
| Public sample | Observed at | Returned post window, UTC | Eligible posts | Format mix | Mean visible interactions | Median visible interactions |
|---|---|---|---|---|---|---|
| Nike | Jul 28, 2026, 1:13 AM PDT | Jun 13 to Jul 23, 2026 | 20 of 20 | 9 Video/Reel, 7 carousel, 4 image | 294,231 | 154,454 |
| Adidas | Jun 9, 2026, 4:01 AM PDT | May 7 to Jun 7, 2026 | 20 of 20 | 2 Video/Reel, 18 carousel | 489,591 | 215,199 |
| Starbucks | May 6, 2026, 2:05 PM PDT | Mar 23 to Apr 30, 2026 | 20 of 20 | 8 Video/Reel, 10 carousel, 2 image | 56,041 | 37,490 |
| Sephora | May 6, 2026, 2:04 PM PDT | Sep 24, 2025 to May 6, 2026 | 20 of 20 | 18 Video/Reel, 2 image | 6,623 | 2,610 |
| Netflix | Jun 9, 2026, 4:04 AM PDT | May 20 to Jun 8, 2026 | 20 of 20 | 10 Video/Reel, 8 carousel, 2 image | 81,867 | 53,032 |
| Red Bull | Jun 9, 2026, 3:58 AM PDT | Oct 1, 2025 to Jun 9, 2026 | 20 of 20 | 20 Video/Reel | 1,135,564 | 151,271 |
The observation timezone shown above is America/Los_Angeles. Publication-window boundaries are UTC timestamps in the downloadable source register.
Reels/video versus carousel results
| Format | Eligible posts | Mean visible interactions | Median visible interactions | Highest observed value |
|---|---|---|---|---|
| Video/Reel | 67 | 433,299 | 50,524 | 17,643,161 |
| Carousel | 43 | 221,515 | 89,331 | 2,816,858 |
| Image | 10 | 272,219 | 46,070 | 1,930,047 |
| All formats | 120 | 343,986 | 59,375 | 17,643,161 |
The pooled Video/Reel mean is almost twice the carousel mean, but its median is lower. The 17.6 million-interaction maximum in the Video/Reel cohort shows why the average needs context. A typical value described by the median produces a different ranking from the arithmetic mean.
What happens inside the same brand samples?
Only four of the six retained samples contain both Video/Reel and carousel posts. Their medians are mixed:
| Sample | Video/Reel posts | Video/Reel median | Carousel posts | Carousel median | Higher median in this sample |
|---|---|---|---|---|---|
| Nike | 9 | 114,088 | 7 | 274,330 | Carousel |
| Adidas | 2 | 1,588,794 | 18 | 144,605 | Video/Reel |
| Starbucks | 8 | 37,661 | 10 | 39,335 | Carousel |
| Netflix | 10 | 65,244 | 8 | 51,051 | Video/Reel |
This 2 to 2 split is a warning against turning the pooled result into a universal rule. The Adidas Video/Reel comparison also rests on only two posts. These are descriptive observations, not a controlled test of format effects.
Posting-frequency distribution
Posting cadence was estimated from the dated posts returned for each sample. It is not a claim about the account's complete posting history.
| Estimated posts per week | Number of samples | Samples |
|---|---|---|
| Under 1 | 2 | Sephora, Red Bull |
| 1 to 2.9 | 0 | None |
| 3 to 4.9 | 3 | Nike, Adidas, Starbucks |
| 5 to 7.9 | 1 | Netflix |
| 8 or more | 0 | None |
The source-window estimates were 3.32 for Nike, 4.23 for Adidas, 3.51 for Starbucks, 0.59 for Sephora, 6.88 for Netflix, and 0.53 for Red Bull. The Sephora and Red Bull windows extend much farther back than the other samples. Historical or pinned posts in a returned set can create long gaps, so those two low estimates should not be read as current schedules.
Download the data
- Anonymized 120-post dataset
- Format summary
- Source register
- Machine-readable summary
- Source manifest and hashes
The post-level download uses sample IDs A through F. It excludes captions, post identifiers, and source-post URLs. The separate source register maps each sample ID to the public account and Instaseer sample page so readers can verify scope without placing identifying post content in the analysis file.
Methodology
The benchmark uses the six retained JSON observations behind Instaseer's published brand samples. Each source contributes 20 returned posts. Formats labeled Video or Reel in the source are grouped as Video/Reel; Carousel and Image remain separate.
For each post, visible interactions are calculated as:
visible interactions = observed likes + observed comments
A post is eligible only when both values are available as safe integers. Mean is the sum of eligible visible interactions divided by eligible posts. Median is the middle eligible value after sorting, or the average of the two middle values for an even-sized group.
The posting-frequency estimate is:
(number of dated posts - 1) ÷ elapsed days between oldest and newest dated post × 7
Posts explicitly marked as pinned are excluded from that estimate. These retained files do not provide a pinned marker, so a historical post returned among recent items can lengthen the apparent window. For broader collection, comparison, and evidence boundaries, read the public Instagram research methodology.
The calculation script, exact source hashes, and output tables are retained with this edition so the result can be reproduced.
Limitations
- This is a selected six-brand sample, not a representative sample of Instagram.
- Accounts were observed on different dates and their 20 returned posts cover different time spans.
- The brands differ in audience size, category, campaign activity, and content strategy.
- The format groups are unbalanced. Red Bull contributes 20 Video/Reel posts, while Adidas contributes only two.
- Visible interactions do not include reach, saves, shares, watch time, completion rate, clicks, conversions, paid distribution, or private audience data.
- Public counters can change after observation. This edition reports the retained values at the stated observation times.
- A pooled association cannot establish that post format caused a result.
- The returned posts may not represent an account's full publication history, especially where an older or pinned post appears in the loaded set.
- No follower-normalized engagement rate is used here, because the research question is about visible post interactions and source-window cadence.
Reuse and attribution
The two charts may be embedded unchanged if their visible attribution remains intact and the publisher links to this benchmark page. Suggested attribution: Source: Instaseer, Instagram Content Benchmarks (six public-brand samples, 120 posts).
Writers should cite the edition, sample size, observation-date differences, and metric definition. A defensible summary is: “In Instaseer's selected six-brand sample, Video/Reel posts had the higher pooled mean, while carousels had the higher pooled median.” It would not be accurate to write that the study proves one format performs better across Instagram.
Continue the research
Use the Instagram account analyzer to inspect another public account with dated coverage. Compare two accounts with the Instagram account comparison tool, review the Instagram Reels competitor-analysis workflow, or calculate a source window with the Instagram posting frequency analyzer.