On the median post, Reels and photos engage almost identically. But the top 1% of posts is 70% Reels - and the top 0.1% is 84% Reels. Same engagement floor. Much fatter right tail. The standard “Reels-first” advice is half right, for the wrong reason.
TL;DR
Two charts that explain the whole thing
The first chart is the one every “Reels vs photos” study cites - the median engagement comparison - and it is genuinely unimpressive. The second chart is the one nobody publishes: how the format mix shifts as you climb the engagement distribution. That is where the story is.
Nearly identical. The format you post doesn't meaningfully change the engagement on a typical post.
Baseline is 54% video. The top 1% of posts is 70% video. The top 0.1% is 84% video.
Per-post ER = (likes + comments) / follower count. Paired comparison requires ≥3 posts of each format from the same account (n=1,343 accounts, 61,607 posts). Tail percentages are computed on the full filtered post pool, not just paired accounts.
The consensus, and what it gets right
Open any Instagram growth guide and the advice is the same: post Reels, photos are dead, the algorithm rewards video. The strong version of this - “pivot to Reels-only or get left behind” - has been the dominant narrative for at least three years, repeated by every social media agency, every algorithm explainer, and most of Instagram's own creator content.
The claim that drives it is partly true. Reels do reach more people. They surface in the Reels tab, in Explore, and to non-followers in ways that photos rarely do. Our data captures the downstream effect of that distribution: when content goes mega-viral on Instagram, it is overwhelmingly Reels. That part of the conventional advice is correct.
The part that is wrong - or at least overstated - is the implication that this advantage shows up on every post. It does not. The typical Reel and the typical photo from the same account engage roughly equally. The cost of pivoting to Reels-only - production effort, abandoning a format the creator may be better at - is paid on every post. The upside it buys is concentrated in a small fraction of posts that go big. Both things can be true at the same time, and they are.
This is the same shape as the cadence study: a metric reported in aggregate hides a more interesting structure underneath. With cadence, the per-post-ER decline obscured the weekly-engagement gain. With format, the headline “Reels go viral more often” obscures the boring median where most creators actually live.
Controlling for follower size
The first objection to the headline finding: maybe larger accounts post more Reels and have more viral posts, so the tail-video skew is just account size repackaged. We split the post pool into five follower cohorts and re-ran the top-1% calculation inside each. The lift holds in every cohort, which rules out the simplest version of that confound.
The video lift in the tail holds inside every cohort - it is not a Simpson's-paradox artifact of cohort mix.
| Cohort | Posts | Baseline video % | Top 1% video % | Lift |
|---|---|---|---|---|
| 1k-2k | 2,109 | 72.2% | 90.5% | 1.25x |
| 2k-10k | 12,940 | 49.6% | 72.1% | 1.45x |
| 10k-50k | 17,483 | 53.3% | 67.2% | 1.26x |
| 50k-100k | 10,481 | 50.8% | 64.4% | 1.27x |
| 100k-500k | 18,594 | 57.2% | 74.6% | 1.30x |
The strongest tail effect is in the 2k-10k cohort (1.45x lift) - the small-to-mid accounts where a viral Reel can fundamentally change an account's trajectory. The smallest effect is in 10k-50k, where the baseline video share is already high enough that the tail does not lift much further. The 1k-2k row has a high baseline (72%) because tiny accounts in this dataset post mostly Reels to begin with, but the tail still tilts even further toward video.
The within-account test
The cleanest version of the question: take an account that posts both formats, look at its single highest-engagement post, and ask which format that post was. This controls for everything that varies between accounts - niche, audience, creator skill, posting time - and isolates the format effect on the same audience.
Out of 1,343 accounts posting both photos and Reels, the highest-engagement post is a Reel 61% of the time.
Tail fatness, within-account
We also measured how much an account's best post in each format outperforms its own median post in that format. This is a direct measure of how “fat” the right tail is for each format on the same account.
At the 90th percentile, an account's best Reel beats its typical Reel by 52x, while an account's best photo beats its typical photo by only 20x. The Reel tail is 2.5x fatter at the top. Same account, same audience.
What this means for you
If an account is genuinely better at photography or carousel-driven content than at video, the data here does not justify a forced pivot. Median engagement is roughly equal between formats. What a photo-only feed gives up is the right-tail upside: the chance of a Reel that breaks out and brings a wave of new followers. Whether that trade is worth it depends on whether the goal is steady performance or growth.
Switching to Reels-only does not raise the typical post's engagement in our data. What it raises is the variance: more shots at a viral hit, and more posts that underperform the median in exchange. Accounts that depend on predictable engagement (brand partnerships paid per impression, e-commerce with steady conversion targets) may not benefit from this trade. Accounts trying to grow audience from zero almost always do.
Mixing photos and Reels captures the upside of one without losing the consistency of the other. The right ratio depends on the production cost difference (if Reels take 5x longer to make, the math shifts), on which format the creator is genuinely better at, and on what the breakout content in the niche actually looks like. There is no universal answer in this data - which is why we end the cadence study and this one the same way: compare yourself to the accounts your audience already follows, not to industry averages.
Apply this to your own niche
Format-niche fit varies more than any aggregate study can capture. Competitor Map analyzes your audience and surfaces the accounts they actually follow, with each account's best-performing posts and median engagement. Look at what format dominates the top posts in your peer set - that signal is stronger than any cross-niche average.
Methodology
Sample
61,607 public Instagram posts from 2,232 accounts. Follower counts capped between 1,000 and 500,000. Pinned posts, posts with comments disabled, and posts with fewer than 5 likes were excluded. Paired per-account comparisons (the 1,343-account subset) require at least 3 image posts AND 3 video posts from the same account.
Why cap at 500k followers
Engagement at the very top of Instagram is driven more by who the account is than by what they post. A celebrity gets hundreds of thousands of likes on a casual photo because the audience came for the person. Including those accounts would let mega-influencer behavior dominate the dataset and obscure what is happening for the creators and brands the study is meant to inform.
Engagement rate
Per-post ER = (likes + comments) / follower count. Paired account values use the median across the account's posts in each format. Tail slices (top 10%, top 1%, top 0.1%) are computed on the full filtered post pool, ranked by per-post ER.
Format buckets - and a limitation
The data layer used for this study stores media type as either “image” or “video.” Instagram's underlying API distinguishes single photo, carousel, and video, but the carousel signal is collapsed into the image bucket in our pipeline. So when this study says “photo,” it means “static post” - a mix of single photos and carousels we cannot separate. Reels are the dominant video format on Instagram in 2026 (Instagram converted nearly all video posts into Reels in 2022), so “video” in this study is effectively Reels content. We will publish a follow-up that splits carousels out once that signal is clean.
Within-account paired comparison
To control for between-account variation (niche, audience size, posting style), we constructed a paired subset of 1,343 accounts that post at least 3 of each format. The paired median ER (0.96% photos vs 1.05% Reels) and the within-account top-post comparison (61% Reels vs 39% photos) are both computed on this subset.
Robustness checks
The tail effect was tested four ways: (1) absolute ER thresholds (ER ≥ 5%, 10%, 25%, 50%, 100%) - all show monotonic video lift; (2) relative-to-account-median multipliers (3x, 5x, 10x own median) - same pattern; (3) within-cohort tail share - lift holds in every cohort with no flips; (4) within-account top-post format - 61.1% video on a sample where the post pool is roughly tied. All four point the same direction.
What this study cannot prove
This is observational data, not an experiment. We can show that Reels go viral more often than photos in this dataset. We cannot prove that an individual account would gain more viral hits by posting more Reels - accounts that post more Reels may differ from accounts that post more photos in ways we cannot measure (creator skill, niche, audience composition). The top-0.1% slice is n=61, which is directional rather than precise; the top-1% slice (n=616) is the more reliable tail metric. Treat the macro finding as a hypothesis to test on your own account, not a deterministic rule.
Questions people ask
Related
Per-post engagement drops 4x as you post more often. Total weekly engagement rises 5.4x. The standard advice optimizes for the wrong metric.
Map every account sharing audience with any public handle. The starting point for niche-aware benchmarking.
Once you know the format mix that wins in your niche, what to actually post. Reel ideas generated from top posts in your competitor set.