Video engagement rate benchmarks for 2026: viewers watch about half, then less

The average video under one minute gets watched only about halfway through: across video types, sub-minute engagement rates run from 44% to 56%, with educational videos at the top at 56%, per Wistia's 2026 State of Video Report, an analysis of more than 13 million videos and 79 million hours of viewing. Engagement rate there means the share of a video people actually watch, on average.

And it only goes down from there: videos over an hour hold 11-17% depending on type. Half is the ceiling, not the norm.

I read the full report (it sits behind a download form, so I am quoting the PDF directly, the numbers below are from its Benchmarks chapter) because these are the only public per-length, per-type engagement benchmarks built on a dataset this size. Here is the usable core, plus the one paradox in it that most summaries miss.

The pattern by type is remarkably consistent: educational content wins almost every length bracket, and the report calls it out directly, short explainers and hour-long deep dives alike outperform the rest. Tutorials and courses sit right behind. At the bottom, customer testimonials hold attention worst at every single length (46% under a minute, 11% past an hour), which is worth knowing before you build a strategy around them.

By industry, food & beverage and retail lead short-video engagement (both at 54% under a minute), while education and software & tech hold up best as videos get longer. Whatever the industry, the report's own summary is blunt: shorter videos hold attention better across every one of them.

First, put the message where the eyes actually are. If your under-a-minute video is watched 50% through on average, the second half is a bonus track: the offer, the proof, the point all belong in the first half. Wistia's own lead-gen data agrees from the other side: forms placed in the last quarter of a video convert best precisely because only the invested viewers are left by then.

Second, mind the context gap. These benchmarks come from videos hosted on companies' own websites, where a visitor already chose to be. Social feeds are a harsher physics: the average Instagram Reel holds a viewer for 8.5 seconds, not half the runtime. If you make short-form for feeds, the Reels benchmarks and length data are the right yardstick; the cross-platform picture is in the 2026 engagement hub.

Engagement rate is a duration metric: it says how much of the video people watched, never what they actually saw, or why they left when they did. Two videos with an identical 50% engagement rate can be failing in opposite ways, one loses viewers at a boring middle, the other holds them while they stare at the wrong thing entirely and remember nothing you needed them to.

That is the layer I built Jeena for. Real viewers watch your video on their phones with the front camera on, before you publish, and the report shows the frame-by-frame attention heatmap, the visibility map of what nobody noticed, and the exact moment attention broke. The benchmark tells you how long people stay on videos like yours. The attention data tells you what to fix in yours specifically, which is the difference between analytics and watching viewers.