How Long Should a Training Video Be? Evidence for Course Builders

A course video should be as long as one idea takes to teach, and for most explanations that means under six minutes. That’s the short answer, and it’s the one most guides give. What they usually leave out is where the six minutes comes from, what it measured, and why tutorials, demos and recorded sessions follow different rules.

This guide sets out what the research actually found, then gives a step-by-step way to decide the length of each video in your course, with a starting length for each type of video.

Where the six-minute rule comes from

The figure traces back to one large study. Philip Guo, Juho Kim and Rob Rubin analysed 6.9 million video-watching sessions from four edX courses in 2012: 862 videos watched by nearly 128,000 people (Guo, Kim & Rubin, 2014). They used the time someone spent on a video as their main measure of engagement, along with whether they attempted the problem that followed it.

Video length was, in their words, “by far the most significant indicator of engagement”. Median watching time was at most six minutes, however long the video was, and learners often stopped before halfway through videos longer than nine minutes. They were also less likely to try the problems that followed longer videos.

A companion study on the same courses looked at where people stopped. About 55% of viewing sessions ended before the video did. The authors’ model predicted 53% dropout for a five-minute video and 71% for a 20-minute one, and much of that dropout came in the first few per cent of the video (Kim et al., 2014).

So far, so clear. Three things complicate it.

It measured watching, not learning. Guo and colleagues say plainly that “it is important not to draw any conclusions about student learning solely from our findings about video engagement”. Watching is usually a precondition for learning from a video, but someone can watch every second and learn little.

Length may stand in for planning. The edX producers the team interviewed suggested that shorter videos were engaging partly because they were better planned. The authors note they didn’t have the data to test this. The course designed for online delivery held learners longer, even in its longer videos, than the course that reused recorded classroom lectures.

Other learners behave differently. A study of working professionals watching a university’s online course videos found viewing patterns, including watching sections on a loop, that “are not reducible to a simple rule of thumb regarding video length” (Lagerstrom, Johanes & Ponsukcharoen, 2015). Free MOOC audiences may not behave like staff completing required training or professionals who paid for a course.

What the learning research adds

The stronger evidence on learning comes from research on segmenting: breaking instruction into pieces the learner takes one at a time, rather than one continuous stream. A meta-analysis of 56 studies (88 comparisons) found a reliable segmenting effect, with small to medium gains in both retention and transfer (Rey et al., 2019).

Segmenting works for the reasons set out in cognitive load theory: a pause at the end of each idea gives working memory a chance to catch up before the next one arrives. One detail is worth knowing. In the meta-analysis, learners with more prior knowledge gained more from segmenting on retention than novices did, so don’t assume it only helps beginners.

The other strand is what happens between segments. In an experiment with Harvard undergraduates, Karl Szpunar and colleagues split a 21-minute statistics lecture into four segments of about five and a half minutes. Students who answered a few test questions after each segment reported half as much mind wandering as those who restudied or weren’t tested, took more notes, and scored higher on the final test: 90% against 76% and 68% (Szpunar, Khan & Schacter, 2013). It was a lab study with a single short lecture, but the direction matches the wider research on testing.

Short tests between segments of a 21-minute lecture halved mind wandering, to 19% of probes

Share of attention probes on which students reported mind wandering during a video lecture split into four segments of about 5.5 minutes, by what they did between segments (Experiment 2, Harvard undergraduates) · % of probes with mind wandering reported

Put together, the evidence says less “keep videos short” than “keep each idea short, and give learners something to do with it before the next one”. Short videos are simply the easiest way to build that into an online course.

How to decide the length of each video

To keep this concrete, we’ll follow one example: a grant writer turning a 40-minute recorded webinar on writing a grant application into a self-paced module. It’s an invented example, but it’s typical of the material course creators start from.

1. Start from one idea or one task

Before you think about minutes, list what the video has to teach. One video, one idea, or one task a learner could do afterwards. If you can’t name it in a sentence, the video isn’t ready to record.

For the grant writer, the webinar covers at least five ideas: reading the funder’s guidelines, writing the need statement, setting outcomes, building the budget, and checking the application before submission. That’s five videos, not one, and each has a clear job.

This step is the one that matters most. Writing learning objectives first makes it almost automatic.

2. Match the length to the type of video

Different videos do different jobs, and the research on how learners watch them differs too.

The tutorial row comes from the edX data too. Learners watched only two to three minutes of each tutorial on average, whatever its length, but often re-watched and jumped to the part they needed. Guo and colleagues’ advice was that for tutorials “length does not matter as much, but support re-watching and skimming”. Kim and colleagues found more replaying in tutorials as well, with most peaks falling where the picture changed, such as a switch from slides to the speaker.

For the grant writer, the budget section is a worked demonstration: filling in a real budget template line by line. That can run ten minutes if it needs to, broken into chapters for each part of the budget. The video on reading the funder’s guidelines is an explanation and should come in well under six.

We have seen a significant increase in the participation of students with shorter lectures (our longest being 10-12 minutes) over the usual ‘film from the back corner of a room and post it online’ approach. It also forces the lecturer to introduce different learning modalities into online learning, or create case studies that focus on one or two key concepts. This creates more focus in the filmed content, and also more opportunities for repetition and participation. For units where this isn’t possible, we have split lectures at natural chapter points and when an in-class discussion occurred, put a forum for online and distance students to participate in. This has resulted in more engagement for online learners, and seen the development of online cohorts and discussions around key learnings.

3. Put something to do between videos

Length is only half the design. What learners do after each video is the other half, and it’s where the Szpunar results come from.

After each short video, give learners a question, a small task or a prompt to apply the idea. For the grant writer: after the need statement video, learners write the first two sentences of their own; after the budget demonstration, they complete three lines of a budget template. The formative assessment entry has more on low-stakes checks, and our guide to writing multiple-choice questions covers quick quizzes that teach.

Your course platform may let you place a quiz or a task between lessons, or questions inside a video. Check what yours supports before you plan the module around it.

4. Earn the first 30 seconds

The dropout data suggests learners decide quickly. In Kim and colleagues’ model, a large share of the people who left a video left in its first few per cent. Open with what the learner will be able to do by the end and why it matters to them, then start teaching. Cut the long title sequence, the personal introduction that repeats on every video, and the recap of the previous lesson that the learner just watched.

5. Plan before you record

The edX producers thought the planning stage had the biggest effect on how engaging a video ended up. Guo and colleagues recommended investing in it, even for instructors who prefer to record live. A short script or a set of notes does most of the work: it keeps a video to its one idea and stops the ten-minute ramble that should have been four.

As a rule of thumb, plan about 150 words of narration a minute, so a six-minute video is roughly 900 words. Our guides to writing an eLearning script and to storyboarding a module cover both steps. The edX team also found that learners engaged more when instructors spoke faster, which they took as a sign of enthusiasm rather than a reason to rush: be natural, not slow and careful.

Christine Brame’s review of the video research adds two habits worth building into the plan: signalling, using on-screen text or symbols to highlight what matters, and weeding, cutting “interesting but extraneous information that does not contribute to the learning goal” (Brame, 2016). Weeding is usually where the minutes go.

6. Check your own data after launch

The studies give you a starting point. Your learners give you the answer. After launch, look at where people stop watching and which videos they replay, if your platform reports it, and at the questions most people get wrong after each video. A video with a sharp drop at the two-minute mark usually has a slow patch there. A video that’s replayed again and again may be teaching something important too quickly.

Common mistakes

1. Cutting by the clock. Splitting a 30-minute lecture into five six-minute pieces, with cuts mid-sentence, gives you the length without the benefit. Cut by idea.
2. Treating six minutes as a target. If an idea takes three minutes, stop at three. Padding a video to a round number costs learners’ attention for nothing.
3. Chopping up demonstrations. A procedure split across four videos with no chapters is harder to follow and harder to find later. Keep it together and make it skimmable.
4. Video after video with nothing in between. A playlist of short videos is still passive. The gains in the research come with a task or question after each one.
5. Reusing recorded lectures as they are. They’re a fine optional extra, but a course built on them asks learners to sit through a format the data says they leave.

In practice

For most self-paced courses, we’d plan explanation videos of three to six minutes, demonstrations as long as the task with chapters, and a question or task after every video. We’d treat anything over ten minutes as needing a reason, and we’d look at the drop-off data after the first group of learners rather than arguing about the ideal length in advance.

Short videos are also easier to fit between other work and to find again later. The same thinking runs through our entry on microlearning: short works when each piece is complete and leads to practice, not when a long lesson is simply chopped up.

What to do next

Open the longest video in your course and write down, in one sentence each, the ideas it teaches. If there’s more than one, mark where each ends. That list is your new video plan, and the gaps between the ideas are where your questions go.