Key takeaways
- Completion falls at every step as length grows, from 85.5% under 5 minutes to 81.5% at 5 to 15 minutes and 70.6% above 15 minutes.
- The drop gets worse in the courses with the most learners, so length costs you most where the stakes are highest.
- Splitting a long course beats shortening it, because each completed course raises the odds a learner finishes the next one.
Where these figures come from
Figures come from an Easygenerator platform dataset of 213,660 created courses and 22.9 million learner-course interactions, recorded between June 2025 and June 2026. The course length analysis covers the 3,802 courses that met the inclusion criteria, which required at least 5 learner results, delivery through an Easygenerator-hosted link, and sessions under 8 hours. Those 3,802 courses carry 730,747 learner results between them.
Full method, sample limits, and the wider findings sit in the 2026 e-learning benchmark report.
What the data says about course length
The key point. Completion drops at every step as courses get longer, and the gap between the shortest and longest courses reaches almost 15 percentage points.
| Course length | Courses measured | Median actual length | Completion rate | Gap against shortest |
|---|---|---|---|---|
| Under 5 minutes | 704 | 2.6 min | 85.5% | Baseline |
| 5 to 15 minutes | 1,232 | 9.3 min | 81.5% | 4.0 points lower |
| Over 15 minutes | 1,866 | 31.9 min | 70.6% | 14.9 points lower |
Completion rate shown as the average per course. Course length measured as the median time spent by learners who reached the end.
85.5%
Courses under 5 minutes on the Easygenerator platform reached an average completion rate of 85.5%, compared with 70.6% for courses running over 15 minutes.
Easygenerator platform data, 3,802 courses and 730,747 learner results, June 2025 to June 2026.
One detail makes these figures more useful than most published advice on course length. We did not ask authors how long their courses should take. We measured the median time that learners spent when they reached the end, so each course carries the length it actually had in practice.
That matters for the comparison. The longest group does not consist of courses that barely crossed the 15 minute line. The median course in that group ran close to 32 minutes, so the gap sits between genuinely short courses and genuinely long ones.
Why length hurts most in your biggest courses
The key point. Long courses that reach many learners perform worse than long courses that reach few, so length costs most in your largest programs.
Two ways exist to average a completion rate. The average per course gives every course equal weight, so it describes what a typical course achieves. The pooled rate gives every learner equal weight, so it describes what a typical learner experienced.
|
87.0% Under 5 minutes, pooled |
67.6% Over 15 minutes, pooled |
Weighted by learner numbers, the gap widens from 14.9 percentage points to 19.4.
Easygenerator platform data, 3,802 courses and 730,747 learner results, June 2025 to June 2026.
Long courses lose more ground once you weight by learner numbers, which means the long courses that reach the most people perform worse than the long courses that reach few. A mandatory 40 minute course rolled out to 3,000 people therefore carries the largest penalty in the dataset.
Where the 15 minute ceiling comes from
The key point. Peer-reviewed research already supports shorter content, though it studied video watching rather than course completion, so our platform data covers the case the literature leaves open.
Philip Guo, Juho Kim, and Rob Rubin analyzed 6.9 million video watching sessions across 862 videos and four edX courses, and they recommended that instructors plan for segments under six minutes. Their engagement measure was how long students watched a video and whether they attempted the assessment afterward. The study appeared at the first ACM Conference on Learning at Scale in 2014.
That research often gets summarized as proof that learner attention runs out after six minutes. It does not establish that. It measured video sessions in university courses on a public platform, and it reported engagement rather than whether anyone finished a course. Corporate e-learning differs on both counts.
Our figures cover what the study leaves open. They measure completion of full courses in workplace training across 3,802 courses, and they show the same direction as Guo and colleagues found for video, with the inflection at 15 minutes rather than six.
The 15 minute figure remains an observed inflection point in one platform’s data. It is not a rule about how human memory works, and several explanations fit the pattern. Long courses tend to cover harder subjects. They carry more assessment. They appear more often in mandatory programs where learners have less control over when they start.
How long each type of training should run
The key point. The research does not break completion down by training type, so the ranges below combine the length data with what each type of training has to achieve. Test them against your own numbers rather than treat them as measured findings.
| Training type | Target length | Why | If it has to run longer |
|---|---|---|---|
| Compliance and regulatory | 5 to 10 minutes | Learners rarely choose this training, so every extra minute costs completion. Legal content also cannot be trimmed for style. | Split by regulation or by scenario, and keep the sign-off assessment in the final course. |
| Onboarding | Under 5 minutes | New starters work through many courses in their first week, and early wins build momentum. | Sequence more short courses instead of extending any single one. |
| Product and process | 5 to 15 minutes | Learners often return to this content as reference, so short courses are easier to find again. | Split by task, so each course answers one question a person might search for. |
| Soft skills and leadership | 10 to 15 minutes | Practice and reflection take time, and the content resists heavy compression. | Split across a series and add practice between courses rather than inside one. |
The pattern behind all four rows is the same. The right unit of e-learning is one thing a learner needs to do, not one topic your organization wants to cover. That principle sits behind microlearning as an approach, and it applies whether or not you use the label.
What to do when the content cannot be shortened
The key point. Split the course instead of cutting it, because each completed course raises the odds that a learner finishes the next one.
Some subjects need 40 minutes. Regulatory training carries required content, and technical processes have steps you cannot skip. Cutting material to hit a length target trades one problem for a worse one.
Three 12 minute courses give a learner three finish lines rather than one, and the platform data supports the approach. Of learners who completed one course, 57.4% went on to complete a second, and 73.3% of those completed a third. Every finish raises the odds of the next one, so a split course reduces drop-off and builds progress at the same time.
“For larger topics, instead of a 30 minute video, I might break it into three 10 minute videos, with a little bit of text or an interactivity or a question in between to re-engage learners as they go.”
Bobby Burchill, Training & Development Manager at ProPharma, speaking on Easygenerator’s video learning webinar, February 2026.
Two practical points come with the approach.
Put the assessment where the decision sits. Where a regulator needs proof that someone passed, keep the assessment in the final course, and let the earlier courses run without a pass mark.
Plan for the maintenance cost. Six short courses mean six objects to update when a policy changes. Courses that reach your LMS through a native integration can update in place, so a change to one course does not require you to re-upload every file by hand. Courses distributed as manual SCORM packages do need that work, and the effort scales with the number of courses you created. Stale content is a common reason that compliance training loses its value between review cycles.
How to measure your own course length
The key point. Authors’ estimates run short, so measure the median time that learners actually spend. The method below produced every length figure in this article.
- Pull the time spent for every learner who reached the end of the course.
- Take the median rather than the average, because a handful of very long sessions will distort an average.
- Exclude sessions over 8 hours, since those almost always come from a browser tab that somebody left open.
- Compare that figure against your completion rate, and repeat once you have at least 5 learner results.
The method works on any platform that reports time spent per learner. A course written for 10 minutes often runs to 18 in practice, and the completion rate follows the real number rather than the intended one.
Four mistakes that cost completion
The key point. Most length problems come from a target set before anyone measured what the course actually does.
Padding a course to hit a required seat time. Some compliance programs specify a duration, and authors add filler to reach it. The filler pushes the course into the group where completion drops by 15 points.
Measuring the length you intended rather than the length you got. The completion rate follows the real number.
Cutting content when the answer was to split it. Removal loses information the learner needed. Splitting keeps all of it.
Treating 15 minutes as a goal. The ceiling marks where completion falls sharply. Nothing in the data rewards a course for approaching it.
What this data cannot tell you
These are observational comparisons, not experiments. Nobody assigned authors to build short or long courses. Long courses tend to cover harder subjects, they carry more assessment, and they appear more often in mandatory programs, so the comparison shows a pattern.
The length analysis rests on a narrower base than the wider research. It covers 3,802 courses against 45,214 for the completion figures in the full report, because length can only be measured on courses delivered through an Easygenerator-hosted link. Courses exported as SCORM packages and played inside an external LMS fall outside it, and many enterprise programs deliver that way.
The training type ranges are inference. The platform data does not segment completion by training type, so those rows combine the length findings with what each type of training has to achieve.
The sample reflects who uses Easygenerator. Our customers lean toward Employee-generated Learning, where subject-matter experts build their own courses. Courses built that way tend to run shorter than courses built by a central team.
Completion is not learning. A learner who finishes a course has demonstrated something on the day. This data says nothing about what they remembered a month later.
About Easygenerator
Easygenerator is an AI-powered e-learning suite that helps organizations create company-tailored training at scale. Built for internal experts and L&D teams alike, its author-first AI captures the depth of organizational knowledge and transforms insights, documents, and ideas into didactically sound courses. Teams can create professional training videos with AI avatars and multilingual voiceovers, and help employees practice real workplace conversations to build skills with confidence. With translation into 75+ languages, Easygenerator makes it simple to scale training consistently across the entire organization.
Easygenerator holds a rating of 4.7 out of 5 on G2 across 175 reviews. Pro plans start at $1,399 per year for a single author and Team plans start at $6,995 per year for five authors, with Enterprise pricing available on request for unlimited authors. Courses publish to any LMS through manual and dynamic SCORM, through xAPI native integrations, and more.
Short courses are also faster to produce. The median course on the platform reached its first publish in under a day, and courses built with EasyAI reached it in about 16 hours against about 23 hours without.
