Online course completion rates: what the research really says
On this page9
What is a normal online course completion rate?
Ask how many people finish online courses and you will hear anything from 3% to over 90%. Both can be true, because they count different people in different courses. The best published data comes from open online courses (MOOCs), where anyone can sign up for free.
A study by Katy Jordan, published in 2015 in the International Review of Research in Open and Distributed Learning, compared completion across many courses. She gathered data on 221 MOOCs and found completion figures for 129 of them. Completion, defined as the share of enrolled students who completed (most often by earning a certificate), ranged from 0.7% to 52.1%, with a median of 12.6%. Even this dataset had to be pieced together: about a third of the courses’ figures came from news articles, and another third from academic reports.
Platform data tells the same story over time. Justin Reich and José Ruipérez-Valiente analysed 565 course runs from MIT and Harvard on edX between 2012 and 2018. In Science, in 2019, they reported that low completion had not improved in six years. Of participants, meaning people who opened a course at least once, 3.13% completed in 2017–18, against 5.85% in 2014–15.
12.6%
Median completion across 129 open online courses
129
MOOCs with completion data, out of 221 studied, ranging from 0.7% to 52.1%
3.13%
Of participants in MIT and Harvard courses on edX completed in 2017–18, down from 5.85% in 2014–15
Why your course is not a MOOC
A MOOC is free, optional and open to anyone with an email address. Many people sign up to browse, with no plan to finish, and nothing happens if they leave. A workplace or paid course differs on four counts: motivation, requirement, cost and cohort. Learners need the skill for their job, the training is often required, someone has paid for it, and people often start together with a date to finish by.
Reich and Ruipérez-Valiente’s own data shows how much this matters. In 2017–18, 3.13% of all participants completed. Among those who said in a survey that they planned to finish, 15.42% did. Among those who paid for a verified certificate, 46.02% did. Same platform, same courses, very different numbers, depending on who is counted.
So a MOOC median is not a benchmark for your onboarding program. We know of no peer-reviewed study that gives a typical completion rate for workplace e-learning. The honest comparison is with your own courses: last quarter’s numbers, counted the same way.
What predicts completion: length, grading and the first two weeks
Jordan tested which course features went with higher completion. Three stood out, and together they explained about 60% of the differences between courses. Longer courses had lower completion. Courses graded only automatically had higher completion than courses that used any peer grading. And newer courses had higher completion, which she thought was most likely because enrolments were falling, and possibly because courses were being improved over time.
She was careful about why peer grading hurt. Among her possible reasons: peer assessments may be more rigorous, or learners may lose interest while they wait for feedback. From the length finding and the early drop-off, she concluded that “a case could be made for shorter, more modular courses.”
Timing matters as much as design. In a subsample of courses with weekly activity data, most of them four weeks long, around half of enrolled students never showed up, and the first two weeks were critical for engagement. After that, the share of active learners levelled out. The people you lose, you mostly lose at the start.
Short videos hold attention
Video is where many courses lose people quietly. Philip Guo, Juho Kim and Rob Rubin studied 6.9 million video sessions in four edX courses. The median time people spent watching was at most six minutes, however long the video was, and in videos longer than nine minutes, viewers often made it less than halfway.
Their first recommendation was to cut videos into chunks shorter than six minutes. The study measured attention, not completion of a whole course, but the logic matches Jordan’s: smaller pieces give people more places to finish and fewer places to quit.
Cohort-based vs self-paced: what the evidence supports
You will often read that cohort-based courses, where a group starts together with a schedule and live sessions, reach 90% completion or more, against single digits for self-paced ones. The low figure usually traces back to MOOC research like the edX data above. The high figure usually comes from companies reporting on their own courses, with no published method: who counted as enrolled, whether people paid, and whether those who left in the first week were counted at all.
The peer-reviewed evidence answers a related question, and it measures something else. A 2009 meta-analysis by Robert Bernard and colleagues in Review of Educational Research pooled 74 studies of distance education. Courses designed for more interaction, between students, with the teacher or with the content, led to better achievement, with an average effect size of 0.38. The stronger the interaction, the higher the achievement, and that link was strongest in courses that ran without live sessions.
That is learning, not completion, and it covers all three kinds of interaction, not only peers. Our reading: a cohort helps because it adds a date, other people, and a teacher who notices when you stop. You can design those in without forcing everyone onto one schedule.
Five levers a training team controls
None of this needs a new theory. From the research above, and our own reading of it, here are five things a training team can change this quarter:
Shorter modules: split long courses into short parts, and long videos into pieces under six minutes.
Early auto-graded checkpoints: a short quiz in the first days gives quick feedback and shows who has really started.
A fixed start date or cohort milestones: a date to begin and a date to be done by turn “some time” into a plan.
Nudges when someone stalls: a reminder at the right moment, to the person who has fallen behind, not to everyone.
Help at the moment of need: when someone is stuck late at night, an answer from the course itself keeps them going instead of waiting for the next session.
What to track, and where Classbase fits
“Completed” on its own misleads. Jordan thought the rise in completion rates over time most likely came from fewer people signing up, not more people finishing. And a required course can show high completion while people click through without learning. Track four numbers together: started, active in the last week, completed, and passed a real test.
Classbase is designed around these levers. The course builder turns your files into short lessons made of interactive cards. Tests draw questions from a pool and set a pass mark. Cohorts schedule live sessions with reminders, and attendance counts toward completion. Automations nudge only the people who are behind, in the app and by email, at most once a day. The AI tutor answers from your own course and passes questions it can’t cover to an instructor. In Reports you ask for each of the four numbers in plain words, and every chart says what was counted.
If you want to see your own completion numbers, counted honestly, book a demo and we’ll show you.
Sources
- Jordan (2015): Massive open online course completion rates revisited: Assessment, length and attrition, IRRODLirrodl.org
- Reich and Ruipérez-Valiente (2019): The MOOC pivot, Sciencedoi.org
- Reich and Ruipérez-Valiente (2019): The MOOC pivot, supplementary datagithub.com
- Guo, Kim and Rubin (2014): How video production affects student engagement: An empirical study of MOOC videos, ACM Learning at Scaleup.csail.mit.edu
- Bernard et al. (2009): A meta-analysis of three types of interaction treatments in distance education, Review of Educational Researchdoi.org
Questions
What is a good completion rate for an online course?
There is no published benchmark for workplace courses. Open MOOCs had a median of 12.6% across 129 courses in one study, but required or paid training is a different situation. Compare each course with your own past runs, counted the same way.
What is the average MOOC completion rate?
In Jordan’s 2015 study, across the 129 MOOCs with completion data, completion ranged from 0.7% to 52.1%, with a median of 12.6%. In MIT and Harvard courses on edX, 3.13% of participants completed in 2017–18.
Do cohort-based courses have higher completion rates?
The very high figures you see are mostly self-reported, with no published method. Peer-reviewed research shows that designing in interaction with peers, teachers and content improves achievement, which is not the same as completion. In our view, a start date, a group and someone who notices are the parts that matter.
How do you increase course completion rates?
Keep modules short and videos under six minutes, add an auto-graded check early, give a start date or milestones, nudge people when they stall, and make help available when they get stuck. Watch the first two weeks most closely.
Why is completion rate a misleading metric?
Because it depends on who is counted. Fewer casual sign-ups raise the rate without anyone learning more, and a course can be completed without being learned. Track started, active, completed and passed together.
