Online Course Completion Rates: 3% Is the Wrong Number

The most repeated completion figure in online education, 3 to 6 percent, is not a research finding. Follow the citation on the page currently ranking first for this query and it ends at a LinkedIn post. The peer reviewed numbers are higher, more specific and far more useful, and you can read every one of them yourself.

What is the average online course completion rate?

The best academic answer is a median of 12.6 percent, with a range from 0.7 percent to 52.1 percent. That comes from Katy Jordan's 2015 paper in the International Review of Research in Open and Distributed Learning, which assembled completion data across 221 MOOCs.

Three caveats that almost nobody who quotes that number mentions:

• The dataset stops in November 2013. It is describing a version of online education that no longer exists.
• It was hand compiled from news articles, academic papers and instructors' social media posts, not from platform telemetry. Roughly a third of the entries came from press coverage.
• Enrolment data existed for 220 courses but completion data for only 129 of them.

It is honest, careful work. The problem is what happened to it afterwards.

Where does the 3 percent number actually come from?

The article ranking first for this query states that completion is 5 to 15 percent for online courses and 3 to 6 percent for MOOCs, and attaches a reference marker to it. That reference is a LinkedIn Pulse post. There is no study underneath.

It gets stranger. Thinkific, a platform that sells software to people running paid courses, tells its own customers that only 5 to 15 percent of students finish, and links that claim to Jordan's free MOOC page. A paid course platform is benchmarking its customers against 2013 data on free university courses that anyone could enrol in with one click and no money at risk.

If you have ever felt that your completion rate is bad, check what you were comparing it to.

What happens when you fix the denominator?

Dividing finishers by registrations is the core error. On a free platform, registration costs nothing, so the denominator fills with people who never opened the course.

Harvard and MIT published four years of edX data in January 2017 covering 290 courses and 4.5 million participants. Their reframing is the single most useful number in this entire subject: among the 662,000 people who accessed at least half the course content, the median certification rate was 36 percent.

Their description of a typical course: about 7,900 people access content, about 1,500 explore half or more, and about 500 certify. Same course, three completely different completion rates depending on which number you put underneath.

Do paying students finish?

Yes, at roughly fifteen times the rate. Reich and Ruiperez-Valiente published an analysis in Science in January 2019 covering 5.63 million learners and 12.67 million registrations across every MIT and Harvard edX course from 2012 to 2018.

Completion across all participants in 2017 to 2018 was 3.13 percent. Among learners who had paid for the verified track, it was 46 percent.

That gap is the whole story. The famous low numbers describe free enrolment behaviour. If someone paid you for a course, the free MOOC benchmark tells you nothing about what to expect.

What actually moves completion?

Two findings survive scrutiny, both with real sample sizes.

Deadlines, by a wide margin

Kaila and Lemstrom published a 2023 study in Informatics in Education comparing identical course content delivered on a schedule and delivered open access. The Java course passed 31.62 percent of 3,972 scheduled students against 9.72 percent of 7,148 unscheduled ones. The Python course ran 33.18 and 44.34 percent scheduled against 11.41 percent unscheduled across 9,998 learners.

Same material, same exercises, same platform. The only difference was whether anything was due on a date. Completion roughly tripled.

Length, negatively

Jordan's regression found that each additional week of course length cost 1.75 percentage points of completion, and that peer graded assessment cost 9.6 points against automatic grading. Both were significant at p below .001 on a model explaining about 60 percent of variance.

Two levers, then, and neither of them is production quality. Put dates on things and make the course shorter. We worked through the second one separately, because course length is also the single biggest driver of what a course costs to build.

The numbers worth quoting, in one place

• Median MOOC completion, 221 courses: 12.6 percent. Range 0.7 to 52.1 percent. Jordan, 2015, data to November 2013.
• Median certification among learners who reached the halfway point: 36 percent. HarvardX and MITx, 2017.
• Completion among paid verified edX learners, 2017 to 2018: 46 percent. Against 3.13 percent for all participants. Science, 2019.
• Scheduled versus open access on identical content: 31.62 percent against 9.72 percent. Kaila and Lemstrom, 2023.
• Cost of each extra week of course length: 1.75 percentage points of completion. Jordan, 2015.

What no platform will tell you

There is no credible aggregate completion benchmark for paid creator courses. None. Teachable, Kajabi, Podia, Udemy and Coursera have published no such figure. Every comparison table claiming that one platform completes 8 to 11 percent better than another traces to affiliate content, not data.

Some vendors publish self reported numbers with no methodology, which is better than nothing and worse than research. Treat them accordingly, and treat any unlinked percentage in this subject as decoration.

The practical consequence is that your own completion rate is the only benchmark you have. Measure it against your last cohort, not against a number from a blog. Coaches and course creators who track it properly usually find the drop happens in the first two lessons, which Jordan also found: weeks one and two are decisive, and after week three the gap between accessing material and submitting work is under 5 percent.

Key takeaways

• The 3 to 6 percent completion figure is not a study. On the top ranking page it cites a LinkedIn post.
• The best peer reviewed number is a 12.6 percent median across 221 MOOCs, from data ending in November 2013.
• Fix the denominator and it changes completely: 36 percent median certification among people who reached the halfway mark.
• Paid learners completed at 46 percent against 3.13 percent for all participants in the same courses.
• Deadlines roughly tripled completion on identical content, from 9.72 to 31.62 percent.
• Every extra week of course length costs about 1.75 percentage points of completion.
• No platform publishes a real completion benchmark for paid courses, so your own last cohort is the only honest comparison.

Common questions

What is a good completion rate for a paid online course?

There is no published benchmark for paid creator courses. The closest evidence is edX verified learners at 46 percent, and that is a useful target for a paid cohort. Judge yourself against your previous cohort rather than against free MOOC figures, which measure something entirely different.

Why are online course completion rates so low?

Mostly because the denominator includes people who never started. Free registration costs nothing, so enrolment lists fill with browsers. Once you count only learners who reached the halfway point, the median certification rate rises to 36 percent in the Harvard and MIT data.

Do cohort courses complete better than self paced ones?

The evidence points that way. In a 2023 study of identical content, scheduled delivery passed 31.62 percent of students against 9.72 percent for open access. Deadlines appear to matter more than format. Adding real dates to a self paced course captures much of the same effect.

Does course length affect completion?

Yes, negatively. Jordan's 2015 regression across 121 courses found each additional week cost about 1.75 percentage points of completion, significant at p below .001. Shorter courses finish better, and they also cost considerably less to produce.

Before you rebuild anything

If your completion rate looks bad, check the benchmark before you rebuild the course. Then check whether anything in your course has a date attached, because that is the cheapest intervention in the evidence. Rebuilding the videos is the most expensive one, and the research does not support it.

EchoPulse produces course video and the pages and sequences that sell it, for coaches and course creators who would rather ship the next cohort than spend a quarter in an editing timeline. The $299 Pilot runs 14 days, you keep everything produced, and there is no contract.