Quick Answer: Rising enrolment numbers can mask a product that isn’t actually teaching anyone anything. Most LMS platforms are architected to optimize signups, not completion or mastery.
The largest study of its kind, covering 290 HarvardX and MITx courses and 4.5 million participants, found a typical course certifies about 500 learners out of 7,900 who access any content. But here’s the part most people miss: among the 662,000 learners who reached at least half the content, the median certification rate was 36%. The enrolment number was never the right denominator. The fix is not more users; it is redesigning around feedback, personalization, and outcome-based metrics like completion to mastery ratio instead of logins and seat count.
There is a specific kind of high that comes from watching your enrolment numbers climb. The dashboard glows green, the board deck writes itself, and every new signup feels like validation. Then someone finally asks the harder question: how many of those learners actually finished anything?
The silence that follows is where most EdTech founders discover the truth. Enrolment is easy to sell. Learning is hard to build. Most learning management systems are quietly optimized for the former while claiming to deliver the latter, not as a design flaw, but as the default architecture of an industry that measures what is easy to track, not what actually matters.
The Enrolment Illusion
The most cited research on this comes from Chuang and Ho’s four-year study of HarvardX and MITx courses: 290 courses, 4.5 million participants, 245,000 certificates, and 28 million participant-hours. In a typical course, roughly 7,900 learners access some content after registering, about 1,500 explore half or more, and around 500 earn a certificate.
Read quickly, that looks like a catastrophe. Read carefully, it is something more useful. Among the 662,000 participants who actually accessed at least half of a course, the median certification rate was 36%. The headline completion rate is not primarily a measure of teaching quality. It is a measure of how many people in your denominator were never really learners in the first place.
That distinction matters commercially. Founders mistake the enrolment curve for product market fit because it is the metric that is visible first and hurts least to report. Enrolment measures curiosity, not commitment, and the illusion breaks predictably: refund requests, churn at renewal, and reviews that praise the marketing while quietly noting “I didn’t finish.”
Passive Content Is a Silent Engagement Killer
The default video, quiz, next module format persists because it is cheap to produce. It is also cognitively weak. In a foundational experiment published in Psychological Science, Roediger and Karpicke had students either restudy prose passages or take recall tests on them. On tests given five minutes later, restudying won. On tests given two days and a week later, the students who had been tested retained substantially more.
The detail that should worry every EdTech founder is what happened to confidence. Repeated studying increased students’ confidence in their ability to remember the material, even as it left them retaining less than the tested group. Passive review feels like learning while producing less of it. That is a vanity metric operating inside the learner’s own head, and a platform built on watch-completion events will faithfully record it as success.
A learner can finish a video at double speed, click “complete,” and retain almost nothing. The system logs a win; the learner logs zero transferable skill.
The Feedback Loop Nobody Built
Most LMS platforms collect data but rarely close the loop with the learner. Timely, actionable feedback is one of the strongest predictors of learner persistence, yet the silence between “enrolled” and “certificate issued” is where motivation quietly dies.
A functioning feedback loop does not need to be elaborate. It needs to exist, to arrive close in time to the attempt, and to tell the learner something more specific than a percentage. This is a build decision, not a content decision, which is why it is usually the first thing cut when a roadmap gets compressed.
Friction Is the Product Killer You’re Not Measuring
Every additional step between intention and action costs you learners. Unclear progress indicators, clunky navigation, buried resume-where-you-left-off states, and neglected mobile experiences all add friction, and mobile matters more each year as a growing share of learning happens on phones rather than desktops.
Founders rarely audit their own product the way a confused first-time user would experience it, but that audit routinely surfaces the friction quietly bleeding completion. The most valuable version of this exercise is watching a real learner attempt to resume a half-finished course on a phone, on mobile data, three days after they last opened it.
Personalization Theater
“Adaptive learning” is one of the most overused claims in EdTech marketing. A genuinely adaptive system adjusts what a learner sees next based on demonstrated performance. Inserting a learner’s name or letting them pick an avatar creates a feeling of customization without changing the actual learning path.
Real personalization requires granular content tagging, a defined competency model, and continuous iteration, which is exactly why it is often skipped in favor of what photographs well in a demo. It is the same pattern we mapped in the Wizard of Oz problem: a convincing surface layer standing in for engineering that was never built.
When Analytics Become a Vanity Mirror
Dashboards built for investors and dashboards built for decisions are not the same thing. Activity metrics like logins and hours flatter a growth narrative. Completion velocity, retention curves, and mastery scores actually predict whether the product works.
Here is how each vanity metric maps to the decision metric that should replace it:
| The vanity metric | What it actually measures | The metric that replaces it |
|---|---|---|
| Total enrolments | Marketing reach and curiosity | Completion rate among learners who passed the first module |
| Logins per week | Habit, or a badly designed resume state | Progress made per session |
| Videos watched | Playback events, including at 2x with the tab hidden | Retrieval attempts passed without a retry |
| Hours on platform | Time spent, which friction inflates | Time to demonstrated competency |
| Certificates issued | Course-end events | Retention of the skill at 30 and 90 days |
Reframing analytics as a diagnostic tool rather than a highlight reel is uncomfortable, but it is the only version that leads to real improvement. A useful test: if your dashboard cannot flag a learner who is likely to churn next week, it was built for the board deck.
Enrolments climbing while completions stay flat?
Techuz builds learning platforms instrumented for outcomes: retrieval-based assessment, real adaptive paths, and analytics that flag at-risk learners before renewal season does.
The Motivation Cliff
Self-paced learning follows a predictable decay curve: high enthusiasm in week one, a sharp drop by week two. Lack of motivation consistently ranks among the top reasons learners abandon online courses, often above content quality itself.
Gamification alone rarely fixes this. Badges and streaks delay the drop-off if the content itself lacks purpose or feedback, because the extrinsic reward has nothing underneath it to sustain. Streaks measure showing up. They do not measure getting better, and learners work out the difference faster than product teams do.
Designing for Outcomes, Not Just Output
The shift founders need to make is philosophical before technical: from “content delivered” to “competency achieved.” Backward design, popularized by Wiggins and McTighe in Understanding by Design, starts from the desired outcome, defines what evidence would prove a learner reached it, and only then builds the path.
Applied to an LMS roadmap, that inverts the usual build order. Assessment stops being the thing bolted on at the end of a module and becomes the thing the module is designed backward from. It is the same discipline that separates a validated MVP development effort from a feature list built on assumption, and the same one we described in what investors really mean by weak product: decisions you can defend, not artifacts you happened to produce.
Measuring What Actually Matters
A metrics framework centered on learning impact should track completion to mastery ratio, retention over time, and applied skill transfer. Kirkpatrick’s four levels of training evaluation offer a structured way to move beyond activity tracking toward measuring real behavior change.
Most LMS dashboards stop at level one. The commercial problem with that is straightforward: level one is the only level a renewing buyer does not care about. Enterprise learning buyers renew on level four, and a platform that cannot produce evidence above level one is asking to be judged entirely on price.
Growth Without Learning Is a Leaky Bucket
Growth without learning outcomes is a leaky bucket dressed up as a rocket ship. The founders who win long term redefine “scale” around impact, not signups, and they instrument for it early, because retrofitting outcome measurement onto a platform architected around content delivery is a rebuild, not a feature.
If your next board meeting only showed completion and mastery data, would you still be proud of the slide?
Build a platform that can prove it teaches
As an LMS development company and a SaaS product development company, Techuz builds learning products with competency models, adaptive paths, and outcome analytics designed in from the first sprint.
FAQs
Why do so many LMS platforms have such low completion rates?
Most are architected around content delivery and enrolment tracking rather than active learning design. It is also partly a measurement artifact: the HarvardX and MITx research found that among learners who reached at least half of a course, the median certification rate was 36%, far higher than the headline number suggests.
Isn’t enrolment growth still a useful signal for an early stage product?
It indicates demand but not value delivered. Pair it with completion and mastery data, and segment completion by learners who finished the first module, so the denominator reflects people who actually started rather than everyone who ever signed up.
What’s the difference between real personalization and personalization theater?
Real personalization adapts what a learner sees next based on demonstrated performance, which requires granular content tagging and a competency model. Theater refers to surface features like names and avatars that do not alter the actual path.
How can a founder tell if their dashboard is built for vanity or decisions?
Check whether the top metrics predict retention. If the dashboard cannot flag a learner likely to churn next week, or identify which module causes the most drop-off, it is built to flatter rather than to diagnose.
Does gamification actually solve low engagement?
Only as a layer on substantive content, not a replacement. Badges and streaks measure showing up rather than getting better, so they tend to delay the drop-off rather than prevent it when the underlying content lacks purpose or feedback. If engagement is structurally low, custom LMS development focused on assessment and feedback design will move the number further than a rewards layer.
Sources
- Chuang & Ho, HarvardX and MITx: Four Years of Open Online Courses, Fall 2012 to Summer 2016
- Roediger & Karpicke, Test-Enhanced Learning: Taking Memory Tests Improves Long-Term Retention, Psychological Science (2006)
- Wiggins & McTighe, Understanding by Design (ASCD)
- Kirkpatrick Model of Training Evaluation, four levels