Data Booster
6 min read
August 2, 2026
Completion on standard corporate e-learning catalogues sits around 12%. Read that as a cost line rather than a statistic: roughly seven of every eight licences you pay for produce nothing at all. Not a partial result, nothing. The learner logged in once and never came back.
Most reporting hides this, because most reporting counts enrolments. Enrolment is a number the L&D team controls and the business does not care about. Completion is the reverse. If your dashboard shows 4,000 people enrolled and nobody has asked which of them finished, you do not yet know whether anything happened.
The drop-off also has a shape. It is not spread evenly across a course. It clusters at the first point that requires actual effort, which is almost always the first exercise rather than the first video. People will watch. Watching is cheap. The moment the course asks them to produce something, the calculation changes, and the course starts competing with the work sitting in their inbox.
Ask learners why they stopped and you get remarkably consistent answers. None of them are about the subject matter.
The first exercise felt generic. A prompting exercise about a fictional bakery teaches prompting in the abstract. It also signals that nobody built this for you. When the exercise instead uses the learner's own report, their own dataset, their own customer email, the same twenty minutes stop feeling like homework and start feeling like work they were going to do anyway.
Nobody noticed they stalled. In most programmes there is no difference between finishing and quietly disappearing. No one follows up, because following up requires knowing who stalled and where, and most platforms only report totals.
There was no deadline. Self-paced means "whenever", and whenever loses to everything with a date on it. A course without a rhythm has no moment at which falling behind becomes visible, to the learner or anyone else.
Their manager never mentioned it. If the manager treats the programme as optional overhead, so will the team. This is the single strongest predictor we see, and it is not a content problem.
Getting in was annoying. An extra login, an enrolment flow through a system people use twice a year, a link that expires. Every step of friction before the learning starts costs a slice of the group, and that slice never comes back.
None of these are exotic. They are the difference between 12% and the numbers below.
1. Give the programme a rhythm. A cohort moving through the material together creates the deadline that self-paced study lacks, and it makes falling behind visible early enough to do something about it.
2. Build the exercises on their own work. Their tools, their data, their cases. This is the most expensive thing to do and the one that changes completion most, because it removes the translation step between the course and Monday morning.
3. Nudge the people who stalled, not everyone. A reminder to the whole group is noise. A message to the eleven people who stopped halfway through section two is a conversation. That requires watching progress per learner rather than reporting a percentage upward.
4. Put a human in the room. Drop-in sessions and applied workshops give people a place to bring the thing they got stuck on. They also create a soft social obligation, which does more for completion than any reminder email.
5. Involve the managers before the learners. Brief them separately, tell them what their team is doing and what they should ask about. A manager who mentions the programme in a one-on-one is worth more than a launch campaign.
6. Remove every step before the first lesson. Single sign-on, pre-enrolment by group rather than self-registration, one link. Count the clicks between the announcement email and the first exercise, then remove half of them.
These are our own programmes, and the numbers are published on the case studies rather than kept in a deck. At Jaarbeurs, 96% of 350 learners completed the programme, with skill growth of 16% measured before and after. At Skyscanner, 77% across more than 1,400 employees. At Jumbo, between 69% and 92% depending on the course, with a group of buyers who were openly sceptical about using data at all.
Set that against 12%. The point is not that our courses are more interesting. It is that a programme designed around rhythm, relevance and follow-up finishes, and a catalogue of videos does not.
There is a budget argument in here too. If you compare suppliers on cost per seat, the cheapest licence usually wins. If you compare on cost per learner who actually finished, the ranking inverts, often by a factor of five or more. That is the number worth putting in front of a CFO, and it is also the one that tells you whether to renew.
If you want to know where your own programme leaks, start by asking for completion per module rather than per course. The module where the line falls off a cliff is almost always the first one that asked people to do something. That is where the fix belongs.
