Key takeaways
- The first question costs almost all of the completion. Courses with 4 or more questions completed at 71.6% against 72.3% for courses with 1 to 3.
- A course without questions counts as complete when a learner clicks to the end, so its 90.7% and a graded course’s 72.3% do not measure the same thing.
- Most learners who fail come back. Of 783,646 learner-course combinations with at least one failed attempt, 70.1% ended in a pass.
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 question count analysis covers the 45,214 courses with at least 5 learner results.
The three tables below draw on different populations. The question count table covers 45,214 courses. The drop-off table counts 321,731 individual learners who did not complete a course. The retake figures count 783,646 learner-course combinations that contained at least one failed attempt.
Full method sits in the 2026 e-learning benchmark report.
What the data says about questions and completion
The key point. Completion falls sharply between zero questions and one, then stops falling.
| Questions in the course | Courses measured | Completion rate, average per course | Completion rate, all learners combined |
|---|---|---|---|
| 0 questions | 11,413 | 90.7% | 94.5% |
| 1 to 3 questions | 9,865 | 72.3% | 84.7% |
| 4 or more questions | 23,936 | 71.6% | 83.8% |
18.4
Adding assessment cost 18.4 percentage points of reported completion, from 90.7% with no questions to 72.3% with 1 to 3. Every question after the first cost 0.7 points in total.
Easygenerator platform data, 45,214 courses, June 2025 to June 2026.
Adding a second, fifth, or twentieth question costs almost nothing, which means the count is not the variable worth arguing about.
Note also that 23,936 of the 45,214 courses carry 4 or more questions. The most common choice on the platform is the one with the most assessment, not the least.
Why the drop is not abandonment
The key point. In courses with questions, most learners who fail to complete have already finished the content and missed the pass mark.
We looked at where learners were last seen when they did not complete a course.
| Courses with no questions | Courses with questions | |
|---|---|---|
| Non-completing learners measured | 41,698 | 280,033 |
| Share who reached the end of the content | 53% | 80% |
| Average point reached by those who left mid-course | 79.7% through | 60.8% through |
Four in five non-completers in a graded course worked through everything and then missed the bar. In a course without questions, only about half of non-completers got that far, so nearly half walked away partway through and never returned.
What the numbers actually say
A course with no questions counts as complete the moment a learner clicks to the end, because there is no bar to clear. The 90.7% figure records arrival. The 72.3% figure records performance.
One detail runs the other way. Among learners who did leave mid-course, those in question-free courses got further before they stopped, 79.7% of the way through against 60.8%.
Knowledge check versus quiz, and why it changes your numbers
The key point. A knowledge check without a pass mark cannot lower your completion rate. A graded quiz can.
A knowledge check asks a question to prompt recall or to break up content. It carries no pass mark and no consequence, so a learner who answers wrong still completes the course. A quiz or assessment carries a score and a threshold, and a learner who falls below it does not complete, however much content they worked through.
That distinction explains the entire 18.4 point gap. The courses in our zero-question group have nothing to fail. The courses in the other two groups do.
“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.
That use of a question is ungated. It exists to re-engage a learner partway through, and it carries no consequence for completion.
So the practical decision is not whether to include questions. It is whether a specific question should gate completion. Use an ungated knowledge check where the goal is attention and recall. Use a graded assessment where somebody needs proof that a learner met a standard.
What happens after a learner fails
The key point. Most learners who fail an assessment retake it and pass.
Among 783,646 learner-course combinations where a learner failed at least one attempt, 70.1% eventually passed. Another 23.0% failed and never returned.
70.1%
Of learners who failed at least one assessment attempt on the Easygenerator platform, 70.1% eventually passed and 23.0% never returned for a second try.
Easygenerator platform data, 783,646 learner-course combinations with at least one failed attempt, June 2025 to June 2026.
A failing grade acts as a pause rather than an exit for roughly three learners in four. The 23.0% who never come back are the group worth attention, and this data cannot say why they stopped. Three explanations fit and they need different fixes. A retake that requires a manager to reset it is a process problem. A retake that nobody tells the learner about is a communication problem. A learner who gave up is a design problem.
Check which applies in your own setup before you change the assessment.
What assessment does for learning, which completion cannot show
The key point. Answering a question improves what a learner retains weeks later, which is the benefit a completion rate cannot record.
Henry Roediger and Jeffrey Karpicke tested this directly. Students studied prose passages and then either took immediate recall tests or restudied the same material. On a final test after 5 minutes, restudying won. On final tests after 2 days and after 1 week, the students who had been tested retained substantially more, even though the students who restudied were more confident they would remember. The paper appeared in Psychological Science in 2006.
That result sits directly against the completion figures above. Assessment costs 18.4 points of reported completion and improves retention at the same time, which is why a team that removes questions to lift a dashboard number makes the training worse while making the report look better.
How many questions should a course have?
The key point. The count barely matters after the first question, so decide what the questions are for instead. The platform data does not tell us the right number by purpose, so the table below combines the completion findings with what each kind of assessment has to achieve.
| What the assessment is for | Questions | Pass mark | What to report |
|---|---|---|---|
| Keeping attention through longer content | 1 to 2 per section | None | Completion rate, which stays near 90% |
| Checking recall before a learner moves on | 3 to 5 | None, or a low one with unlimited retakes | Completion rate and average score |
| Proving competence for a manager or regulator | 8 to 12 | 80%, or as the regulation requires | Pass rate and end-of-content reach, separately |
| Diagnosing what a group does not know | 10 or more | None | Score distribution per question |
The row that causes trouble is the third one. Teams often apply a regulatory pass mark to a course that only needed a knowledge check, then report the resulting completion rate as though it measured the same thing as an ungated course.
Once you have decided what an assessment is for, our guides cover how to write each format, including the hotspot question, the true or false question, and the fill in the blanks question.
What to report instead of a single completion rate
The key point. Report two numbers, because completion alone cannot separate a learner who quit from a learner who tried and missed.
Report the share of learners who reached the end of the content. Report the share who passed. The first tells you whether the course held attention. The second tells you whether anyone learned enough to clear the bar.
A dashboard column holding 90.7% for an ungated course next to 72.3% for a graded one makes the weaker course look stronger. Splitting the number removes the problem, and in graded courses 80% of the apparent failures were people who finished. Compliance programs that need documented evidence of competence depend on that second number rather than the first.
Three mistakes that distort assessment data
Cutting questions to lift the completion rate. The rate rises because the bar disappears, and the retention benefit disappears with it.
Comparing graded and ungated courses in one report. They measure different things, and the ungated course wins every time.
Adding a pass mark by default. A threshold on a course that nobody has to certify creates failures with no purpose, and it removes your ability to read the completion rate.
What this data cannot tell you
These are observational comparisons, not experiments. Nobody assigned authors to add questions. Courses with assessment tend to cover regulated or technical subjects, and they appear more often in mandatory programs.
The question count grouping ignores pass marks. Our zero-question group has no assessment at all. The other two groups mix ungated knowledge checks with graded assessments, so the 18.4 point gap covers both.
Drop-off points are estimated. We take the last course section a learner visited and divide its position by the total number of sections. Section order comes from the order in which authors created the sections, which does not always match the order a learner moves through, since some courses let learners jump between sections.
The recommended question counts are inference. The platform data does not segment completion by assessment purpose.
Completion and pass rates are not learning. Our data says nothing about what a learner remembered a month later. The Roediger and Karpicke result speaks to retention, and it comes from laboratory study rather than workplace training.
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.
