Microlearning gets delivered as short video almost by default. The evidence behind it points somewhere else: what is well established is retrieval practice — being asked to produce an answer or make a decision, repeated across gaps. Three minutes spent deciding is retrieval. Three minutes spent watching is not. That makes video the microlearning format with the weakest claim on its own evidence base. A short scenario, a flashcard set, a knowledge check on last month's process change — all of these are micro, none of them is video, and all of them ask the learner to produce something.
Key takeaways
- Microlearning is treated as a length rule. The evidence is about what the learner is asked to do, not how long it takes.
- Retrieval practice - producing an answer rather than reviewing one - is among the better replicated findings in learning research.
- Spacing those retrieval episodes out adds a large further benefit. A 2021 meta-analysis of 29 studies puts spaced retrieval ahead of massed retrieval at g = 0.74.
- A short video asks the learner to watch. A short scenario asks them to decide. Only one of those is retrieval.
- Micro does not mean video: scenarios, flashcards, knowledge checks and click-to-reveal walkthroughs are all short and all active.
- One scenario is a lesson. A library of them, released over time and returned to, is a practice programme - and the library is the part that compounds.
What the evidence actually supports
Our companion piece sets out where microlearning earns its place in a corporate setting and why the retention statistics attached to it deserve suspicion. This page picks up the part that decides the format: two findings sit underneath short-form training, and they are usually blurred together.
The first is the testing effect. Being asked to retrieve something — to answer, to decide, to produce — leaves a stronger trace than reviewing the same material again. This is one of the more durable findings in the field and it has been replicated across subjects, ages and settings.
The second is spacing. Spreading those retrieval episodes across gaps beats packing them together. A 2021 meta-analysis in Educational Psychology Review pooled 29 studies comparing spaced retrieval practice against massed retrieval practice and found a large benefit for spacing, g = 0.74. Note what the comparison is: both conditions involve retrieval. The 0.74 is what you gain by spreading it out, on top of whatever retrieval itself is already buying you.
Put together, the practical reading is unglamorous: ask people to produce something, and ask more than once, with gaps.
Now hold that against how microlearning is usually implemented. A five-minute video, watched once, marked complete on play-through. It is short. It contains no retrieval and no spacing. It is not a small version of good training; it is a small version of a lecture. That is the case against video as a default, and it is why the rest of this page is about what to build instead.
Three minutes watching
- The learner receives; nothing is produced
- Completion is recorded on play-through
- No signal about whether anything was understood
- Nothing to come back to, because there was no answer to get wrong
- Feels efficient to ship and to consume
- Easy to scale, hard to evidence
Three minutes deciding
- The learner commits to a choice - that is retrieval
- Feedback lands at the moment of the decision, when it is wanted
- The choice is recorded, so you know where judgement is thin
- Worth repeating later, and worth repeating the ones they got wrong
- Takes more thought to build - this is the real cost, and it is a design cost
- Harder to fake having done
Micro does not mean video
Video became the default for microlearning for a practical reason rather than a pedagogical one: it was the cheapest short thing to produce at volume. That constraint has loosened, and the format options worth considering are wider than the category's habits suggest.
A scenario with two or three decision points
A situation the learner recognises, a choice, a consequence, a short debrief. Two or three decisions is usually enough; the cost of a scenario climbs steeply with branch count, because every branch needs writing, review and testing. A tight scenario that takes three minutes is worth more than a sprawling one nobody finishes.
A flashcard set
The most direct expression of retrieval practice there is, and the easiest to repeat at intervals. Unfashionable, and effective.
A knowledge check on the thing that changed
Four questions about the process that changed last month, sent to the people it affects. No course, no module — just the check. This is the highest-yield piece of microlearning most organisations are not doing.
A click-to-reveal or step-through walkthrough
For procedures, where the learner predicts the next step before seeing it. The prediction is the retrieval; revealing the answer immediately after is what makes it stick rather than frustrate.
All four are short. None is a video. All ask the learner to produce something before they are told the answer.
- 1Start from material that already exists
The policy document, the SOP, the incident note, the recording of a call that went badly. The raw material for a scenario is almost always already written down somewhere in the business. A blank page is the expensive way to begin.
- 2Pull an outline from it
An AI-seeded outline turns a document into a first structure - sections, an order, a starting set of questions. Treat it as a draft to argue with, not an answer. It removes the blank-page cost; it does not know which decisions actually matter in your operation.
- 3Decide what the learner has to get right
This is the judgement step and it does not delegate. Which two or three calls, made badly, cause the trouble you are trying to prevent? Those are your decision points. Everything else is context.
- 4Write the wrong answers as carefully as the right one
A scenario is only as good as its distractors. The wrong options should be the mistakes people genuinely make - the plausible shortcut, the over-promise, the escalation that should have happened one step earlier. Obviously wrong options teach nothing.
- 5Add the retrieval check
A short knowledge check after the scenario, or a flashcard pair drawn from it. This is what you will repeat later, so it needs to stand on its own away from the scenario that produced it.
- 6Publish it into the course people are already in
Not a link, not a separate portal. Inside the course, so completion and scores land in the same learner record as everything else and your existing reporting already covers it.
One scenario is a lesson. A library is a programme.
The single scenario is where most teams stop, and it is where the value has not started yet. A library is a different asset, for three reasons.
Coverage of the awkward cases. Any one scenario covers one situation. The situations that cost money are usually the unusual ones — the angry customer, the half-complete handover, the exception nobody documented. A library can hold twenty of those. A course cannot, and a video certainly cannot.
It gives you something to space. This is where the evidence at the top of this page cashes out. With one asset there is nothing to schedule; with twenty, you can release them across weeks instead of assigning them all on day one. The gap between exposures is the mechanism, and a library is what makes the gap possible.
Learners can return to what they got wrong. Because a scenario records a decision rather than a play-through, you know which ones each person fumbled. Those are the ones worth surfacing again a fortnight later. That is spaced retrieval, running on data you already collected.
What the platform has to do for any of this to work
The design is only half of it. Sequenced or scheduled release, so material arrives over time rather than all at once. Per-learner state, so spacing follows the individual rather than the cohort. Reporting that distinguishes completed from got it right, because those are different facts and only one of them is interesting. And delivery on the device people actually have with them.
If the platform cannot do those things, a library of excellent scenarios will still be delivered like a catalogue of videos — assigned once, completed once, never returned to. The design decides whether the content is any good. The platform decides whether the mechanism ever runs.
Frequently asked questions
Is microlearning just short videos?
Short video is the most common format, not a definition. Scenarios, flashcard sets, knowledge checks and step-through walkthroughs are all short-form and all ask the learner to produce something, which short video does not. On the evidence, the active formats have the stronger claim.
What makes a good wrong answer in a training scenario?
The distractors decide whether a scenario teaches anything. Wrong options should be the mistakes people genuinely make - the plausible shortcut, the over-promise, the escalation that should have happened a step earlier. If an option is obviously wrong, choosing against it costs the learner nothing and tells you nothing.
How many decision points should a scenario have?
Two or three is usually right. Cost climbs steeply with branch count because every branch needs writing, review and testing, and completion falls as scenarios get longer. A tight scenario people finish beats an elaborate one they abandon.
Can we do this in Moodle?
Yes. The building blocks - scenarios, flashcards, knowledge checks, click-to-reveal, scheduled release, per-learner tracking - are standard LMS capabilities rather than exotic ones. What varies between deployments is whether release scheduling and per-learner spacing are actually configured, and whether reporting separates completion from correctness.
How many scenarios before it counts as a library?
Fewer than most teams expect. Once you have enough to release over several weeks rather than assign in one go, the scheduling benefit starts, and that is the point at which a set of scenarios behaves differently from a single lesson. Start with the situations that generate the most avoidable mistakes and add as they surface.
How is this different from drip content?
Drip content is the delivery half - releasing material on a schedule instead of all at once. This is that plus the other half: making each released item something the learner has to answer rather than watch. Drip without retrieval is just a slower catalogue.
Where to start
Pick the process that generates the most avoidable mistakes, and write one scenario about the decision people get wrong in it. Not a course — one scenario, three minutes, two decision points, honest wrong answers. Put it in the course those people are already enrolled in and look at what they chose.
That one artefact will tell you more than a completion report for a module ever has, and it is the first item in the library.
Related reading: where microlearning earns its place in corporate training — the companion to this piece — plus why completion is not capability and building SCORM lessons inside the LMS.
Source
Latimier, A., Peyre, H., & Ramus, F. (2021). A Meta-Analytic Review of the Benefit of Spacing out Retrieval Practice Episodes on Retention. Educational Psychology Review, 33(3), 959–987. The g = 0.74 figure is that paper's pooled effect for spaced versus massed retrieval practice across 29 studies. No other figure in this post is a statistic, deliberately — the retention percentages that circulate in this category do not trace to sources we were able to verify.
Want to build one scenario and see what your team actually chooses?
Book a Free Demo