Microlearning statistics, with the studies behind them
Most microlearning statistics you will find are recycled between marketing pages with no study at the end of the chain. These are the findings that come from real papers, and a note on the famous numbers that do not.
Disclosure: Learning Cadence publishes this article and makes Cadence, one of the apps listed. We have a commercial interest in Cadence. Every other app is described from its official pages, linked in each entry, and we say plainly where a different app is the better choice.
If you search for microlearning statistics you will find the same handful of numbers everywhere: retention improves by 50%, completion rates hit 80%, engagement rises 130%. Follow the citations and they lead to another marketing page, then to a statistics aggregator, then to nothing.
The underlying idea is nevertheless well supported, just not by those numbers. There is a large experimental literature on spacing, retrieval, and forgetting going back to the 1880s. This page collects what those studies actually found, with a link to each, so you can cite the real thing.
What the research says
Findings, with the study behind each one.
Every claim below names the paper it comes from and links it. Where a widely repeated number has no traceable source, this article says so instead of repeating it.
01
Forgetting is fastest immediately after learning, then slows down
Ebbinghaus established this in 1885 by memorising nonsense syllables and measuring how much relearning effort was saved at delays from twenty minutes to a month. The curve drops steeply at first and then flattens. A 2015 replication reproduced the original result more than a century later.
Study: Murre, J. M. J., and Dros, J. (2015). Replication and Analysis of Ebbinghaus' Forgetting Curve. PLOS ONE, 10(7).
What it means in practice: Most of what you lose from a lesson, you lose in the first day or two. That is an argument for revisiting material soon after the first exposure, not weeks later.
Spreading study across sessions beats massing it, across 317 experiments
A meta-analysis examined 839 assessments of distributed practice drawn from 317 experiments across 184 articles. It found a reliable benefit for spaced over massed study, and that the gap producing the best retention grew as the delay before the final test grew.
Study: Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., and Rohrer, D. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132, 354 to 380.
What it means in practice: Ten minutes a day for six days beats an hour on Sunday, and the advantage gets larger the longer you need to remember the material.
The best gap between reviews is a proportion of how long you need to remember
More than 1,350 people learned a set of facts, reviewed them after a gap of up to 3.5 months, and were tested up to a year later. Increasing the gap first improved and then reduced final performance, and the optimal gap declined from roughly 20 to 40% of a one-week delay to roughly 5 to 10% of a one-year delay.
Study: Cepeda, N. J., Vul, E., Rohrer, D., Wixted, J. T., and Pashler, H. (2008). Spacing effects in learning: A temporal ridgeline of optimal retention. Psychological Science, 19, 1095 to 1102.
What it means in practice: If you want to still know something in a year, revisit it around a month later. If you need it for something next week, review after a day or two. There is no single correct interval, only one relative to your deadline.
Being tested beats rereading, but only when you measure later
Students read prose passages then either restudied them or took recall tests without feedback. On a final test after five minutes, restudying won. On final tests after two days and one week, the students who had been tested retained substantially more. The advantage of testing is invisible if you measure immediately.
Study: Roediger, H. L., and Karpicke, J. D. (2006). Test-Enhanced Learning: Taking Memory Tests Improves Long-Term Retention. Psychological Science, 17(3).
What it means in practice: A quiz at the end of a lesson is not assessment, it is the part that makes the lesson stick. It also explains why rereading feels more effective than it is: it wins the short-term comparison you can feel.
The forgetting curveEbbinghaus measured how much relearning effort was saved after a delay. This shows the shape of the result and not exact values: the steep part is the first day, which is the argument for revisiting material early.The best review gap scales with your deadlineFrom the 2008 study of more than 1,350 people. These are the two intervals the paper reports, shown as a share of how long the material had to last.
Check before you cite
The microlearning numbers with no source.
Several figures circulate as established fact across training vendor pages and app marketing. We could not trace any of them to a primary study, and we are not going to repeat them.
If you are writing a business case, do not use these without finding the original yourself. A reviewer who checks one citation and finds a dead end will discount the whole document.
Microlearning improves retention by 50%, or by 25 to 60%. Widely quoted, usually attributed to nobody, or to an aggregator page that cites another aggregator.
Completion rates of 80 to 83% against 20 to 30% for traditional courses. The comparison is not like for like even in principle, since a five-minute module and a six-week course are not competing for the same commitment.
Learners retain 145% more, or recall is 150% stronger. Percentage-of-a-percentage claims about memory are a signal to check the source, because retention is usually reported as a proportion recalled, not a multiplier.
Employee engagement rises 130%. Engagement is not a standard measure, so a number without a definition and an instrument attached cannot be interpreted.
The average attention span is now 8 seconds, less than a goldfish. This one has been repeatedly investigated and traced back to no identifiable study.
What it adds up to
What the evidence actually supports.
Taken together the studies support a narrower claim than most marketing makes, and the narrower claim is more useful because it is true.
Short study blocks spread over days beat the same total time in one sitting, and the advantage grows the longer you need to retain the material. Reviewing after a gap works better than reviewing immediately, and the right gap scales with your deadline. Being made to recall something beats reading it again, though you will not feel that benefit for a couple of days.
What the evidence does not say is that short lessons are better for everything, that any specific length is optimal, or that a five-minute module can replace supervised practice. Much of this work was done on verbal recall with student participants, so applying it to a management skill is a reasonable extrapolation, not a proven result.
Where we fit
How this shapes Cadence.
Cadence is our product, so discount this accordingly.
These findings are the reason Cadence is shaped the way it is: lessons of about ten minutes rather than long modules, arranged across days, not offered as a library, with a check inside the lesson, not an optional quiz at the end.
The research above supports a format, not a product. Everything here is equally an argument for a spaced-repetition flashcard app, a textbook read on a schedule, or a course you build yourself. We would rather cite it honestly than imply the studies were about us.
See it in concrete terms
Open a real Cadence course.
If the underlying question interests you, these Cadence courses cover habit design and reading evidence critically.
The underlying mechanisms are well supported. A meta-analysis of 839 assessments across 317 experiments found a reliable advantage for spaced over massed study, and the advantage grows as the delay before testing grows. What is not well supported is any specific headline percentage, since the commonly quoted figures rarely trace to a primary study.
Is it true that microlearning improves retention by 50%?
We could not find a primary source for that figure, or for the similar 25 to 60% range, despite both being quoted widely. They generally trace to statistics aggregator pages citing each other. Use the underlying spacing and retrieval studies instead, which report their methods and sample sizes.
How long should the gap between study sessions be?
It scales with how long you need to remember the material. In a study of more than 1,350 people, the best gap fell from roughly 20 to 40% of a one-week retention interval to roughly 5 to 10% of a one-year interval. Practically, review after a day or two for next week, and after about a month for next year.
Is testing yourself better than rereading?
After a delay, yes. In a 2006 study, restudying beat testing when the final test came five minutes later, but testing beat restudying at two days and one week. This is also why rereading feels more effective than it is: it wins the short-term comparison you can actually perceive.
Can I cite this page in a business case?
Cite the underlying studies and not this page. Each finding above names its authors, journal, and year and links to a publisher or author-hosted page, so you can read the original and quote it directly. That will hold up to scrutiny in a way that citing an app's blog will not.
Last reviewed
How we sourced this
What we checked, and what we did not.
We only included findings we could trace to a named, published study with authors, a journal, and a year, and we linked each one so a reader can check it. Where a number is widely repeated across the industry but we could not find a primary source for it, we say that explicitly rather than repeating it. Descriptions of each study reflect what the paper reports, not a stronger version of it.
Direct checks: We checked the title, authors, journal, year, and headline result of every study cited here, including the sample sizes and the specific intervals reported, and linked a publisher or author-hosted page for each.
Testing limits: We did not run any original research, and we did not re-analyse the underlying data. Most of this work was carried out on verbal learning tasks with student participants, so applying it to workplace skills is a reasonable inference, not a demonstrated result. We are not a research organisation and this is not a systematic review.
Disclosure: Learning Cadence publishes this article and makes Cadence, one of the apps listed. We have a commercial interest in Cadence. Every other app is described from its official pages, linked in each entry, and we say plainly where a different app is the better choice.
Learning Cadence is not affiliated with any other product named on this page. Product names belong to their respective owners. Prices and free tiers change, so open the official page before you buy.
