Ebbinghaus forgetting curve: what it is and how to improve retention

You train your team, the course gets completed, the satisfaction surveys come back fine. A month later, nobody remembers the procedure. It isn’t a lack of interest and it isn’t a problem with your content: it’s how human memory works, and someone measured it almost 140 years ago.

The Ebbinghaus forgetting curve is probably the most cited and worst understood concept in corporate training. In this article you’ll see what the original experiment actually says, why the percentages going around don’t describe your employees, and which evidence-backed strategies really do flatten the curve.

What is the Ebbinghaus forgetting curve?

The forgetting curve is the graphical representation of how retention of something learned decays as time passes without review. Its shape is very characteristic: a steep drop in the first hours and days, followed by a progressive flattening. Whatever survives the first week tends to stick around for quite a while.

It was described by Hermann Ebbinghaus, a German psychologist, in his work Über das Gedächtnis (“On Memory”), published in 1885. It was the first serious attempt to measure forgetting experimentally rather than describe it philosophically, and that’s why it’s still a reference.

The core idea it contributes is simple and still holds: forgetting isn’t a failure, it’s memory’s default behavior. Without intervention, what’s learned is lost. Designing training while ignoring that is like filling a leaky bucket.

The original experiment: what Ebbinghaus actually measured

This is where it’s worth reading the fine print, because it explains almost every misunderstanding that followed.

Ebbinghaus worked with a single experimental subject: himself. And what he memorized wasn’t concepts, procedures or skills, but lists of nonsense syllables along the lines of “WID”, “ZOF” or “KAJ”. He chose them that way deliberately, precisely to strip out any prior meaning, any context and any chance of associating the material with something he already knew.

To measure forgetting he didn’t use a test, but the savings method: he calculated how much less effort it took to relearn a list he had already studied. That “saving” is the percentage you see on the curve.

The reference modern replication was published by Jaap Murre and Joeri Dros in PLOS ONE in July 2015. They repeated the experiment with the same method, around 70 hours of learning spread over more than two months, and again with a single subject. Their conclusions are interesting both for what they confirm and for what they qualify:

• The shape of the curve replicates. The fast initial decay followed by flattening is real.
• At 31 days the measured saving was 4.1%, a considerably harsher figure than the original experiment’s.
• The curve isn’t smooth. There’s an upward jump at around 24 hours. The sleep explanation is plausible, but the authors themselves caution that it isn’t established yet for this type of experiment.

The percentages everyone quotes (and why they aren’t yours)

Search for “forgetting curve” and you’ll find the same paragraph repeated across dozens of sites: we forget roughly half in an hour, 70% in a day and 90% in a week. They’re presented as if they described someone who has just finished your workplace safety course.

They don’t, and the difference matters when you’re making decisions:

Does that mean the forgetting curve is useless? Quite the opposite. It means you have to use it as a mental model, not as a data table. What’s actionable in Ebbinghaus’s finding isn’t any specific percentage, but three things:

• Forgetting is fast at the start. The useful window for intervening opens in days, not months.
• Forgetting is inevitable without reinforcement. No course, however good, holds up on its own.
• Every review flattens the curve. It doesn’t just recover what was lost: it makes the next drop slower.

And there’s an uncomfortable corollary: the more abstract and less applicable your content is, the more it resembles Ebbinghaus’s syllables and the faster it will be forgotten. The best defense against the forgetting curve starts with the design of the content, not with the review schedule.

What it really means for corporate training

Translated into concrete decisions, the forgetting curve calls into question the most widespread training format in companies: the single, intensive course, consumed in one sitting and never revisited.

That format is convenient to manage —you plan it, you deliver it, you close the file— and it’s precisely the worst possible one from a retention standpoint. It concentrates all the learning at the moment the curve starts to fall, and then never touches it again.

There are three practical implications that tend to get overlooked:

• The completion rate doesn’t measure anything relevant. Someone completing a course tells you they consumed it, not that they remember it or apply it. If that’s your main indicator, you’re measuring the highest point of the curve and ignoring everything else.
• Annual compliance training is an extreme case. Twelve months between reviews is, in practice, starting from scratch every time.
• Timing matters as much as content. Training in November on something that will be used in March guarantees you’ll have to train again in March.

None of this requires throwing out your existing courses. It requires adding a reinforcement strategy that today, in most training plans, simply doesn’t exist.

How to flatten the forgetting curve: 6 evidence-backed strategies

These are the interventions with solid evidence behind them, ordered by impact relative to the effort they cost.

1. Spaced repetition

This is the best documented strategy of all. Instead of concentrating the learning, you spread it across sessions separated in time. The meta-analysis by Cepeda, Pashler and colleagues, published in Psychological Bulletin in 2006, gathered 839 effect estimates from 317 experiments and confirmed something key: the optimal interval between reviews grows as the retention horizon grows.

In other words: there is no such thing as “the” reinforcement schedule. It depends on when you need the person to remember.

2. Retrieval practice

Forcing people to retrieve information from memory retains far more than re-exposing them to it. Roediger and Karpicke demonstrated this in 2006 with an elegant design: after five minutes, rereading the material produced better results than having taken a test; but after two days and after a week, the result clearly reversed in favor of those who had been tested.

It’s a cheap change with a big effect: replace the end-of-module summary with three questions that force people to answer without looking. Reinforcement activities serve exactly this function.

3. Microlearning and on-demand content

Microlearning fits the curve naturally: short, repeatable units you can look up at the moment of need. A three-minute learning pill someone reviews before a client call is worth more, in retention terms, than an hour-long course consumed months earlier. Microlearning videos and mobile learning are especially useful for teams that don’t work in front of a computer.

4. Reinforcement applied on the job

Nothing flattens the curve like using what you’ve learned. When knowledge is put into practice, each use works as a spontaneous review. Hence the value of simulation training and realistic scenarios: they shorten the distance between learning and applying, which is where most of the learning is lost.

5. Interleaved learning

Alternating related topics instead of exhausting them one at a time makes studying harder in the short term, but improves long-term retention. It’s counterintuitive and it works. We cover it in depth in our article on interleaved learning, or interleaving.

6. Continuous formative assessment

Formative assessment turns measurement into part of the learning: every check is both retrieval practice and a signal of where reinforcement is needed. It’s the strategy that connects the previous five and lets you decide, with data, what’s worth insisting on.

Designing content as reusable learning objects makes all of the above vastly easier: it lets you recombine pieces for reviews without rebuilding the whole course.

Reinforcement schedule: when to place each review

The practical question is when to place the reinforcements. The study by Cepeda and colleagues published in Psychological Science in 2008, with more than 1,350 participants, gives the most useful rule there is: the optimal gap between sessions grows with the horizon, but shrinks as a proportion of that horizon. The authors measured an optimum of between 20% and 40% of the horizon when the test was a week away, dropping to between 5% and 10% when it was a year away.

Translated into a training plan:

Two warnings about this table. First: these are starting points, not a law; complex content and content that isn’t used daily need more reinforcements. Second: a reinforcement only counts if it forces retrieval. Resending the course PDF isn’t a reinforcement, it’s an attachment.

Common mistakes when applying the forgetting curve

• Quoting the percentages as if they were yours. Using “90% in a week” in an internal presentation works as a selling point, but it sets a false expectation that the real data won’t support later.
• Confusing repeating with reinforcing. Showing the same content again creates a feeling of familiarity, which the brain mistakes for knowing. Retrieving from memory is what consolidates.
• Spacing too little. Three reviews in the same week and none afterwards concentrate the effort where it’s least needed.
• Reinforcing everything equally. Not all content deserves the same treatment. Prioritize what’s critical, what’s rarely used and what has consequences if forgotten.
• Measuring only at the end of the course. An immediate assessment measures the peak of the curve. If you don’t measure again at 30 or 60 days, you know nothing about retention.
• Treating it as a content problem. The forgetting curve isn’t solved by making a better course, but by designing what happens after the course.

How to measure retention in your organization

Without measurement, any reinforcement plan is guesswork. The good news is you don’t need to set up a study: it’s enough to change what you look at and when.

• Assess later, not only at the end. A short check at 30 days and another at 90 turns the curve into something observable. Without that second point there’s no curve, just an isolated data point.
• Compare groups with and without reinforcement. If you introduce spaced reviews into a program, leave one group without them for a cycle. It’s the only way to know whether the effect is yours or chance.
• Look at application indicators, not consumption ones. Process errors, incidents, resolution time or interaction quality say more than any completion rate. It’s the difference between the L&D metrics that describe activity and the ones that describe impact.
• Use the platform’s analytics. Learning analytics lets you see which content gets looked up again and which is never reopened, which is a direct signal of where retention is failing. LMS analytics and calculating training ROI rest on those same data.

In long programs, such as an online onboarding, measuring retention at 30, 60 and 90 days is also the fastest way to spot which part of the content is surplus and which needs reinforcing.

Conclusion

The Ebbinghaus forgetting curve describes something true and easy to ignore: without reinforcement, what’s learned is lost, and lost fast. What it doesn’t describe is your employees, because the percentages in circulation come from one man memorizing nonsense syllables in 1885 and from a 2015 replication also run with a single subject. The shape of the curve is an excellent mental model; the figures aren’t. Used well, the curve doesn’t justify training more, but training in a different order: less intensity all at once and more spaced reviews that force people to remember.

The obstacle is almost never pedagogical, it’s operational. Scheduling reinforcements at three days, one month and three months, across dozens of programs and thousands of people, is unworkable by hand. isEazy LMS lets you build learning paths with spaced reinforcements, formative assessments and on-demand content, and then measure whether retention actually improves at 30 and at 90 days. Want to see it with your own programs? Request a demo.