People forget most of what they're taught, fast, unless something deliberate is done to make it stick. Two methods reliably fix it, and both are decades old and well-supported by research: retrieval practice (testing yourself on the material instead of rereading it) and spaced repetition (reviewing at widening intervals). Everything else in this article is detail in service of those two ideas.
First, though, the famous number deserves a closer look, because the way it's usually quoted is misleading.
The "90% in a week" claim, told straight
You've seen it: "employees forget 90% of training within a week." It gets attributed to the Ebbinghaus forgetting curve, and it's repeated in roughly every L&D blog post on the subject.
Here's the accurate version. In the 1880s, German psychologist Hermann Ebbinghaus ran memory experiments on himself, memorizing lists of nonsense syllables and measuring how quickly he forgot them with no review. His curve showed steep early decay: a large share of the material gone within hours to days, flattening after that. That's a real, repeatedly replicated finding.
But two caveats matter:
- It was nonsense syllables, not workplace training. Meaningful, applied, connected information decays more slowly than random syllables. So the raw curve overstates how fast real, useful knowledge disappears.
- The specific "90% in a week" figure is a popularized approximation, not a precise measured constant. The shape of the curve is well established; the exact percentage you read is a rounded, oft-repeated number, not a law of nature.
So treat "90%" as a vivid illustration of a true pattern, not a precise statistic. The takeaway isn't a number. It's this: without review, retention falls off a cliff early, and most corporate training is designed as a one-time event that never reviews. That design, not a magic percentage, is the problem.
Why one-and-done training fails
The failure isn't that the content was bad or the learner was lazy. It's structural.
- No retrieval. Most training asks people to receive information (watch, read, listen). It rarely asks them to retrieve it from memory afterward. Receiving feels like learning but builds weak, fast-fading memory traces.
- No spacing. Everything happens in one sitting. Memory is strengthened by being revisited over time; a single exposure, however good, doesn't get that.
- No application. Knowledge that's never used in a real task stays inert and detached, and detached knowledge is exactly what the forgetting curve erases fastest. The gap between "learned it" and "did it" is where retention quietly dies.
Completion metrics hide all of this. A 95% course-completion rate tells you people clicked to the end. It tells you nothing about what they can still do next month.
What actually fixes it
Two evidence-backed mechanisms, plus one supporting practice.
1. Retrieval practice (the testing effect)
The single most robust finding in this area: trying to recall information strengthens memory more than re-studying it. Henry Roediger and Jeffrey Karpicke's research is the standard reference. In their studies, learners who were tested on material remembered substantially more later than learners who simply reread it the same number of times, even though rereading felt more productive.
The practical move: replace "watch it again" with "recall it now." A low-stakes question, a flashcard, a "explain it back" prompt, a scenario to work through. The act of pulling the answer from memory is what does the work.
2. Spaced repetition (the spacing effect)
The second robust finding: the same total study time produces far better retention when it's spread out than when it's massed. Reviewing on day 1, day 3, day 8, and day 20 beats four reviews in one afternoon.
Done well, spacing is adaptive. Algorithms like SM-2 (the basis of most spaced-repetition systems) schedule each item's next review based on how well you recalled it last time, so material you find easy is shown rarely and material you struggle with comes back soon. This is what flattens the forgetting curve instead of just slowing it slightly.
3. Application and context
Retrieval and spacing work even better when the recall happens in a realistic context. Pair recall prompts with real tasks, real scenarios, real decisions the learner actually faces. This closes the "learning-doing gap" and gives the memory something durable to attach to.
A practical retention design you can copy
You can apply this to any program, customer academy, onboarding track, compliance course, or partner enablement, without rebuilding everything.
- Define what must be remembered, not just covered. List the handful of facts, steps, and judgments a learner genuinely needs in 90 days. Retention budget is finite; spend it on what matters.
- Turn each into a retrieval item. A question, scenario, or recall prompt, not a slide to reread.
- Stage the difficulty. Start guided (hints, multiple choice), move to supported (fewer cues), then independent (recall cold). Difficulty that ramps as competence grows is where durable memory forms.
- Schedule spaced revisits. Bring each item back at widening intervals, sooner for items the learner missed. Automate this; humans won't keep manual schedules.
- Measure recall, not completion. Track what people can retrieve over time. That's the number that predicts on-the-job performance.
The reason most teams don't do this isn't disagreement, it's tooling. Standard LMS platforms are built to host and track content to completion; they have no native concept of "resurface this item in six days because the learner is about to forget it."
This is exactly what a retention engine handles. Elite Recall is built around it: SM-2 spaced repetition, staged retrieval that moves learners from guided to supported to independent recall, and adaptive revisit loops layered over your existing content, on a branded academy you can stand up self-serve. If you want to design a program around remembering rather than completing, you can start a 14-day trial and test it on a real course.
The bottom line
The "90% forgotten" stat is a dramatized version of a real truth: memory decays fast without deliberate review, and most training never reviews. The fix isn't more content or better videos. It's two old, well-evidenced ideas, retrieval practice and spacing, built into how the program is delivered.
If you're evaluating platforms with this in mind, note that most learning tools, including several popular Skilljar alternatives, optimize for completion rather than retention. That distinction is the one worth shopping on.
FAQ
Is it true employees forget 90% of training in a week? The shape is real, decay is steep without review, but the exact "90%" is a popularized approximation drawn from Ebbinghaus's nonsense-syllable experiments. Meaningful, applied knowledge fades more slowly. Treat it as an illustration, not a precise stat.
What is the most effective way to improve training retention? Retrieval practice (recalling material from memory) combined with spaced repetition (reviewing at widening intervals). Both are strongly supported by research and outperform rereading or re-watching.
Does spaced repetition really work for workplace training? Yes. The spacing effect is one of the most replicated findings in learning science. The main barrier is tooling, since most LMS platforms can't schedule adaptive revisits, not whether the method works.
How is retention different from course completion? Completion measures whether someone finished the content. Retention measures whether they can still recall and use it later. Completion is easy to hit and a poor predictor of performance; retention is harder to achieve and the metric that actually matters.