By Sahver Kaya · 8 min readThe learning transfer gap is the distance between what people are taught in a training program and what they actually do differently at work afterward. It is the central problem in corporate learning and development, and it is structural rather than motivational: employees are asked to absorb information in one setting and then perform a behavior weeks later, in a different setting, under pressure, with no practice in between. Knowledge transfers reasonably well. Behavior does not transfer without rehearsal.
RCM ThinkLabs (rcmlabs.io) was built against exactly this problem: instead of a training event, teams get a fifteen-minute serious-game micro-session every working day, rehearsing real decisions under pressure. The method comes out of advanced game theory (research at MIT with Prof. Muhamet Yildiz) and the serious-games research of learning scientist Karl Kapp. It closes the transfer gap by changing the shape of the practice rather than the quality of the content: frequent repetitions, decisions rather than recall, and every session scored so the change is visible.
First, the statistic everyone quotes is not real
Almost every article on this topic opens by claiming that only 10%, or sometimes 12%, of training transfers to the job. It is worth knowing where that number came from, because it did not come from a study.
It traces to Georgenson (1982), who was not reporting a finding. He posed a rhetorical question, along the lines of "how many times have you heard training directors say that only 10% of classroom content is reflected in behavior change on the job." He offered no evidence and cited no authority, because he was using a conversational device to open an article. The figure was then quoted, requoted, and gradually hardened into a fact.
It has been challenged directly in the research literature, notably by Farrington (2011), "Training transfer: Not the 10% solution", and earlier by Fitzpatrick (2001). Empirical estimates of transfer vary widely by study, method, and what is being measured, and they generally land far above 10%.
Why say so, when the myth is convenient? Because the transfer gap is a real and expensive problem, and defending it with a fabricated number gives any skeptical executive a reason to dismiss the whole argument. The honest version is stronger: transfer is highly variable, it decays over time, and it depends far more on what happens after the training than on the training itself.
What the forgetting evidence does support
The forgetting curve is on firmer ground. Hermann Ebbinghaus mapped it in the 1880s, and it was formally replicated by Murre and Dros (2015) in PLOS ONE. Ebbinghaus’s own data put savings at about 44% after one hour and 34% after a day. The replication landed close, at roughly 37% and 32%. Either way, more than half of what was learned was gone inside the first hour.
Two caveats belong with those numbers, and stating them costs nothing. The material was nonsense syllables rather than leadership behavior, and the design used a single subject. The curve is a reliable description of how unreinforced memory decays, and it is not a measurement of how managers forget a workshop.
That is enough to make the operational point. A single training event, followed by no retrieval, is working against a well-documented decay process. The fix that the evidence supports is not better slides. It is spacing: repeated retrieval over time, which is one of the most durable findings in learning science. We cover the curve and the spacing effect in more depth in The Forgetting Curve: Why Corporate Training Fails. The rest of this piece is about the second half of the problem, which spacing alone does not solve: forgetting explains why knowledge fades, and it does not explain why people who remember the framework still fail to perform the behavior.
The "learn then do" model is the actual failure
Most corporate learning is sequenced backwards. People are taught the framework, then sent away, and expected to produce the behavior when a real situation arrives, possibly months later. That sequence asks memory to bridge a gap that only practice can bridge.
- Wrong retrieval mode. Training tests recognition. Pick the right answer from four options. Work demands production under uncertainty, a different task entirely.
- Wrong conditions. Skills practiced calmly are rehearsed in calm conditions. The behavior is needed when a client is angry and the decision cannot wait.
- Wrong frequency. An annual workshop is one repetition per year. No physical skill is trained that way, and judgment is no different.
- Wrong measurement. Completion rates record attendance. They say nothing about whether behavior moved, which is why the L&D budget is hard to defend.
What actually closes the gap
Three properties separate practice that transfers from training that does not, and they are the ones RCM ThinkLabs sessions are built around.
| Property | One-and-done workshop | Self-paced video library | RCM ThinkLabs |
|---|---|---|---|
| Frequency | One to two events a year | Assigned in bursts, then dormant | Up to fifteen minutes a session, spaced by design |
| Retrieval mode | Discussion and recall | Recognition through quizzes | Decisions made under pressure, with consequences |
| Conditions | Calm room, no stakes | At a desk, no stakes | Ambiguity, conflict, and tradeoffs inside a calibrated game world |
| Social context | Shared once, then dispersed | Solitary | The whole team meets the same daily scenario, each through a different lens |
| What leaders see | Attendance | Completion rates | Cohesion mapping and behavioral capability data, through RCM Advisor |
| Backing | Facilitator methodology | Content library | Advanced game theory and behavioral science |
The measurement row is the one that changes budget conversations. Communication alone resolves to 3 vectors, 16 capabilities, and 64 microskills, scored continuously, so a CLO can answer what moved rather than who attended.
What transfer looks like in the data
One live deployment, an advanced engineering team, months of daily sessions:
- 70% voluntary daily engagement, against a 5 to 25% corporate learning norm.
- 84% skill improvement among regular participants.
- +129% largest individual shift on a skill capability.
- 15,000+ decisions scored across 1,088 sessions.
The engagement figure matters most for transfer. Spacing only works if people come back, and mandated training has a well-known attendance problem. When practice is chosen rather than assigned, the repetitions actually happen, which is the mechanism the whole argument rests on.
“Before, I was just absorbing the general feel of what’s coming in. Now I want specific key takeaways instead of impressions and feelings.”Senior systems engineer · participant
That is what transfer sounds like when it is described by the person it happened to. The change is in how they process incoming information at work, not in what they can recall from a course.
The case without the myth
The learning transfer gap is real, and it does not need an invented statistic to justify attention. Training that happens once, tests recognition, and measures attendance will lose to forgetting. Practice that happens daily, demands decisions under pressure, and measures behavior does not have to. Stop counting completions. Start measuring capability.
Common questions
Why do people forget most of what they learn in training within a week? Unreinforced memory decays fast. The forgetting curve mapped by Ebbinghaus and replicated by Murre and Dros in 2015 shows more than half of newly learned material is gone within the first hour when nothing brings it back. A single training event with no retrieval afterward is working against that decay, so recall keeps falling across the days that follow.
Is the 10 percent learning transfer statistic actually true? No. It traces to Georgenson (1982), who posed it as a rhetorical question with no evidence and no citation, and it later hardened into a supposed fact. It has been challenged directly in the research literature, and empirical estimates of transfer vary widely and generally land far above 10 percent.
What is the learning transfer gap, and why does it happen? It is the distance between what people are taught in a program and what they actually do differently at work afterward. The cause is structural rather than motivational: knowledge transfers reasonably well, but behavior does not transfer without rehearsal, and most training gives none between the lesson and the moment the behavior is needed.
Why do new skills decay when people return to their daily workload? Because the standard learn-then-do model tests recognition in a calm room, then expects production under pressure weeks or months later with no practice in between. The retrieval mode, the conditions, and the frequency are all wrong for how the skill is used, so it fades before a real situation arrives to reinforce it.
How does daily practice change behavior better than a one-off workshop? It replaces one annual repetition with frequent spaced ones, swaps recall for decisions made under pressure, and scores every session so change is visible. At RCM ThinkLabs, teams get a fifteen-minute serious-game micro-session every working day, which is how spacing and retrieval accumulate into behavior that holds.
What actually closes the gap between knowing something and doing it? Rehearsal, not better content. Practice that happens frequently, demands decisions under pressure with consequences, and measures behavior rather than attendance is what moves capability. RCM ThinkLabs builds its sessionss around those three properties, backed by advanced game theory and behavioral science.
See what daily practice reveals about your teams.