TL;DR. Learning transfer often fails because there is no planned practice phase between understanding something and being able to apply it. Reminders, action plans and peer partnerships keep the content fresh, but they do not let people repeat the behaviour itself. The evidence is also less conclusive than much of the advice suggests: well over 80 percent of transfer studies measure what learners say they applied rather than what they actually did (Thalheimer, 2020). This article looks at what the research supports, what it does not, and the three questions every transfer programme needs to answer.
Key Takeaways
- Well over 80 percent of studies on training transfer measure learner perceptions rather than observed behaviour on the job (Thalheimer, 2020, research review), a weakness also named by Ford, Baldwin & Prasad (2018).
- The widely repeated claim that only 10 percent of training reaches the job traces back to a rhetorical question in Georgenson (1982), not to a study. Robert Fitzpatrick reconstructed the citation chain.
- Implementation intentions, meaning concrete if-then plans, produce the largest and most consistent effects in the literature (Gollwitzer & Sheeran, 2006), tested on transfer by Friedman & Ronen (2015). Goal setting on its own shows only small effects (Blume, Ford, Baldwin & Huang, 2010).
- At Schwäbisch Hall, the group that practised with Sleak scored 72.5 percent against 61.4 percent for the course-only group, an improvement of 11.1 percentage points after two conversations.
- In the same rollout, the gap between what participants knew and what they could apply fell from 32.9 points to 13.9.
Ask a sales manager what one person on the team has done differently since the last training. The answer will usually be a completion rate, a satisfaction score or a general comment about engagement. What you rarely hear is a specific change in behaviour. That is not the fault of the participants or the trainer. It is a structural problem: a workshop can explain a method to twelve people in a day, but it cannot give all twelve people twenty opportunities to practise it.
Other high-performance professions take a different approach. Pilots use simulators, surgeons rehearse, and musicians repeat a passage until they can perform it under pressure. Commercial and leadership skills, by contrast, are often explained once and then tested for the first time in a real conversation with a customer or employee. This article explains what learning transfer actually means, what the research does and does not support, where the well-known 10 percent figure came from, and what a practice phase needs in order to change behaviour rather than simply refresh memory. One important caveat is often missing from vendor content: decades of transfer research have not identified a single factor that dramatically increases transfer. Claims to the contrary should be treated with caution.
Why does so little of what people learn show up in what they do?
So little of what people learn shows up in what they do because knowing a method and performing it under pressure are different capabilities, and only one of them gets trained. Baldwin & Ford (1988) framed transfer as the product of training design, trainee characteristics and work environment, and that framing still holds. In practice it collapses into one break point: the content is delivered, the behaviour is never repeated.
Three things separate knowing from doing. The first is recall under pressure: an argument that makes perfect sense in the classroom may not come to mind during a live conversation. The second is responding to the unexpected, because no playbook can cover every moment when the other person goes off script. The third is correction. Without specific feedback on what someone has just done, bad habits can become established just as easily as good ones. Without all three, people may be familiar with a topic, but they are not yet competent in applying it.
Most transfer programmes address only the first issue by keeping the material fresh. But remembering the material was never the main constraint.
What does the research on training transfer actually support?
Very little, and what it does support is narrow: concrete if-then plans work reliably, and most other levers are either weak or poorly measured. The meta-analysis by Blume, Ford, Baldwin & Huang (2010) remains the reference point on which factors matter, and it finds only small effects for goal setting on its own. Will Thalheimer's 2020 research review states the position plainly: research can point with confidence to only a few factors that support or harm transfer, and it has not yet found factors that produce large changes.
There is a measurement problem underneath all of it. Thalheimer estimates that well over 80 percent of transfer studies do not measure transfer at all, they measure learners' perceptions of transfer. Ford, Baldwin & Prasad (2018) name the same weakness in their review of the field. The research and the practice share one habit: both ask people whether they applied something instead of watching them do it.
| Lever | What the evidence supports | What it does not do |
|---|---|---|
| Implementation intentions (if-then plans) | Large, consistent effects (Gollwitzer & Sheeran, 2006), tested on transfer by Friedman & Ronen (2015) | Plans the trigger for the behaviour, never rehearses the behaviour |
| Post-training goal setting | Small effects (Blume et al., 2010) | Says nothing about whether the person can perform the action |
| Manager support | Supported, but measured inconsistently across studies (Blume et al., 2010) | Does not scale past the hours a manager actually has |
| Reminders, nudges, learning nuggets, peer partnerships | Rarely tested against observed behaviour | Keeps content available, repeats no behaviour |
In short, the best-supported approach plans when the behaviour should be applied, but does not measure the behaviour itself. That gap exists in both research and practice, which helps explain why so much advice about transfer sounds more certain than the evidence allows.
Why do the standard transfer tactics fall short?
The standard transfer tactics fall short because they are addressed to memory, and behaviour change is not a memory problem. Open any five articles on improving learning transfer and the same list appears: pre-training alignment, action plans, realistic case studies, microlearning, spaced reminders, peer learning, manager check-ins, follow-up sessions. Every item is reasonable. Together they build a system that reliably produces people who can describe the method when asked.
Yet three questions remain unanswered, and they determine whether a transfer programme changes anything in practice. Who observes the behaviour after the training? How often? And against which standard? Even prominent guidance on learning transfer rarely answers them. The manager is usually named as the observer, but without a defined frequency or set of criteria.
That is a capacity constraint, not an oversight. Observation costs time from exactly the people who have least of it. Before Sleak, one-to-one practice sessions at Energieversum could run to 90 minutes each, with a team lead playing the customer, watching the performance and then giving feedback. At Schwäbisch Hall, a typical group session had around ten participants and one coach: while one person practised, the others watched, and each might get a single attempt. No transfer concept survives contact with that arithmetic.
Where does the "only 10 percent transfers" figure come from?
The 10 percent figure comes from a rhetorical question, not from data. In 1982, David L. Georgenson asked readers how often they had heard training directors estimate that only around 10 percent of content shows up as behaviour change on the job. He cited no evidence and named no source, and his article contained no dollar figure at all. Robert Fitzpatrick traced the citation chain and showed how later authors converted the attention-grabbing question into a finding, eventually attaching a 100 billion dollar loss that Georgenson never wrote. Will Thalheimer has documented that reconstruction publicly.
This matters because figures like this influence budgets. An organisation that believes 90 percent of its training is lost may invest in transfer services simply to limit the damage. Once it knows that the figure has no empirical basis, it can ask a more useful question: what exactly should people do differently after the programme, and how could that change be observed?
A simple rule follows: any transfer statistic presented without a study, a sample and a measurement method should be treated as a starting point for discussion, not as evidence. That also applies to figures that support the case for practice.
What does a practice phase that secures transfer look like?
A practice phase that secures transfer repeats the behaviour under realistic conditions before it counts, and scores it against a standard defined in advance. This is not another workshop. It is a phase between the explanation and the live situation in which the action happens several times, each time with feedback, each time measured the same way.
The standard is the part nearly every transfer model leaves out. Without a defined evaluation standard, repetition produces no information, because nobody can say whether the second attempt beat the first. At Sleak that standard is called a Scorecard, or Standard of Excellence: an evaluation rubric a leader defines up front, against which every conversation is scored. An Initiative is the development goal a leader sets, covering what a team has to know and what it has to be able to do.
The practice itself runs in Training Mode. A person picks a scenario, holds a voice-based practice conversation with a virtual counterpart, and afterwards receives an evaluation from the AI Coach citing evidence from the transcript rather than general praise. A Development Program sequences those scenarios into a path with levels and tasks. Over 600 organisations and more than 76,000 conversations have run on the platform, counted across every organisation on it, including internal, demo and trial accounts.
For L&D or leadership development, the key benefit is not the technology itself. It is that observation no longer depends on finding time in several people's calendars. Ten people can each practise ten times without requiring ten observers.
What do the numbers from real rollouts show?
Two practice conversations were enough to cut the gap between knowing and doing by more than half. In a controlled comparison at Schwäbisch Hall, one group worked through an adaptive learning course, and a second group did the same course plus two practice conversations with Sleak. In the transfer test the practice group scored 72.5 percent against 61.4 percent, an improvement of 11.1 percentage points. The gap between what participants knew and what they could actually apply fell from 32.9 points to 13.9.
The shape of the curve matters more than any single figure. Scores rose 14.1 points between the first and the second practice conversation, then a further 6.5 points in the final simulated customer conversation. The gain arrives early, and it arrives from repetition rather than from more content. This is one organisation's rollout, not a meta-analysis, and it should be read as a directional result rather than a general effect size.
Volume is where the change is most visible. Before Sleak, many participants at Energieversum reached their final onboarding day having practised between one and ten sales conversations. Today a representative completes around 35 on average. Weekly one-to-one sessions that used to take about an hour now take roughly 15 minutes, because the repetition no longer has to happen inside the session. More than nine in ten users at Schwäbisch Hall said the training was useful and that they wanted to keep training with it.
Isn't this just another post-training nudge?
A practice phase differs from a post-training nudge because the behaviour is repeated and scored against a standard, rather than recalled. A reminder email and a practice conversation are not the same category of intervention: one addresses memory, the other addresses performance. Apply the three questions from earlier and the distinction is immediate. Every conversation is observed, the frequency is set by the person practising, and the standard exists before the first attempt.
The honest limit belongs here too. A practice phase does not fix a problem that lives in the work environment. If the new behaviour is punished in daily work, if process or incentives pull the other way, or if a manager rewards the old approach, no amount of rehearsal changes the outcome. That is the part of the Baldwin & Ford model called work environment, and it stays a leadership responsibility.
And the concession: the classroom workshop is still the right place for first understanding, for mindset, for debate, and for everything that needs people in a room together. When a new methodology is introduced, that is where it should be introduced. It is simply not the place where it gets mastered.
FAQ
What is learning transfer?
Learning transfer is the application of knowledge and skills from a learning situation to actual work behaviour. The second half is what counts: not recall of content, but changed action. The modern framing comes from Baldwin & Ford (1988), who described transfer as the result of training design, trainee characteristics and the work environment.
How do you measure learning transfer properly?
Through observed behaviour scored against criteria defined in advance. Participant surveys measure self-perception rather than transfer, which is the methodological weakness shared by well over 80 percent of studies in the field (Thalheimer, 2020). A workable internal sequence is a rubric first, then an occasion to observe, then a repetition.
How many practice repetitions are needed?
The Schwäbisch Hall rollout shows the largest jump between the first and second conversation, with a further gain by the third. Two to three attempts at the same situation is a realistic starting point. There is no universal number, and it scales with the complexity of the conversation.
Does digital practice replace classroom training?
No. Classroom training remains the place for first understanding, mindset and discussion. The practice phase comes afterwards and closes the gap a single training day structurally leaves open, because no single day lets every participant repeat the behaviour enough times.
Related reading
- Completion Is Not Competence: Why Every L&D Metric You Report Is Wrong
- Scorecard-Based Coaching: How Structured Feedback Transforms Sales Training
- Initiatives, Not Courses: A New Organizational Model for People Development
- The Coaching Gap: Why Managers Are Accountable for Outcomes They Have No Tools to Produce
