It's also, as far as researchers have been able to trace it, not a real finding.
For decades, L&D conversations have repeated a specific number: only 10 percent of training investment actually transfers to the job. It shows up in conference talks, consulting decks, and the opening line of countless articles on this exact subject. It's also, as far as researchers have been able to trace it, not a real finding. The figure originated in a 1982 remark by a training consultant, was never based on measured data, and was picked up and re-cited by author after author until it hardened into an accepted fact nobody had actually verified. A 2011 paper by researchers J. Kevin Ford, Stephen Yelon, and Abigail Billington examined the claim's origins directly and found no empirical study behind it at all — the number simply propagated through citation.
That the field's most-repeated statistic about itself turns out to be unsupported is, in a strange way, a better starting point than the number itself would have been. It says something real: this is a research area prone to appealing, memorable claims that outrun the evidence, which is exactly the discipline a Professional-facing piece on the subject should model rather than repeat. The actual, defensible starting point is simpler and less dramatic — transfer varies widely across studies, training types, and how "transfer" gets measured, and no single clean percentage describes it. What the research does establish with more confidence is not how much fails, but why.
This is where transfer of training as a formal research area actually begins. It's been studied since at least 1988, when Timothy Baldwin and Kevin Ford published the framework that most subsequent research still builds on. Their model split the problem into three parts: characteristics of the trainee, design of the training itself, and the work environment the trainee returns to. Nearly forty years later, that third factor is the one most organizations still underinvest in, and it's the one that has received the most sustained research attention since.
Trainee characteristics matter, but they're the part organizations can least control and most tend to over-blame. "Some people just don't apply what they learn" is a real observation and a convenient one, because it locates the failure in the individual rather than in the system around them. The research doesn't support treating it as the primary lever. Ability and motivation predict some of the transfer gap. They don't predict most of it.
Training design matters more, and it's improved. Most competent programs now build in practice, feedback, and some attempt at real-world relevance rather than pure lecture. This is progress. It's also not sufficient on its own, which is the part that tends to get lost in program design conversations: research on the Baldwin and Ford model suggests that even well-designed training, delivered to motivated people, does not reliably produce durable transfer when the third factor — the environment the trainee returns to — fails to support it.
Researchers have a more formal name for this environmental factor: transfer climate. It refers to the cues, consequences, and constraints in the work environment that either support or block someone using what they learned. A supportive transfer climate includes things like opportunities to practice the new skill soon after training, resources to actually do it, and social signals — from peers as well as managers — that the new behavior is valued rather than a distraction from real work. A hostile or simply indifferent climate can substantially undermine even well-designed training, because the employee returns to a system with no room, no reward, and no visible interest in whether anything changed.
That third factor is where a substantial body of evidence concentrates, and manager (or supervisor) support is one of its most consistently documented findings. Not manager approval of the training budget. Manager behavior after the session ends — whether a supervisor asks about what was learned, creates any opportunity to apply it, or simply notices when an employee tries and either reinforces or ignores the attempt. A 2010 meta-analysis by Brian Blume, Kevin Ford, Timothy Baldwin, and Jason Huang, synthesizing 89 empirical studies, confirmed supportive work environment factors — including supervisor support specifically — as a significant, positive predictor of transfer. Several subsequent reviews building on that body of work describe supervisor support as among the strongest organizational-level predictors identified in the literature.
Consider two composite, illustrative scenarios built from the general pattern this research describes, not a specific study or a reported effect size. Two employees attend the same negotiation-skills workshop, delivered by the same trainer, on the same day. One returns to a manager who asks, a week later, how the first negotiation using the new approach went, and follows up again after the next one. The other returns to a manager who never mentions the training again. The direction the research consistently points to is that the first employee has a meaningfully better chance of still using the new approach months later — not because the training affected them differently, but because one of them had a structure making the attempt visible to someone who reacted to it, and the other didn't.
The mechanism of making a specific attempt visible — and attaching a response to it — connects to a parallel body of evidence at the individual level: Implementation Intentions: The Research That Doubles Goal Achievement explains why specificity of trigger and planned response, rather than general intention, is what distinguishes attempts that hold from attempts that quietly dissolve.
This produces an uncomfortable asymmetry for how organizations typically allocate effort. L&D departments spend enormous energy on instructional design, materials, delivery format, and the training experience itself — all things the department controls directly. One of the most consistently documented levers sits almost entirely outside that department, in manager behavior nobody in L&D is positioned to require or even reliably observe. The training can be excellent and still transfer at low rates because nothing changed about the six weeks after it ended.
There's a second, related pattern worth being precise about, because it often gets flattened into a vague claim about "reinforcement": the research on opportunity to perform suggests transfer isn't primarily about repetition of the training content. It's about opportunity to apply combined with some form of feedback on the attempt. A common pattern described in this literature is that an employee who tries a new approach and gets no signal — positive or negative — from anyone who matters tends to quietly revert to the old method, not because the new one didn't work, but because nothing distinguished trying it from not trying it. The absence of a reaction can function, in practice, similarly to discouragement.
This is where the underlying evidence problem echoes something larger than training specifically. The gap between an insight and a sustained behavior isn't unique to corporate learning — it shows up anywhere someone is expected to carry a new understanding forward without a mechanism for making the attempt visible to anyone, including themselves. Training transfer research essentially describes, in organizational terms, why a specific kind of continuity — someone noticing, someone responding — determines whether a change holds. The mechanism generalizes even though the research literature that names it is specific to workplace learning.
What this suggests for practice is narrower than most transfer-of-training advice usually lands on. It isn't "make training better," though that's rarely bad advice. It's that the highest-leverage intervention available to most L&D functions isn't inside the training room at all — it's a structured, low-effort mechanism for managers to notice and respond to attempted application in the weeks after a program ends. That doesn't require a new training program. It requires making an invisible behavior — did anyone try this, and did anyone notice — visible enough to act on, which is a measurement and evidence problem before it's a training-design problem.
This is also where the transfer-of-training literature intersects with a broader measurement problem that shows up well beyond corporate learning: most systems are built to record whether an event happened — attendance, completion, a signature on a form — rather than whether anything downstream of the event actually changed. A completion record answers "did this occur." It says nothing about "did this get used," and the two questions have very different answers in most organizations, precisely because only one of them is being tracked. Closing that gap doesn't require replacing manager judgment with a new system. It requires giving managers and employees a low-friction way to make the six weeks after training as visible as the training day itself already is.
Most organizations don't have that mechanism. They have a training completion record and, at best, a satisfaction survey collected the same day. Neither one captures whether anything changed six weeks later, on the job, in front of a manager who may or may not have noticed, which means the organization is measuring the training event closely while leaving the factor the research most consistently points to — what happens afterward — almost entirely unmeasured and unmanaged.