There is a figure that almost never comes up in conversations about training budgets, and it should open them. Robert Brinkerhoff's research on the transfer of training to the job established a distribution that gave rise to the term scrap learning: about 20% of participants apply nothing they learned, 65% try but slip back into their old habits, and only 15% actually change the way they work.
CEB, now Gartner, offers a more conservative estimate: 45% of what is delivered is never applied. Take the more cautious of the two figures and the problem is still enormous.
It's not that training doesn't work
It's that the dominant format is designed in a way that guarantees this result.
The one-off, lecture-based event, with no practice and no reinforcement, produces exactly what the evidence describes. The waste is not an accident of the method: it is its predictable consequence.
A longitudinal study published in Human Resource Management tracked the real-world application of what was learned in corporate training. 62% applied what they had learned immediately after the course. After six months, 44%. After a year, 34%.
The certificate, on the other hand, does not decay. It is just as valid in the file twelve months later as on the day it was issued. That is the structural gap in the system: the record documents a past event, not what is actually being done in the field.
Now the arithmetic
ATD's State of the Industry 2025, based on 539 organizations, reports $1,054 in direct learning spend per employee and 13.7 hours of training per year. The cost per learning hour reached $165, up 34% from $123 the previous year.
Cross that cost with the transfer data and you get the real size of the problem: if between 45% and 80% of that investment never turns into behavior, between $470 and $840 per employee per year is evaporating. Across a workforce of 300 people, that is a number no finance manager would let slide on any other line of the budget.
The 34% increase in a single year also changes the arithmetic of the decision. When a traditional training hour was cheap, the waste was tolerable. At this price, the cost of not measuring it has overtaken the cost of measuring it.
The easy way out that makes the problem worse
Faced with this diagnosis, the usual reaction is to go digital: upload the content to a platform and let everyone move at their own pace.
An analysis published in the International Review of Research in Open and Distributed Learning reviewed 221 massive open online courses, with completion data available for 129. The median completion rate was 12.6%, ranging from 0.7% to 52.1%.
This is not a content problem or a platform problem: it is structural to the self-paced format with no support and no consequences. When nobody expects anything on the other side of the screen, the course competes with everything else on the computer. Uploading PDFs to a platform is not digitizing training: it is digitizing dropout.
What does transfer
The right question is not how to train more cheaply. It is which format produces behavior, and how you prove it.
Three conditions show up consistently in the transfer literature: repeated practice instead of one-time exposure, scenarios that resemble the real task instead of generic content, and an objective record of performance instead of satisfaction surveys.
A simulated exercise meets all three by design. It can be repeated at no meaningful marginal cost, it is built on the plant's actual procedures, and with no extra work it produces a record of what each person did, how many attempts they needed, and where they made mistakes.
That does not make the simulator the only answer. It makes the waste measurable, which is the step almost no one has taken.
The question for your next budget meeting
It is not how much the training program costs. It is what percentage of that program is turning into behavior, and how you know.
If the answer is a satisfaction survey, you already have your diagnosis.
Sources: Brinkerhoff, cited by ATD — scrap learning · ATD — State of the Industry 2025 · Saks & Belcourt — Human Resource Management (Wiley), 2006 · Katy Jordan — IRRODL, 2015