Imagine a cohort of eighteen newly promoted supervisors learning a three-step feedback model on Monday morning. They can explain it, practice it, and pass the knowledge check. By Friday, most are still postponing difficult conversations because their calendars are packed, their managers have not asked to see the skill, and the team rewards speed over coaching. My position is simple: most training transfer failures are work-design failures before they are memory failures. A course can introduce a skill, but the job determines whether that skill gets used, supported, and maintained.
Training transfer means applying what was learned to the job and continuing to use it over time. Learning happens in the session; transfer happens in the workflow. That distinction explains why a strong post-course score can coexist with little observable change.
The memory argument is right—and incomplete
The strongest opposing view deserves a fair hearing. People forget. A learner may understand a concept on Tuesday and fail to retrieve it during a tense customer call two weeks later. Hermann Ebbinghaus's early experiments made forgetting visible, and later research has consistently supported the value of spacing, retrieval practice, feedback, and repeated application. If a safety course teaches a critical procedure once, buries it in a slide deck, and never lets anyone rehearse it, the course design deserves blame.
There is also a legitimate risk in blaming the workplace for every failed outcome. A learner who cannot perform a skill in a realistic practice session may have a learning problem, not merely a support problem. Poor explanations, vague practice, weak feedback, and badly chosen examples do not become acceptable because a manager is busy. Training has to teach something useful, and learners have to do some of the work.
Still, memory is only one part of the diagnosis. Research on transfer has treated the work environment as a central variable for decades. Baldwin and Ford's 1988 model defined transfer through generalization and maintenance: can people use the skill in situations beyond the lesson, and can they keep using it later? Reviews by Burke and Hutchins, Grossman and Salas, and Blume and colleagues have all pointed to the combined influence of training design, learner motivation, supervisor support, opportunity to perform, and organizational climate.
For anyone responsible for learning transfer in the workplace, the practical distinction is this: did people fail to learn, or did they learn something that the job did not permit them to use? Those require different remedies. More instruction may help the first. A changed workflow, manager expectation, or performance measure may be the answer to the second.
The job decides whether learning survives
Transfer should be designed before the course is booked. Start with one behavior that another person could see or hear, rather than a broad intention such as “communicate better” or “be more strategic.” A customer-support representative might acknowledge the customer's concern, ask one diagnostic question, and confirm the next action before closing the interaction. A project lead might name the decision owner and deadline at the end of every handoff. Those are behaviors. “Improve collaboration” is not.
A useful transfer brief can fit on one page. Name the behavior, the situation that should trigger it, the standard for doing it well, the opportunity the learner will have to practice, and the person who will observe it. This forces awkward but valuable questions. Where will the new behavior happen? How often? What will make it difficult? Does the employee have the authority, time, information, or system access needed to act differently?
The question I would ask a manager before any course is, “Where will this person use the skill by Thursday?” If the answer is “whenever the opportunity comes up,” there may be no transfer plan. The opportunity may be rare, or the employee may not know what counts as a good attempt. A manager who can name the situation can also create one: schedule a customer debrief, assign a real presentation, or make the next one-to-one a practice ground.
This is where incentives matter more than most learning teams want to admit. Suppose a contact-center team is trained to slow down, ask better questions, and resolve the underlying issue. If supervisors judge performance almost entirely by average handle time, the training is competing with the operating system. Sending a refresher module will not settle that conflict. The organization must adjust the metric, clarify when the metric can be ignored, or teach the skill only in cases where it improves the business outcome.
The counterintuitive move may be to remove a barrier rather than add reinforcement. A manager deleting one contradictory target can produce more transfer than a learner completing three additional practice exercises. That does not mean practice is unimportant. It means practice cannot win a fight against a process that penalizes the desired behavior every time.
Manager support also needs a more precise definition. “Encourage your team to use the training” is too soft to guide action. A manager can set an expectation, provide an opportunity, observe one attempt, and discuss what happened. That may take fifteen minutes. It may also require the manager to stop interrupting the employee's first attempt with a correction. Support that becomes constant monitoring can make people hide mistakes, especially when the skill involves judgment rather than a fixed script.
Practice for the real problem, not the ideal one
Training design has to balance near transfer and far transfer. Near transfer means using a skill in a situation that closely resembles practice. Far transfer means adapting it when the context changes. Both matter, but they are built differently.
A service-desk analyst learning an escalation process should first handle a straightforward case with the correct facts and decision path. That gives the learner a clean repetition. The next cases should include missing information, an impatient internal customer, a conflicting priority, or a handoff to another team. The analyst needs to practice the underlying judgment, not memorize the order in which a facilitator presented the steps.
This is one place where realistic training can go wrong. More realism is not automatically better. A simulation that reproduces fifteen system screens, five policies, and an impatient manager may feel authentic while leaving too little attention for the one behavior that matters. A highly detailed scenario can also reproduce a bad local norm. Fidelity is useful when it changes the decision the learner must make; it is expensive decoration when it does not.
The best sequence is often simple: demonstrate the behavior, rehearse it under clean conditions, add one complication at a time, then practice in the actual work setting. A short checklist, prompt card, or supervisor cue at the point of use can help bridge the final step. The evidence for spacing and retrieval is strong, but there is no universal number of repetitions or perfect simulation formula for every corporate role. Organizational research is thinner than the confident claims made around many training recipes.
That limit matters. A high-stakes procedure may justify extensive simulation, certification, and observation. A low-frequency administrative task may be better served by a clear job aid and an annual drill. Building a six-week transfer program for every skill wastes time and teaches employees that learning initiatives always arrive with extra paperwork. Choose the level of support according to the risk, frequency, and complexity of the behavior.
Measure behavior, then admit what you cannot prove
Training effectiveness is the extent to which a target behavior improves in the job and remains useful there. A quiz, satisfaction score, or confidence rating can tell you something about the learning experience. None proves transfer.
Kirkpatrick's third level—behavior—offers a helpful discipline, though it is not a complete evaluation strategy. Before training, establish a modest baseline for the behavior. After training, observe the same behavior under normal conditions. Then check whether it persists after the initial enthusiasm has faded. For a new de-escalation skill, a quality team might review a handful of approved calls before training and another handful a week or month later, looking for whether the representative acknowledged the issue, asked a useful question, and confirmed the next step.
The sample does not need to pretend to be a scientific experiment. It needs to be consistent enough to reveal a pattern. If the behavior improves but only when a supervisor is present, the issue may be pressure rather than transfer. If it appears in practice but not in live work, inspect the workflow. If it improves and then disappears, look at manager attention, workload, or competing targets. If possible, compare the change with a similar team that has not yet taken the course. If that is not possible, describe the result as directional rather than claiming the course caused it.
Do not use confidence as a substitute for evidence. Learners often feel more capable immediately after training, which can be useful motivation but is not the same as competent performance. Nor should a business outcome such as sales, retention, or customer satisfaction be attributed to one course without accounting for other changes. Use those outcomes as context and look for the nearer behavior that the training was supposed to influence.
This approach fails when the job offers no genuine opportunity to use the skill. It also fails when leaders ask employees to demonstrate a behavior without granting the authority or time required to do it. In a rare emergency, a checklist and a well-rehearsed drill may outperform a broad course. In a regulated role, transfer observations should supplement—not replace—formal qualification and safety controls.
There is a real cost here. Managers need time to create opportunities and observe performance. Learning teams need access to operational data and the willingness to train fewer behaviors more thoroughly. Some organizations will decide that the cost is not justified for a low-risk task. That is a sensible decision. The dishonest alternative is to run a cheap course, report completion, and call the absence of workplace change a learner problem.
This week, take the next course on your calendar and pause before sending the invitation. Write down one observable behavior, the first real situation in which it should appear, the manager who will see it, and the evidence you will review seven days later. That half-hour of preparation will tell you whether the course has a destination—or merely a date.