How do you find the gaps a learning program should close?

Ask an expert what salespeople should do in a pricing conversation and you will have a clear, confident answer in five minutes. Ask what salespeople actually do in pricing conversations today, and the room goes quiet. That asymmetry is the central problem of needs analysis. The should-do side is easy: experts carry it around fully formed and can recite it on request. The today side is invisible: nobody writes it down, managers describe it in generalizations, and the people doing the work cannot see their own habits from the inside. A learning program aimed at an unknown starting point spends its runtime on what the audience already knows and misses what they actually do. This guide is about the hard half: what a gap is, what kinds exist, and where the evidence of today's behavior actually lives.

What exactly is a gap?

A gap is the distance between what people typically do today and what they should be doing, in a specific situation. Each part of that definition matters.

A gap is not a topic. "Pricing" is not a gap. A gap is not a competency label either. "They need better negotiation skills" is not a gap; it is a conclusion someone drew from behavior they never described. A gap is a behavioral distance, measured between two observable points. In the renewal call where the customer pushes back on price, typical reps offer a discount within the first two minutes; the best reps ask what is driving the budget pressure before conceding anything. Situation, typical behavior, better behavior. That is a gap, and a program can be aimed at it.

The definition also explains why the room goes quiet: the should-do end of the distance is cheap, because experts state it on demand, and the typical-today end is expensive. The rest of this guide is about paying that cost.

What kinds of gaps are there?

Not every gap is the same kind of problem, and the kind determines what closing it will take. This guide sorts gaps into three types by how the new material relates to what the learner already believes, a frame built on Seifert's published argument about how mental models accept or reject new information.

• Information gaps. Something is missing: the learner has never encountered the information, and once it arrives it fits into their existing mental model without a fight.
• Misconceptions. Something is wrong: the new material contradicts a belief the learner already holds, and that contradiction produces resistance, disbelief, and misunderstanding rather than adoption.
• Discontinuities. Nothing connects: the new material does not contradict the learner's mental model, but it does not attach to it either, so there is no path to assimilate it and it never becomes something the learner can use.

The three types fail differently. Information gaps close with good content. Misconceptions and discontinuities do not, because the obstacle is not missing material but the learner's existing model, which is why so many programs deliver accurate content that changes nothing. The full treatment of this taxonomy, and of how mental models decide what learners can accept, is the subject of its own companion guide to mental models in learning design.

The taxonomy also sharpens the instructional designer's assignment: to know which type of gap you are facing, you have to know what learners currently believe and do.

Why is the today side the hard half?

Four reasons, and every L&D team runs into all of them.

Nobody documents it. Organizations write down procedures, policies, and the occasional postmortem. Nobody writes down how the pricing conversation actually went or which step of the protocol quietly gets skipped on busy days. The behaviors a program most needs to know about have no paper trail.

Managers generalize. Ask a manager where their people fall short and you get conclusions: "they need better negotiation skills," "their judgment is weak under pressure." These are verdicts, not observations. The behavior that produced the verdict is valuable; the verdict itself is unusable as design input.

Self-reports describe intentions. Ask people what they do in a difficult situation and they tell you, sincerely, what they mean to do, which usually resembles the training they last received. The distance between intention and behavior is often the very gap the program needs to close, so asking performers directly measures the wrong end of it.

The behavior happens unobserved. The moments that matter most, the objection on the call, the bedside conversation, the one-on-one where the coaching should have happened, occur in rooms with no observer present. There is no recording to review, only traces.

Where does today-side evidence actually live?

No single source holds the truth. Four reliable ones, combined, get close.

Specific recent incidents. The founding evidence here is old and still decisive. In 1941, the U.S. Army Air Forces analyzed why 1,000 pilot candidates had washed out of flight training. The reasons recorded by the elimination boards were clichés and generalizations: "poor judgment," "unsuitable temperament," "lack of inherent flying ability." Much of it was too vague to use. The remedy, which Flanagan later formalized as the critical incident technique, was to ask qualified observers for specific observed events instead, ending with the request, "Describe the officer's action. What did he do?" In later studies of combat leadership, that style of question produced several thousand usable incidents (Flanagan, 1954). The same move works in any needs analysis: not "what do people struggle with," but "tell me about the last time someone mishandled a price objection." Ask for recent incidents; Flanagan's field data showed observers lose most incidents from memory within weeks. Getting these incidents out of experts, with the right probes and passes, is its own craft, covered in our companion guide to extracting what experts actually know.

Manager complaints, read as field reports. "They know it but they don't do it," "they revert under pressure," "she can recite the framework but freezes in the room." Complaints like these are usually treated as noise. They are data: eyewitness accounts that a specific behavior is not happening in a specific kind of situation. One follow-up converts the complaint into design input: when did you last see that happen, and what did the person do instead?

The completer/non-completer comparison. Brandon Hall Group's needs-analysis guidance poses three diagnostic questions: has the needle moved on key business metrics since training was implemented, are employees actually developing the skills the training is meant to impart, and, sharpest of all, "Is there a clear difference in performance between those who complete training and those who don't?" If completers and non-completers perform the same, the existing program is not touching the behavior. Their illustration: imagine a group that completes customer engagement training and showed conversion rates identical to those who had not, revealing that motivation and tools, not knowledge, were the real barriers (Brandon Hall Group, 2025).

The trails behavior leaves. Escalations, unplanned discounts, rework, repeated support tickets, audit findings: each is the visible end of an invisible behavior. Follow an escalation backward and you find the conversation that failed; follow a rework queue backward and you find the step that gets skipped. These operational traces are the closest thing to a recording of the rooms nobody observes.

Content developers retrofitting an existing program use exactly the same sources; the playbook for adding an application step shows where this interview work slots into that workflow.

What if the gap is not about information at all?

One caution before writing anything down. Some gaps are not information problems in any form, and clinicians offer the clearest illustration. Holtrop and colleagues studied why physicians resisted shared decision making for implantable defibrillators, an approach the clinicians knew well. On the surface they said things like "I already do shared decision making" and "I just don't have time for this." Underneath, the researchers found a quieter belief: the physicians saw their job as helping patients live as long as possible, wanted patients to get the device, and expected that a decision aid might steer patients away from it (Holtrop et al., 2021). No amount of content closes that distance, because the gap lives in how the clinicians see their role, not in what they know. When your evidence shows people who understand the better behavior and still do not adopt it, you have found this kind of gap, and getting underneath it belongs to the mental-models guide rather than another module.

What should a written gap look like?

The output of all this evidence gathering is a set of gap statements, each with four parts:

1. The situation. Specific enough that a practitioner would recognize it: the renewal call after a price increase, the handoff between shifts, the second missed deadline.
2. What people typically do. The observed missteps, in behavioral language, drawn from the incidents and trails above.
3. What the best performers do instead. Also behavioral, also observed, not the idealized version from the manual.
4. Who can explain why. A named expert, coach, or top performer who can articulate why the typical move falls short and the better move works.

These gap statements live one layer below learning objectives in a well-built program: the objectives state what the program commits to changing, and each objective stands on the specific gaps beneath it, a two-layer structure formalized in our companion guide to writing learning objectives. Downstream, each gap becomes exactly one scenario decision, the handoff that Designing Decision Points picks up in full.

What do the gaps become?

The reason the today side is worth this much effort is what it turns into. When a gap statement carries real typical-today behaviors, those behaviors become the tempting suboptimal options at a scenario decision, the choices learners recognize because they have made them, or watched a colleague make them, in the past month. The best-performer moves become the optimal options, and the named expert's reasoning becomes the coaching at the moment of choice. This is the design approach behind Guided Scenarios on AliveSim: start from the learner's current behavior rather than the expert's finished model, let their first instinct at a decision reveal which kind of gap is present, and close it right there. None of that works if the today side was guessed at a desk, which is why finding the gaps is where every program worth building starts.

References

• Flanagan, J. C. (1954). The critical incident technique. Psychological Bulletin, 51(4), 327–358.
• Holtrop, J. S., Scherer, L. D., Matlock, D. D., Glasgow, R. E., & Green, L. A. (2021). The importance of mental models in implementation science. Frontiers in Public Health, 9, 680316.
• Rochelle, M. (2025). How to conduct a learning needs analysis for 2026: A step-by-step guide for L&D leaders. Brandon Hall Group.
• Seifert, D. (2023). Mental models: The secret to effective education and training outcomes. Syandus.
• Seifert, D. (2024). The Science of Guided Scenarios, Part 5: Building Stronger Mental Models. Syandus.

Published July 17, 2026 · 9 min read