All lessonsBusiness & Leadership

Why Your Best Performer Can't Teach It: Modelling the Skill Nobody Can Explain

Your top performer cannot hand the skill over in words, and that is exactly where the business stalls. The NLP modelling process extracts the skill from the doing rather than the description, then tests it on someone who was never in the room.

2026-10-13 2,566 words 13 min read

The Eight Shooters And The Man Who Could Not Explain It

In 2001 a national sports psychologist asked me to bring Neuro-Linguistic Programming into the Singapore Armed Forces shooting contingent. The team had slipped from second place to fourth. Singapore is a small country, and those shooters carried more than their own scores.

The squad was made up largely of full-time national servicemen, boys of eighteen and nineteen. Their competitors were regular soldiers from other armies with years on the range behind them.

I was given eight shooters to record on video. Two M16 gunners, two pistol shooters, two GPMG gunners, two carbine shooters. My job was to find what the best of them did, so the rest of the squad could copy it.

What I found was stranger than a technique. Each man had built a different internal route to the same target. They worked with pictures, with sound, with inner speech, and not one of them had been trained in any of it.

One carbine shooter stood out. Nothing unsettled him. He rehearsed his physical movements while he slept, to the point where his body twitched in the bunk. When rain came down on the range and the other shooters packed up, he stayed inside his mental range and kept firing.

Then I made a mistake. I asked him how he did it.

He said, "I focus."

That was the whole answer. The most consistent performer in the squad could not hand me the skill in words. He had it. He could run it. He could not describe it. That gap is what this article is about, and every founder who has tried to clone their own best operator has stood inside it without a name for it.

The Skill That Runs Below The Sentence

Michael Polanyi put the problem in seven words in 1966: we can know more than we can tell. His book The Tacit Dimension described how much of competence lives in the body and the habit, and how the attempt to force it into language distorts it as often as it captures it.

Donald Schön named the same thing knowing-in-action in The Reflective Practitioner in 1983. The professional at work is not following a checklist. The professional is reading a live situation and adjusting in real time, and the adjustment happens faster than the explanation can form.

The mechanism is ordinary, and it is the reason this matters. When you repeat a skill enough, the brain compiles it. What started as twenty conscious steps runs as one unit. The compilation is the achievement. It is also the loss of the instruction manual. The steps are no longer available to the part of you that speaks.

Ask a skilled person to explain the compiled routine and you get a reconstruction. The reconstruction is a story about the skill, assembled after the fact, and it is often a good story and a wrong one.

That compilation is why a ten-year operator can be worse at teaching a six-month operator than a two-year operator is. The longer the skill has run on its own, the further the spoken description has drifted from the doing. The person nearest to the skill is frequently the one who can least see it.

What Modelling Actually Is

Richard Bandler and John Grinder did not begin with explanations. In the early 1970s they sat with three exceptional therapists, Fritz Perls, Virginia Satir and Milton Erickson, and recorded what those people did. They watched the work, transcribed it, and built a description of the structure underneath it. The result was The Structure of Magic in 1975 and the Erickson volumes that followed.

That method is modelling. You decline the expert's lesson. You gather samples of the performance itself, find the difference that makes the difference, and write it in a form another person can run.

They borrowed the unit of analysis from George Miller, Eugene Galanter and Karl Pribram, whose 1960 book Plans and the Structure of Behavior described behaviour as test-operate-test-exit, a loop rather than a line. A skilled performance is a set of these loops, and modelling is the work of drawing them out.

Robert Dilts carried the method into business and education and set it out in Modeling with NLP in 1998. John Grinder and Carmen Bostic St. Clair restated it in Whispering in the Wind in 2001 as the core of the discipline, and they were blunt that much of what followed had drifted away from modelling and into collecting techniques.

I will be straight about the evidence here, because the field has earned a reputation for overclaiming. NLP is a modelling discipline. The descriptions it produces of how a skilled person works are hypotheses about structure. Some of the mechanisms they point at, such as the role of internal imagery, self-talk and physiological state in performance, have support in mainstream psychology. The specific NLP patterns are modelled descriptions rather than validated instruments, and I will not dress them up as controlled trials.

The Three Places The Skill Escapes

One: automaticity. The routine is compiled and it no longer reports its own steps. The expert's account is a memory of the skill, and the memory edits itself every time it is retold.

Two: the accuracy gap. Timothy Wilson and Richard Nisbett showed in 1977 that people confidently report reasons for their own behaviour that they could not have used, and they have no idea the report is wrong. The expert is not lying. The expert is describing a plausible process, and over time the plausible process and the real one have drifted apart.

Three: the sample problem. One great performer is a sample of one. His calm may come from temperament rather than from a move you can teach. Model the person and you inherit their accidents. Model the function and you inherit the skill.

Put those three together and you have the reason training most high performers fails at the first attempt. You asked the wrong person the wrong question at the wrong moment, and you mistook a description for an instruction.

What The Research Says About Turning Ability Into Transfer

K. Anders Ericsson and his colleagues argued in 1993 that expert performance grows out of deliberate practice: effortful work on specific weaknesses, with immediate feedback, sustained over years (Ericsson, Krampe and Tesch-Römer, Psychological Review, 1993, expanded in Peak, 2016). Repetition on its own is not the engine. Feedback on the repetition is.

John Hattie and Helen Timperley reviewed four decades of feedback research in 2007 and found enormous variation in effect. The feedback that changes performance answers three questions: where am I going, how am I going, and where to next. Vague praise does nothing measurable. Avraham Kluger and Angelo DeNisi had shown in 1996 that a substantial share of feedback interventions actually lowered performance, which is worth remembering before you hand a performer a score.

The training design that matches modelling has a name. Allan Collins, John Seely Brown and Susan Newman called it cognitive apprenticeship in 1989. The master makes their thinking visible, the learner attempts the task, the coach scaffolds and then fades the help. Albert Bandura's work on modelling had shown years earlier that people acquire behaviour by watching a competent model, and that watching someone similar succeed raises the observer's own belief that they can do it (Bandura, 1977, and Social Foundations of Thought and Action, 1986).

Daniel Kahneman and Gary Klein closed a long argument in 2009 by agreeing on when skilled intuition can be trusted. The environment has to be regular enough to learn, and the practitioner has to have had the chance to practice with feedback. Remove the feedback and the expert's certainty becomes confidence without calibration.

Gary Klein's earlier work on cognitive task analysis, with Beth Crandall and Robert Hoffman, took the method into the field: interview people about real incidents, push on the decision points, and rebuild the expertise they cannot summarise (Hoffman, Crandall and Shadbolt, Human Factors, 1998). That body of work is the professional version of what I was trying to do on a firing range.

The Modelling Protocol: Five Steps

This is the sequence I run now. It works for a sales play, a leadership behaviour, or a skill you personally cannot hand off.

Step one. Name the outcome and the exemplar. Pick one performer and one thing they do that you would pay to duplicate. Write the outcome as observable behaviour at a specific moment, not as praise. Sell the complex renewal in the room with the sceptical buyer.

Step two. Capture the doing, not the story. Record it. Watch the performance, and note what is on the recording that the person never mentions. The details they omit are usually the ones that matter.

Step three. Find the difference. Contrast a good day with a bad day. Contrast the expert with a competent non-expert doing the same task. Ask after the doing, never before: what did you see, what did you say to yourself, what did you feel in the second before it worked. The contrast is where the structure shows itself.

Step four. Write the smallest teachable unit. One page. The trigger, the internal move, the decision point, the recovery. Test the page on a second person who does not have the skill yet.

Step five. Teach a third person and watch where it breaks. Their failure is your next question. Fix the description, run it again. The description earns its place the day it makes a stranger competent under pressure. This fifth step is the calibration loop, and it is the step almost everybody skips.

The order matters, and the discipline sits in the last two steps. Collecting the performance is observation. Writing the page is description. Only the third person shows whether you built a transferable skill or a private memory of one person on a good day.