Teach after you are paid, not before

What are the five levels of mastery?

Mastery of a skill, in the order I have watched it happen, moves through five levels. You learn it, you do it on yourself, you do it for other people for free, you get paid for it, and then you teach it. Each level teaches you something the one before it could not.

A level can only teach you something if the level before it left you material to work with. Doing it on yourself gives you room to fail cheaply and change the experiment. Doing it for someone else adds what you never meet alone. Getting paid adds an invoice.

Why does getting paid stop making you better?

Getting paid stops making you better because paid work punishes experiments. When a client is paying, you do what worked last time, because a failed experiment now costs them money and costs you a deadline.

This is not only my observation. In 1993, K. Anders Ericsson, Ralf Krampe and Clemens Tesch-Römer published the violin study the 10,000-hour idea later came from. It counts "services rendered for pay" as work, and separates work from practice. At work, they wrote, people "rely on previously well-entrenched methods rather than exploring alternative methods with unknown reliability", and "the costs of mistakes or failures to meet deadlines are generally great, which discourages learning". Their verdict on work as a place to learn: "far from optimal".

Ericsson came back to it in 2006. Most professionals, he wrote, "reach a stable, average level of performance, and then they maintain this pedestrian level for the rest of their careers." The skills that got them paid become automatic, and "when the behaviors are automatized, mere additional experience will not lead to increased levels of performance." His example is tying your shoelaces, which people have difficulty describing. That difficulty is also the reason teaching works.

What does teaching do that client work does not?

Teaching makes you hand over what you know, and the handover fails in ways client work never shows you. A client pays for the result. A learner tries to produce it. They ask questions, poke holes in what you said, and when they try it, they do not get what you get, because you have not handed everything over. You did not leave it out on purpose. You did not know it was there.

Michael Polanyi put the problem in one line in 1966: "we can know more than we can tell." It has been measured since. In a 1990 Stanford dissertation, Elizabeth Newton had people tap out the rhythm of well-known songs for a listener, then estimate whether the listener had named the tune. Their estimates averaged about half. The listeners named 3 of the 120 tunes tapped, 2.5%. In 2014, three expert surgeons were filmed teaching a cricothyrotomy, and what they said was checked against a full step list built with three other surgeons. On average they left out 51% of the action steps and 71% of the clinical-knowledge steps. The knowledge had gone automatic, and automatic knowledge does not show up when you explain.

So the teacher rebuilds. You simplify what you do and write it down as steps, then as a process, then as a framework. Then the questions push past anything the work ever required, and you go and research what you never needed to know.

Rod Roscoe and Michelene Chi, who study tutoring, call the two ways this can go knowledge-telling and knowledge-building. In a 2007 review they concluded tutors mostly do the first: "Peer tutors, even when trained, focus more on delivering knowledge rather than developing it." What tipped tutors into building was the learner. In their 2008 study, "when tutees asked questions that contained an inference or required an inferential answer, tutors' responses were more likely to be elaborative and metacognitive."

Imagine you understand muscle recovery the way most people who train do. The muscle breaks down, you eat, you sleep, and it rebuilds. That is enough to train yourself and to help a friend. Then you teach it. Someone asks which foods actually change recovery. Someone asks whether it differs for women. Someone asks about age. Training never needed those answers, so you go and find them, and connect them to what you have already watched work. The next time you teach it, you answer those questions before anyone asks.

What goes wrong when you teach too early?

Teaching too early fails because the questions have nothing to land on. Someone learns a thing from videos, books and conversations and goes straight to a webinar, a workshop or a course. They skip the two levels where knowledge gets tested: doing it on themselves, and doing it for someone else. When a learner asks a hard question, there is no experience to connect the answer to. The answer is another article.

If you have taught something you first read about this year, you probably remember the question you answered with a summary and looked up afterwards.

The strongest case against this comes from the laboratory. Learning by teaching works. In a 2014 study by Nestojko and colleagues, 56 students read a passage for 10 minutes. Those told they would teach it recalled more of it, and organised it better, than those told to expect a test. None of them ever taught. Kobayashi's 2019 meta-analysis of 28 studies found that preparing to teach helped, and teaching after preparing helped more, most of all when it was interactive. Read quickly, that says teach as early as you like.

Read the method and it says something narrower. In the typical study, students learn a short text or lesson, often in minutes, and are tested on that material. Teaching helped them remember and understand what they had just learned. None of the studies set teaching against doing the work, so none can say teaching stands in for it. A beginner who teaches will remember the book better, and will still have nothing to say when the question is about a client.

The most repeated number on the subject does not survive even that much reading. Searching "does teaching help you learn better" from Saudi Arabia on 3 October 2026, the first related question Google offered was "How to retain 90% of what you learn?" The figure comes from the learning pyramid, which claims we retain 90% of what we teach others. Letrud rebutted it in 2012, and in 2018 he and Hernes concluded that models like it "did not originate from empirical research".

What does Robert Greene's Mastery add?

Robert Greene's Mastery (2012) describes mastery from the apprentice's end. He splits it into three phases: "The first is the Apprenticeship; the second is the Creative-Active; the third, Mastery." In the apprenticeship you learn the way apprentices always have, "by watching Masters and imitating them as closely as possible", because "we learn best through practice and repetition". Only "as you gain in skill and confidence" does Greene tell you to move to experimentation. That is my first three levels in a different vocabulary.

Greene's numbers run from 10,000 hours, which he puts at seven to ten years, roughly a traditional apprenticeship, to "some 20,000 hours and beyond" for mastery. Even the 10,000 is soft. In the 1993 violin study it was roughly the average practice of the best group by age 20, and in 2016 Ericsson and Robert Pool wrote that half of those ten violinists had not reached it, and that "there is nothing special or magical about ten thousand hours". Malcolm Gladwell's Outliers (2008) turned the average into a rule.

Greene does not make teaching one of his phases. Where it appears in the book, it comes after mastery, as a way of repaying a master. The closest he comes is the mentor, a relationship he calls "a two-way emotional connection". He tells the apprentice: "You give them feedback to their instruction, perhaps adjust some of their ideas." He wrote that for the apprentice. Read it from the mentor's chair and it is the fifth level.

Where does teaching happen when nobody calls it teaching?

The most useful teaching I do as a marketing consultant does not look like teaching. I run workshops, and I have worked with 40+ marketing departments, but the place I learn the most is the first conversation with a potential client. I treat it as a one-to-one session. I ask questions and poke holes, to see if there is an opportunity. I listen for what people believe, and for the words they use.

In one of those conversations, about personal branding, a belief came up that I had not planned for. The person was afraid they might say something stupid. That fear was the reason none of the strategy would ever be used. It later went into a piece on founder personal brands as the fear of saying something wrong in public, and it changed what I say now. When someone carries that fear, I tell them the first thing can be a recording or a written piece, edited before anyone sees it, rather than anything live. Or they can appear in someone else's content before they make their own.

Isn't teaching what people do when they cannot do the work?

The idea that people teach because they cannot do is old enough to have an author. In 1903 George Bernard Shaw gave it to a fictional revolutionary, John Tanner: "He who can, does. He who cannot, teaches." At full strength it describes something I have seen: people whose main way of earning is teaching a thing they are not good at doing. Teaching pays sooner than practice does, and a room of beginners rarely asks a question a book cannot answer.

Teaching also has a pull that has nothing to do with mastery. It builds credibility and awareness. People start to know you. That benefit is real, and it is exactly the benefit that tempts people to teach before they have done the work. I do not have a fix for that temptation.

What the objection gets wrong is the test. People teach for different reasons: to pass on experience they already have, to reach a level of mastery they do not have yet, and to be known. From the audience, all three look the same. And two maxims below Shaw's famous one, under the same heading, Tanner's handbook says: "Activity is the only road to knowledge."

What to do with this

If you learned something this year, do it on yourself first, then for a few people for free, then for money. If you have been paid for the same work for a while, teach it. Write down every question you cannot answer, and treat that list as the next thing to learn.

When should you start teaching what you know? When the work has stopped teaching you.