Developed by Allan Collins, John Seely Brown, and Susan E. Newman (Cognitive Apprenticeship, 1989), with Ann Holum (1991).
Teaches judgment-heavy skills by having an expert think aloud while working, coaching the learner's attempts, adding scaffolds, asking for explanations, comparing approaches, and then withdrawing support.
Six methods for teaching a valuation, from showing the thinking to letting go
A senior analyst teaches a new hire to value a company. At each of the senior's turns, pick a move: the method that stage calls for, or the shortcut that skips the thinking. The reply shows what the learner gets, and the bar tracks how much of the work is now hers. Illustrative.
Senior analyst
New analyst
Senior analyst
New analyst
Her share of the work0of 6
- Modeling (Senior analyst). The new analyst has never built a valuation. Pick how the senior starts.
- Send her last year's finished model to study. I see the numbers. I don't see why any of them are there. A finished product: the judgment that made it is invisible.
- Build one live, saying out loud why each assumption goes where it does. So you cut their growth forecast because their last three forecasts missed? Modeling: the expert's reasons, and dead ends, become visible.
- She watches (New analyst). The thing worth learning was the senior's distrust of management's numbers, which no finished file shows.
- Coaching (Senior analyst). She is building her first model for a second company. Pick the senior's move.
- Watch her build it, and hint when she reaches for the wrong margin. Wait, gross margin? Right, it should be the operating margin. Coaching: she catches the error herself, with a nudge.
- Take the keyboard and fix it so it's done right. Thanks. I'm not sure what you changed. Taken over: the model is right, and she learned nothing.
- First try (New analyst). She finished the model herself. The hints caught two errors she can now spot alone.
- Scaffolding (Senior analyst). She will build three more this quarter. Pick the support.
- A template with every formula filled in, to use from now on. Great, I'll just change the inputs. Permanent support: she learns to fill a form, not to build a model.
- A template with the revenue drivers laid out, removed after two deals. So I build the rest, and next time I build the drivers too. Scaffolding: help for the hard part, with a date it comes off.
- With support (New analyst). She is now doing most of the build. The template carries only the part she has not yet learned.
- Articulation (Senior analyst). Her discount rate looks reasonable. Pick the senior's line.
- Walk me through why you chose that discount rate. I took it from the last deal. I didn't check whether this company is riskier. Articulation: explaining it out loud exposes a step she skipped.
- Your discount rate looks fine. Moving on. Good. Approved: the copied number stays, and so does the habit.
- Her reasoning (New analyst). Saying it aloud made her see a choice she had not made on purpose.
- Reflection (Senior analyst). Her model and the senior's differ by 15 percent. Pick the move.
- Grade it: good work, a B plus. Thanks. Which parts were the minus? A grade: a verdict with no comparison, so the gap stays unexplained.
- Put our models side by side and find where they differ, and why. Most of the gap is the customer churn assumption. I see why you set yours higher. Reflection: she sees her process against an expert's.
- The gap (New analyst). She now knows which assumption carries the most weight, and why the senior set it where he did.
- Exploration (Senior analyst). She is ready for more. Pick the next assignment.
- A company in an industry neither of us has valued, with no template. Okay. I'll start with the churn and the forecasts, and bring you questions. Exploration: she frames the problem herself, and support comes off.
- Another deal just like the last one, until she gets it perfect. Sure. Same template? Held back: she gets faster at one kind of deal and no better at judgment.
- On her own (New analyst). She builds the model and brings questions, not drafts for correction. The senior's role has shrunk to answering them.
An example
A mid-market advisory firm hired an analyst from accounting and gave her three months to value companies alone. The first plan was the usual one: a folder of past models and a review of whatever she built.
A senior analyst ran it differently. He built one valuation in front of her, saying out loud why he cut the client's growth forecast (their last three had missed) and where he had gone wrong before. He watched her build the next one and hinted when she reached for gross margin instead of operating margin. He gave her a template with the revenue drivers laid out, and took it away after two deals. He asked her to explain her discount rate, and she found she had copied it from the last deal. They put her model next to his, and most of the 15 percent gap came down to one churn assumption.
In the third month he gave her a company in an industry neither of them had valued. She built the model herself and brought him questions instead of drafts.
Why it works
Expertise in judgment-heavy work lives in decisions nobody sees: what to distrust, which number matters, when to stop. Finished examples hide all of it, and lectures describe it in terms a novice cannot yet use. Traditional craft apprenticeship solved this for physical skills, where the learner could watch the hands. Cognitive apprenticeship adapts it to thinking, by making the expert's reasoning audible and the learner's reasoning visible.
The methods are ordered for a reason. Modeling shows the thinking; coaching and scaffolding let the learner try it with help; articulation and reflection make the learner put her reasoning into words and set it against the expert's; exploration hands over the framing of new problems. Support is withdrawn on purpose at each step, so the learner ends up doing the whole task, judgment included, instead of a supported version of it.
The thinking behind it
Allan Collins Allan Collins, a Northwestern University learning scientist whose work spans cognitive psychology, artificial intelligence, and education, described cognitive apprenticeship with John Seely Brown and Susan E. Newman in a 1989 essay on teaching reading, writing, and mathematics, written for a volume honoring Robert Glaser. The model makes an expert's thinking visible to a learner through six methods: modeling, coaching, scaffolding, articulation, reflection, and exploration.
John Seely Brown John Seely Brown, who directed Xerox PARC and served as Xerox's chief scientist, proposed the model with Collins and Newman as an adaptation of craft apprenticeship to school subjects. With Collins and Ann Holum he brought it to teachers in Cognitive Apprenticeship: Making Thinking Visible (1991), an article in American Educator built on modeling, coaching, scaffolding, and fading, the gradual removal of support.
How to run it
- Name the skill. Pick the judgment to pass on, such as setting valuation assumptions, not the whole job.
- Model it aloud. Do the task in front of the learner, narrating each choice, doubt, and dead end.
- Coach the first attempts. Watch the learner try, and give hints, not fixes.
- Scaffold, with an end date. Give support for the hardest part, and set when it comes off.
- Ask for the reasoning. Have the learner explain choices out loud, which surfaces the ones made by habit.
- Compare with the expert. Set the learner's work beside the expert's and trace where and why they differ.
- Hand over new problems. Assign a task with no template, so the learner frames it as well as solves it.
Where it goes wrong
The expert models the result, not the thinking. A senior who builds the model quickly and silently has shown speed, not judgment. Slow down and say the reasons, including the ones that feel too obvious to mention.
The scaffolding never comes off. A template left in place produces someone who fills templates. Set the removal date when the template is handed over, and keep it.
Coaching turns into correcting. Fixing the learner's work is faster and feels helpful, and it takes the learning away. An expert who ends up at the keyboard should hand it back and ask a question instead.
When to use something else
- Deliberate Practice. When the skill is known and the learner needs focused repetition with feedback to get better at it.
- Zone of Proximal Development. When the question is how much help a learner needs right now, and what they are almost ready to do alone.
- Toyota Kata. When the skill is improving a process, coached in short daily cycles by a manager.
Where it sits in SOLVE eX
Go to the source
- Cognitive Apprenticeship: Teaching the Crafts of Reading, Writing, and Mathematics, Knowing, Learning, and Instruction (Lauren B. Resnick, ed.), 1989
- Cognitive Apprenticeship: Making Thinking Visible, American Educator, 1991