Mental Models: Munger's Latticework, 12 Models Operators Actually Use, and the 3 the Stoics Added First

A mental model is a tool you think with. Collecting them is a hobby. Using them is a discipline.

Most operators run their business on two or three models they never named: whatever worked last time, whatever the loudest advisor said, and whatever the spreadsheet seems to show. Those are models too. They are just unexamined, and an unexamined model is the one that quietly steers you into the wall.

The phrase "mental models" has become its own content industry, with lists of a hundred and posters for the wall. Lists are not the point. The point is a small set of tools you actually reach for when a pricing decision, a hire or a bad quarter is on the desk.

This article gives you what a mental model is, the latticework idea in its author's words, twelve models operators actually use with one application each, the three the Stoics built two thousand years earlier, and the install protocol that turns a list into a habit.

What a mental model is

The term is older than the business-book version. The reference entry records that it "was coined in 1943 by Kenneth Craik, who suggested that the mind constructs 'small-scale models' of reality that it uses to anticipate events."

Small-scale models of reality, used to anticipate events. That definition does more work than any list. A model is a compressed version of how some part of the world behaves, held in your head so you can predict what happens next without running the experiment.

Every operator already has them. The question is whether they are the right ones, whether there are enough of them, and whether you know which one you are using when you decide.

The latticework, in Munger's words

The business version of the idea comes from a 1994 talk Charlie Munger gave at the USC Business School, titled "A Lesson on Elementary, Worldly Wisdom," transcribed at Farnam Street. His first rule:

you can't really know anything if you just remember isolated facts and try and bang 'em back. If the facts don't hang together on a latticework of theory, you don't have them in a usable form.

Then the instruction: "You've got to have models in your head." And the warning about having too few, which is the one most operators need. With only one or two, "the nature of human psychology is such that you'll torture reality so that it fits your models, or at least you'll think it does." He borrows the old saying for it: "To the man with only a hammer, every problem looks like a nail."

Two details from the same talk keep the idea practical. On the number: "80 or 90 important models will carry about 90% of the freight in making you a worldly-wise person. And, of those, only a mere handful really carry very heavy freight." You do not need a hundred. You need the handful, used.

And on knowing your limits: "Every person is going to have a circle of competence. And it's going to be very hard to advance that circle." Which is itself one of the models below.

Twelve models operators actually use

These are grouped by the decision they serve. Where Apex has written the full article, the link is there instead of a repeat.

For seeing clearly

1. First principles. Strip a decision to what is physically or contractually true, then rebuild from the floor instead of copying the market. Most pricing, hiring and training decisions are copies of someone else's. The full method is in first principles thinking.

2. Inversion. Instead of asking how to succeed, ask how you would guarantee failure, then avoid those things. It finds the failure modes a forward plan hides. Covered in full in the inversion mental model.

3. Second-order thinking. Every decision has a consequence, and the consequence has a consequence. A discount lifts this month's sales (first order), trains customers to wait for the next one (second order), and compresses your margin for a year (third order). Ask "and then what?" three times before committing.

4. The map is not the territory. Your dashboard, your forecast and your org chart are maps. The business is the territory. When the map and the territory disagree, the territory is right. Operators who forget this manage the dashboard and lose the customer.

For allocating

5. Opportunity cost. The cost of any decision is the best alternative you gave up. A free consult call is not free; it costs the hour of Build work that did not happen. Price every yes in the currency of what it displaces.

6. Margin of safety. Build in more room than you think you need: cash reserve, delivery capacity, training recovery. The model exists because forecasts are wrong in the direction that hurts. Three months of operating cash is a margin of safety. So is a deload week.

7. Compounding. Small, consistent gains on a growing base outperform large, occasional ones. It is the mechanism behind the whole Apex frame: training, Stoic practice and systems each make the others stronger over years. The business version is progressive overload applied to growth.

8. Pareto. A minority of inputs produces a majority of outputs: a few clients, a few offers, a few habits. Find the minority and protect it. The rest is a candidate for an SOP, a tool or a no.

For judging people and numbers

9. Regression to the mean. Extreme results tend to be followed by less extreme ones. The record sales month was partly luck, and so was the terrible one. Decide on the trend, not the spike, and never restructure the business after a single outlier week.

10. Incentives. Show the incentive and you can predict the behaviour. A salesperson paid on bookings will book. An agency paid on spend will spend. Before judging anyone's advice, including your own, ask what they are paid for.

11. Circle of competence. Munger's rule above. Know the edge of what you actually understand and stay inside it, or hire the person whose circle covers the gap. Most expensive founder mistakes happen two steps outside the circle.

12. Hanlon's razor. Do not attribute to malice what is adequately explained by oversight. The client who went quiet is usually busy, not angry. The model saves relationships and, more importantly, saves the week of rumination that follows the wrong assumption.

The three models the Stoics built first

The Stoics did not call them mental models. They called them exercises, and they built them to do exactly what Craik described: small-scale models of reality used to anticipate events. Three of them belong in any operator's handful.

The dichotomy of control

Epictetus opens the Enchiridion with it: "Some things are in our control and others not. Things in our control are opinion, pursuit, desire, aversion, and, in a word, whatever are our own actions. Things not in our control are body, property, reputation, command, and, in one word, whatever are not our own actions."

As a model, it is a sorting function. Every input to a decision goes into one of two columns, and effort goes only on the first. It is the fastest filter for cognitive load an operator has. Applied to a team, it is covered in the dichotomy of control for team management.

Chapter 5 of the same text supplies the companion model: "Men are disturbed, not by things, but by the principles and notions which they form concerning things." The event and your judgement of the event are two different objects. Only one of them is editable.

Premeditatio malorum

The deliberate rehearsal of what could go wrong before it does. It is inversion with a two-thousand-year head start, and the Stoics ran it daily rather than at the planning offsite. The business implementation is the pre-mortem. The practice is in premeditatio malorum for entrepreneurs.

The view from above

Marcus Aurelius returns to this one repeatedly. In Book VII of the Meditations, in the Long translation, he approves Plato's line that "he who is discoursing about men should look also at earthly things as if he viewed them from some higher place." In Book IX he instructs himself: "Look down from above on the countless herds of men and their countless solemnities, and the infinitely varied voyagings in storms and calms."

As a model it is a scale correction. A lost client at the scale of this week is a crisis; at the scale of a ten-year business it is a data point. Regression to the mean is the statistical version of the same move. The practice is in the view from above for business setbacks.

Why these are models and not moods

A mood is something you have. A model is something you run. Each of the three above has an input, a procedure and an output: sort the inputs, rehearse the failure, change the scale. They produce different decisions than you would have made without them, which is the only test of a model that matters. Stoicism is an operating system, not a feeling, and these are three of its core routines.

How to install a model

Reading about a model installs nothing. Munger's point was that experience has to be hung on the latticework, and that is a practice, not a reading list. The Apex install protocol:

1. One model per month. Pick one from the fifteen above. Write its one-sentence definition on a card and put it where you decide things: the desk, the notebook, the top of the weekly review.
2. A decision journal. For thirty days, every decision above a threshold (money, people, time over an hour) gets three lines: the decision, the model applied, the expected result. Date it.
3. The Friday read-back. In the weekly review, reread the month's entries. Mark where the model changed the decision and where you forgot to use it. The second column is the data.
4. The ninety-day check. Once a quarter, reread the oldest entries against what actually happened. This is where regression to the mean, second-order effects and your own circle of competence become visible, in your own handwriting.
5. Train the same way. The body teaches models faster than the business does. Progressive overload, margin of safety and regression to the mean all show up inside a twelve-week training block, with a log to prove them. An operator who already runs those models under a barbell has less to learn at the desk.

Twelve months of this gives you twelve models with evidence attached, which is a latticework. Twelve months of reading lists gives you a bookshelf.

The collector's trap

The failure mode of mental models is the same as the failure mode of supplements: acquiring them feels like progress. Munger's own number was that a handful carry the heavy freight. If you cannot name the model you used on your last three decisions, you do not have a latticework yet, whatever is on the poster.

The second trap is the hammer. One model, applied to everything, is worse than none, because it comes with confidence. An operator who sees only incentives becomes cynical; one who sees only compounding never cuts a loss. The latticework exists so the models check each other.

The third is mistaking the vocabulary for the work. Saying "second-order effects" in a meeting is not the same as writing down the second-order effects before the decision. The journal is the difference.

Frequently asked questions

What is a mental model?

A mental model is a compressed representation of how some part of the world works, held in the mind and used to predict outcomes and make decisions. The term is traced to Kenneth Craik in 1943, who described the mind constructing small-scale models of reality to anticipate events. Operators use them, named or not, every time they price, hire or plan.

What did Charlie Munger mean by a latticework of mental models?

In his 1994 USC talk he argued that isolated facts are useless unless they "hang together on a latticework of theory," and that you need multiple models from multiple disciplines so you do not "torture reality" to fit the one or two you have. He estimated that 80 or 90 important models carry most of the value, and only a handful carry the heaviest load.

Which mental models matter most for a business owner?

For most operators the heavy-freight set is first principles, inversion, second-order thinking, opportunity cost, margin of safety, compounding and incentives, plus the three Stoic models: the dichotomy of control, premeditatio malorum and the view from above. The right test is which ones changed a real decision in the last quarter.

Are Stoic practices mental models?

Yes, in the exact sense Craik described. The dichotomy of control sorts inputs into controllable and not, premeditatio malorum rehearses failure before it happens, and the view from above changes the scale at which a problem is judged. Each has an input, a procedure and an output, which is what separates a model from a mood.

How do you learn mental models so you actually use them?

One model a month, written on a card where you decide things, plus a decision journal that records the decision, the model applied and the expected result. Review it weekly for where the model changed the decision and where you forgot it, and quarterly against what actually happened. Twelve months of that builds a latticework; twelve months of reading does not.

Think with the tool, then check the tool

A mental model earns its place by changing a decision you would otherwise have made differently. That is the whole standard. Fifteen are listed above, three of them two thousand years old, and a handful of them will carry most of the weight in your particular business.

The practice is the same one the Stoics ran every evening: write down what you did, name the principle you applied, and check it against the result. Training gives you the same loop in the body, with a log that does not flatter. Together they make the latticework real instead of decorative.

The free 5-Day Stoic Operator Challenge installs the daily review that makes this work, alongside the training anchor that teaches compounding and margin of safety in the body first. Start there, then add one model a month.