Why do my models break every time something changes?

You build a model. It works once, for one set of numbers. Then an assumption shifts and the whole thing falls apart, or worse, keeps running and lies to you quietly. You call this bad luck, or not enough time. It is neither. It is the structure you chose when you built it. A flexible, driver-based model is not a gift some people have and you lack. It is a habit of keeping what you know separate from what you guess. Most builders bury both in the same formula, the same cell, the same wall of logic no one — including future you — can move without breaking something else. This is not a mystery of advanced technique. It is a discipline you have not yet practiced on purpose. You can start on the very next model you open, not the ideal one you plan to build someday.

◆ How this problem reads on the two dials

GuidanceKnowledge

Coaching

More to learn

1:1 with AureliusWith others (a Pod)

Mostly you & the coach

A little with peers

There is real technique to teach — driver trees, assumption isolation, sensitivity testing — but the failure is mostly habit, so coaching must push the reader to apply it immediately.

How the two dials adapt to you →

What’s really going on

Your model breaks because you built an answer, not a machine that produces answers. A robust model lets inputs change without you rebuilding the house. Separate your assumptions from your formulas today. Pull every number someone might question into one place. That discipline — not talent — is the whole skill.

🔒 What you’ll build togetherUnlock by starting

A moveOpen your last model. Find every number that isn't a hard fact. Pull each one to its own tab, labeled 'assumption.'

A movePick one output cell. Trace it back through every formula to its root driver. If that takes more than two minutes, the model is not driver-based yet.

A moveBefore building the next model, write down the three or four drivers that actually move the outcome. Build the skeleton from those first.

A moveChange one assumption and watch what breaks elsewhere. Fix it. Do this three times before calling the model finished.

A moveStop burying logic inside formulas. Name each piece of logic in plain words in its own cell, so someone else can read it without you.

What changes unlock by starting

You change one number and trust the whole model updates correctly, without checking every tab.
Someone else opens your model and understands the logic without asking you to explain it.
You stop rebuilding from scratch when the business question shifts slightly.
You build new models faster because the structure is already decided before you add detail.

One object, two jobs: a public answer to a real problem, and — the moment you start the chat — Aurelius’s live plan for your version of it.