Talk it through with Aurelius
Library›Aurelius›The problem
Aurelius · Work & Leadership
Knowledge + Guidance

Why do my models break down when the data runs thin?

You can run the numbers on most things now. Your team builds a model, it holds, the client nods, everyone moves on. Then a case comes in with almost no history behind it — a new market, a rare customer, something nobody has seen five times before. The model still gives an answer. It still sounds sure of itself. That is the moment that should worry you. Not because you lack skill, but because the skill you have was built where mistakes get caught fast and corrected faster. Thin data doesn't forgive you that way. You can be wrong for a quarter, a year, a whole client relationship, before anyone finds out — including you. What you owe your team, and the people depending on your answer, isn't a better guess. It's honesty about which parts of the model are load-bearing and which parts are hope wearing the clothes of analysis. Say which is which, out loud, before it ships. That part is within your power today.

◆ How this problem reads on the two dials
GuidanceKnowledge
More coaching
Some to learn
1:1 with AureliusWith others (a Pod)
Some one-to-one
Practise with peers
The concept is simple to explain once; the hard part is building a team habit of naming thin-data moments before deciding, which takes practice, not more teaching.
How the two dials adapt to you →
What’s really going on

The problem isn't your skill — it's that your instincts were trained where data is thick and mistakes get caught fast. Thin data breaks that loop. You can be confidently wrong for a long time before anyone notices. Say that out loud before you guess, not after.

🔒 What you’ll build togetherUnlock by starting
A moveBefore you present a model, make someone say out loud which numbers come from five data points, not five thousand — say it to the client, not just the team.
A moveWhen you hit a thin-data case, stop chasing the best guess. Ask instead: what is the worst defensible answer? Write that floor down first.
A moveRun a two-minute pre-mortem: if this number is wrong, how long before anyone notices — a day, a quarter, never? Treat 'never' as a live wound, not a small one.
A moveAssign one person on the pod to argue the case for being wrong, out loud, before the model goes out the door. Rotate who plays that part.
A moveAfter the decision, write down the one piece of evidence that would have changed your mind. Check it against what actually happens, later, without excuses.
PractiseName the Thin Ice · a Pod of 4 · 30 min

What changes unlock by starting

  • Your team can point to exactly which parts of a model rest on thin ice, in real time.
  • Confident-sounding guesses stop passing for solid results in your reviews.
  • Thin-data decisions get flagged and revisited instead of quietly forgotten.
  • Clients hear the truth about uncertainty before it becomes an expensive surprise.

Related problems

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.