Model Thinking
Model thinking runs a decision through several different mental models at once, so you can see where they agree, which raises confidence, and where they clash, which flags real uncertainty.
Five steps carry a real decision through several models in turn, starting with the decision itself and ending with a call made from where the models agree and where they diverge.
Reach for this when…
- A big call keeps getting made on a single favourite framework.
- Smart people in the room disagree and you suspect they're each using a different lens.
- You want to stress-test a decision before committing budget to it.
How to run it
- Take the real decision or problem in front of you.
- Apply at least three genuinely different models to it, not three versions of the same one.
- Write down what each model predicts or reveals separately.
- Mark where the models agree and where they diverge.
- Use the pattern of agreement and divergence to decide, and to know where to watch closely.
A worked example
Situation. Arjun Mehta was deciding whether to expand Setu, his SaaS scheduling tool, from Bengaluru into Hindi-speaking North India, working off a single competitor-comparison spreadsheet.
Applied. He ran the decision through three models: a diffusion-of-innovation view of adoption speed, a network-effects view of switching costs, and a simple unit-economics model of support cost per new region; diffusion and unit-economics both argued for caution, network effects alone argued for speed.
Result. He saw his one confident model was the outlier and slowed the launch by a quarter to fix onboarding cost first. The delayed launch converted at twice the rate.
The catch
It is easy to run three models badly and call it rigour, when really you have picked three that all agree with your gut. Models simplify by design, so the exercise can manufacture false precision if you forget what each one leaves out. It takes real fluency across models to run this well, which most people do not have.
If every model you chose happens to agree with the decision you already wanted, check which models you left out.
Origin: Scott Page (University of Michigan)