Decide Now, Rationalize Later

I took a course at Stanford called Decision Analysis. We employed systems analysis and statistical models to produce decision diagrams and stuff. This rubbish is wonderful for management consultants hired to evaluate M&A deals, but in the real world all it does is give the illusion of certainty.

It’s important to make a distinction between risk and uncertainty.

• If I know the outcome of an event, that is certainty. If I eat the blowfish in the office fish tank, I will die.
• If I don’t know the outcome of an event, but I know the probability of each possible outcome, this is risk. If I buy a lottery ticket I have a 0.000007% chance of jackpot.
• If I don’t know the outcome of an event nor the probability, this is uncertainty. If I invest in startup A, I could make a 100x return, or I could lose my money.

His name is Fugu and he is DERICIOUS.

Startup investing is pretty low-risk if I end up making 100x. But that’s uncertain. As are most things in the world.

In the face of uncertainty, decisions are necessarily made with incomplete rationale. The best decisions are driven by intuition and guidelines.

Professional decision-makers have to act like the universe operates with known probabilities and outcomes, because it’s their job to produce a justifiable conclusion. It would be unacceptable for a McKinsey consultant to come back to the managers and say, Yeah, my spidey senses tell me this is a very good deal. You guys should totally go for it.

Intuition cannot be defended in a group setting. In uncertain situations, committee decisions are tyrannized by the individual who assigns the highest probability to adverse outcomes. We’ve seen what happens when products are designed by committee.

How can a leader drive intuitive decisions?

One solution is to take the dictator approach, much like Steve Jobs: The iMac will be the exact shade of the water at Bondi Beach because I say so.

A less antagonizing solution is to just invent rationale: I chose Bondi Blue because I tested 41 shades on 25,000 users and created these fancyass statistical models. Who’s actually gonna check your work?

VCs are especially good at this, because early-stage startups are completely uncertain. Funding decisions are based on whether they like the founders, but justified with probable cause: We measured your traction with our data-driven dipstick and it appears to be a quart low. Also we need evidence that this will scale.

We like you, but we’re not sure if it will scale. Come back when you have more traction.

Can’t argue with logic and statistics.

See Also: