(Second Thought)

A small toolkit for working through a hard decision when you have an hour and a quiet room.

Before anything else

Can this be reversed? Most decisions can. Important decisions that can't be reversed deserve careful attention — slow down, write things down, sleep on it.

Could it ruin me or my organization? If yes, no expected-value calculation matters. Use the tail-risk lens. Are you sure you're in a normal-distribution world?

What am I assuming about how this turns out? Most decisions rest on an implicit forecast. Make it explicit before you commit. A short discipline for forecasts you can actually score.

What kind of decision is this?

Choosing among defined alternatives
Candidates, vendors, properties, deals. The question is which one.
The shape of outcomes matters most
Big upside or downside, fat tails, irreversibility, payoffs that bend sharply somewhere.
Too big to settle tonight
A real, complex question with no clear alternatives yet. Tonight's task is to get clear on what it actually requires.
Pick one to read more

The Mediating Assessments Protocol

↑ Back to top

Pick five to seven attributes that capture what matters about the decision. Score each alternative on each attribute, independently, before forming any overall impression. Only after every cell is filled in do you allow yourself a holistic judgment.

The discipline is in the order. Score first, feel second. Halo effects ruin intuitive comparisons; this protocol blocks them.

Source. Daniel Kahneman, Nobel laureate in economics, developed this protocol for Israeli army officer selection and refined it across five decades of judgment-and-decision research.

Reference. Kahneman, Sibony, and Sunstein, Noise (2021), Part VI.

When not to use it. When alternatives differ wildly in shape — when one option could ruin you and another can't — averages across attributes mislead. Use the tail-risk lens.

The Tail-Risk Lens

↑ Back to top
Mediocristan or Extremistan? →
First, recognize what world you're in. The questions below only matter once you've made the call.

When the payoff curve is bowed or ruin is on the table, expected-value thinking fails. Diagnose first, then act.

Diagnose
  • Convexity. Does variance help or hurt? If the upside is uncapped and the downside is bounded, take more risk. If the downside is uncapped and the upside is bounded, take less.
  • Ergodicity. Is this "many people each do it once" or "I do it many times"? For anything with ruin risk, the second governs and the first lies. Great average outcomes don't help if any single trial wipes you out.
Act
  • Barbell. When uncertainty is high, prefer combining extreme caution and extreme aggression to moderate exposure. Be 90% safe and 10% wild rather than 100% medium.
  • Via negativa. Subtract bad options before adding good ones. "Don't do anything stupid" outperforms "be brilliant" over long stretches.

Source. Nassim Nicholas Taleb, former options trader, has spent thirty years working on decisions under uncertainty and the asymmetric properties of fat-tailed distributions.

Reference. Taleb, Antifragile (2012), especially Book IV.

When not to use it. When the decision is small, reversible, and the distribution is well-behaved. Don't bring tail-risk thinking to lunch choices.

The Framework Builder

↑ Back to top
Open the Framework Builder →
Tonight's work is the decomposition. The tests come after — over days, with your team, on your calendar.

Some questions are too big to answer in an evening — and trying to force a conclusion anyway is how you end up with whatever you were already inclined to believe, dressed up with whatever facts were handy. The move instead is to spend the evening making the question concrete. Break the main question into three to five considerations — second-level questions that, if answered, would resolve it. Under each, list the hypotheses that would have to be true or false. That's tonight's work, and it's enough: you end the evening knowing exactly what the question requires rather than pretending you've settled it.

The tests come later — over days, with your team, on your calendar. What you've built tonight is the map of what to investigate. You can do the whole decomposition on paper. The Framework Builder page adds a prompt that drafts it for you, plus guidance on designing tests worth running when you get to them.

Source. Roots in the scientific method and the issue-tree tradition from strategy consulting (MECE — mutually exclusive, collectively exhaustive). No single canonical reference.

Reference. Barbara Minto, The Pyramid Principle (1987).

When not to use it. When you already know the question and the alternatives. You're probably in MAP territory — go there instead.

After the decision. Write down what you decided and what you expected to happen. One line in your calendar entry, the meeting notes, or a Word doc — wherever you'll find it later. Not a journal. A breadcrumb. Future you will want to know what past you was thinking.