Notes on deciding well.

Regulators and firms tend to equate informing people with helping them decide. The evidence says otherwise. Giving someone more information assumes they read it, understand it, and weight it correctly — when in fact attention is scarce, and comprehension peaks early then falls as volume grows.

What actually moves behaviour is structure: what is made salient, what the default is, how risk is framed, how much effort a good choice takes. A single well-chosen sentence can shift more decisions than ten pages of fine print.

So we test comprehension and outcomes — not page counts — and redesign the moment of choice, not just the document around it.

Some choices can be revised. The ones that matter usually can't. When a decision is irreversible, its value depends not only on the expected payoff but on what you give up by committing before you have learned enough — an option value that is invisible on a spreadsheet and routinely ignored.

The discipline is knowing when more analysis is worth it and when it is not: gather evidence while it still changes the decision, and commit the moment it stops.

We make that trade-off explicit — weighing the cost of waiting against the cost of being wrong — so the call to cross or wait is made on purpose, not by default.

You can't train your way out of a badly designed choice. Decades of experiments show that biases are systematic and predictable — losses loom larger than gains, defaults stick, the first number anchors the rest. They are not failures of character; they are features of how minds meet choices.

That is good news. What is systematic can be designed for: change the default, the frame, or the effort required, and behaviour shifts — reliably, and at scale.

We treat the choice architecture as the unit of work. Fix the structure, and the "irrational" customer, citizen, or investor starts to look a great deal more sensible.

If you can't measure it, you're deciding blind. Trust, risk perception, attention, disorientation — the variables that actually drive decisions are usually asserted, not measured. We build and statistically validate instruments that turn them into numbers you can act on.

The same rigour applies to systems: how information, reputation, and behaviour really move through a market or a network — rather than how an org chart says they should.

Measuring first is slower for a week and faster for a year. It stops teams from optimising the wrong thing.