How to Be Freakishly Decisive

Make your best decision now. Then find out whether it deserves to change.

Choose now. Revisit for a reason.
  1. Decide nowChoose with what you know.
  2. Name the gapsWhat could beat that choice?
  3. Price the researchCheck only when it pays.
  4. UpdateKeep or replace the incumbent.

Stop when another check cannot pay. New information or a new option can start the loop again.

A useful default for changing your results is to reverse the order of decision-making. Instead of researching until you feel ready to choose, choose with the information you already have. Then investigate what could make a different choice better.

You immediately have something useful: a decision you can act on, plus a reason to change it. You can be uncertain and still know which action currently looks best.

This is a decision about the next action. A reversible trial, a necessary safety check, or waiting for a specific result can be that action. Choosing immediately does not require making an irreversible commitment immediately.

Have a decision in mind? Run the six yes/no questions.

1. Choose the best action available now

Ask: “If I had to decide with only what I know now, have I picked what I would do?” If no, pick. That choice is your incumbent: the best answer you have so far.

The incumbent need not be what you were already doing. Keeping the current arrangement has to compete with changing it. Inaction has consequences too.

Define how you will compare alternatives. What outcome matters, over what period, and what would a bad result cost? Choose the option with the best expected result, weighting plausible gains and losses by how likely you think they are.

Then include switching costs: setup, disruption, lost progress, and the cost of reversing a mistake. Compare what each option costs from here. Money and effort you cannot recover are already gone; they do not make a worse option worth keeping.

2. Name what could beat it

Ask: “Could something I do not know change which option wins?” If no, you can stop researching. If yes, write a short list of those unknowns.

Each item needs an answer that would change your action. “I want to understand the market better” is open-ended. “If this tool saves more than two hours a month after setup, I would switch” gives the investigation a job.

Keep the list short by testing relevance: if either answer leaves you making the same choice, resolving that question has no immediate decision value. It may still be interesting. It is not holding up this decision.

The decision loop

Know when to stop looking

Begin with your best available choice. Compare options over the same time horizon, including the cost of switching.

Start or revisit a decision

Does an option beat your incumbent after switching costs?

YesMake it the incumbent.

NoKeep this choice.

Either way, check what could change the answer.

Could an unknown change which choice wins?

Yes

Is the next check worth more than it costs?

Compare its expected decision benefit with time, attention, money, and delay.

Yes Run the smallest useful check.

Back to 1 with what you learned
Use incumbent

Act on your best choice now. You can stop researching.

Reopen for relevant new information or a new option

3. Research when it pays in expectation

Ask: “Is the expected improvement from finding out greater than the cost of finding out?” Count the chance that the result changes your action and how much better that action would be. Subtract investigation time, attention, money, and the cost of delay.

You may rationally stop while important things remain unknown. You may also investigate a very unlikely problem if its consequences are large enough. The amount of uncertainty is not the deciding variable; its effect on your choice is.

Do the cheapest check that can usefully change the decision. That might be one call, one measurement, or one bounded trial. Afterward, price the next check again. The first useful answer does not authorize an unlimited research project.

4. Update when another choice wins

Ask: “Given the new evidence, does another option beat the incumbent after the cost of changing?” If yes, switch. If no, keep going.

New evidence can change your beliefs without changing your action. You do not need to reopen the decision every time you learn something.

Write down what you will watch and when you will check it. A useful trigger is observable: repair time exceeds the expected savings, a deadline slips, or a trial meets the switching threshold. “I feel unsure again” does not tell you whether the expected result has changed.

When you switch, the new choice becomes the incumbent. Run the same comparison from that position.

Run your decision through six yes/no questions

Use one decision. Read each question and follow your answer. When a branch says to stop, act on your incumbent and watch for the evidence you named.

  1. Have I chosen what I would do with only what I know now?

    Yes
    That is your incumbent: the choice to beat.
    No
    Pick the best available action now. Include continuing, waiting, or a small trial among the options.
  2. Have I defined how another option would beat it?

    Yes
    Use the same outcome, time horizon, and downside for both.
    No
    Write the result you want and compare expected gains and losses. Name the strongest alternative.
  3. Have I counted the cost of changing course?

    Yes
    Require the alternative’s expected advantage to cover that cost.
    No
    Count setup, disruption, lost progress, and reversal. Exclude costs already paid that neither choice can recover.
  4. Could an answer I do not have change the winner?

    Yes
    Write that question and the answer that would make you switch.
    No
    Use the incumbent. You can stop researching.
  5. Is finding that answer worth more than it costs?

    Yes
    Do the smallest useful check. Count time, attention, money, and delay.
    No
    Use the incumbent. Uncertainty alone is not a reason to keep searching.
  6. Does the new evidence make another option better after switching costs?

    Yes
    Switch. The new choice becomes the incumbent.
    No
    Keep the incumbent. Reconsider when relevant evidence arrives.

Write your result in one sentence: “I choose ___ for now; ___ would make me switch; the next check is ___, or no further research is worth doing.”

For a concrete model-selection example, see When Is a Cheaper Model Actually Cheaper?

The decision theory behind the routine

Your current beliefs are your prior. Evidence updates them through a likelihood: how probable would this observation be under each explanation? The updated beliefs are your posterior. Choose the action with the greatest posterior expected utility, including losses and transition costs.

Use ranges when a point estimate would hide uncertainty. A credible interval describes uncertainty under your model; it is not a confidence threshold you must cross before acting. Different beliefs within a wide interval can still favor the same action.

The value of a proposed check is the expected value of choosing after seeing its result, minus the value of choosing now. Subtract the cost of the check and any delay. This accounts for the possibility that you learn something and keep the same choice.

For claims about what an action will cause, make the comparison explicit: the same task and conditions with action A versus action B. Task difficulty can affect both which action you choose and the outcome, so an observed difference alone need not be the effect of switching. A randomized trial can help separate them; a handpicked example cannot establish the average effect.