Decisions
15 min

Bad outcome versus unsurvivable outcome

Separate outcomes you recover from and outcomes that end the activity, and apply different decision rules to each.

Time cost
15 min
Output
Two lists, and a constraint on the second.
Steps
5

Use when

  • A risk discussion treats all downside on one scale.
  • Before applying expected value to anything with a large loss branch.
  • Someone says “worst case” and means “bad”, or means “terminal”, and nobody has asked which.

Do not use when

  • Nothing at stake is irreplaceable, in which case everything is bad and nothing is terminal.

Inputs required

  • The list of possible outcomes
  • An honest definition of what you cannot lose

Procedure

  1. 01

    Write what cannot be lost

    Health, the reserve that covers fixed costs, a licence, a relationship, the ability to try again. Be specific — vagueness here makes every risk feel terminal.

  2. 02

    Sort every outcome into two piles

    Recoverable, and not. The test is whether you can still act afterwards, not how much it hurts.

  3. 03

    Apply expected value to the recoverable pile

    Averages are meaningful when you get to play again. Optimise here freely.

  4. 04

    Apply a constraint to the other pile

    Not a weight — a constraint. Terminal branches are removed or capped, not priced. No upside makes them acceptable.

  5. 05

    Re-examine anything that moved piles

    The useful output is usually one outcome that everyone treated as merely bad and is actually terminal, or the reverse.

Characteristic failure mode

Inflating the terminal pile. If everything is treated as unsurvivable, the constraint blocks every option and the method becomes an argument for never acting.

Worked example

A person weighing whether to leave stable employment to work independently.

  1. 01Cannot be lost: housing, and the ability to return to the field.
  2. 02Sorted: a bad first year is recoverable. Exhausting savings entirely is not, because it forces a rushed return on bad terms.
  3. 03Constraint added: a hard floor of nine months' costs, at which point the attempt ends by rule rather than by crisis.

Result

The decision becomes a bounded experiment. The terminal branch was removed by a pre-committed stopping point rather than by deciding not to try.

Where to go next