Decisions
15 min

Cynefin

Sort the situation into clear, complicated, complex or chaotic, because each one rewards a different response and punishes the other three.

Time cost
15 min
Output
A named domain and the response it implies.
Steps
5

Use when

  • You do not know whether to bring in an expert or run an experiment.
  • A method that worked before is failing and you cannot see why.
  • A group is arguing about approach rather than about facts.

Do not use when

  • The domain is obvious. Naming it wastes the meeting.
  • You want a decision, not a diagnosis. Cynefin tells you how to decide, not what to decide.

Inputs required

  • A description of the situation
  • Knowledge of whether cause and effect are visible in advance

Procedure

  1. 01

    Test for clear

    Cause and effect are obvious to anyone. Sense, categorise, respond. Apply the known practice and move on. The risk is complacency: clear situations shift without announcing it.

  2. 02

    Test for complicated

    Cause and effect are knowable but need analysis or expertise. Sense, analyse, respond. Bring in someone who has seen this before. The risk is expert bias — specialists see their own domain.

  3. 03

    Test for complex

    Cause and effect are visible only afterwards. Probe, sense, respond. Run small, safe-to-fail experiments and amplify what works. The risk is treating this as complicated and buying analysis that cannot exist yet.

  4. 04

    Test for chaotic

    No discernible cause and effect. Act, sense, respond. Do something to establish stability first, then look for pattern. The risk is staying in command mode after the situation has calmed.

  5. 05

    Name the confusion

    If you cannot place it, you are in the confused centre. That is a real state and the response is to break the situation into parts and place each part separately.

Characteristic failure mode

Treating a complex situation as complicated. It looks like diligence — more analysis, more experts — and it produces confident plans for a system that will not hold still long enough for them to be right.

Worked example

A clinic’s patient no-show rate has climbed over six months.

  1. 01Not clear: no obvious cause, and the previous fix stopped working.
  2. 02Not complicated: three analysts would give three different answers, because behaviour depends on factors that interact.
  3. 03Complex: probe with four small changes — reminder timing, booking window, transport information, a waitlist — run each on a subset.

Result

Two probes move the number, two do not. Amplify the two. No analysis performed before the probes could have identified which two.

Where to go next

Also cited by
Issue trees and MECE decomposition · OODA loop