Issue trees and MECE decomposition
Break a question into sub-questions that do not overlap and leave nothing out, so that the parts can be answered separately and reassembled.
- Time cost
- 45 min
- Output
- A tree with a number against each branch.
- Steps
- 5
Use when
- The question is too large to answer directly.
- Several people must work on parts of the same problem without colliding.
- You suspect the problem has one dominant driver but cannot say which.
Do not use when
- The problem is complex in the Cynefin sense — the parts interact, so decomposing them changes them.
- You already know the answer and are building the tree to justify it.
Inputs required
- One question at the top
- A way to check the branches are exhaustive
Procedure
- 01
Write the top question
One sentence, answerable in principle. “Why has retention fallen?” works. “What should we do?” does not.
- 02
Split once
Break it into three to five branches. Test each split: mutually exclusive means no fact belongs in two branches; collectively exhaustive means no fact belongs in none.
- 03
Prefer an arithmetic split
Where the question has a formula, split along it. Revenue splits into volume × price. That split is exhaustive by construction and needs no judgement.
- 04
Split again where it pays
Go one level deeper only on branches that could plausibly hold the answer. Depth everywhere is wasted work.
- 05
Size the branches
Put a rough number on each branch before investigating. A branch that cannot account for the gap does not need investigating at all.
Characteristic failure mode
Worked example
A student’s grades have fallen and the cause is unclear.
- 01Top: why did the average drop 12 points?
- 02Arithmetic split: is it fewer marks per assessment, or fewer assessments completed?
- 03Sizing: completion is unchanged, so the whole gap sits in marks per assessment.
- 04Split that branch: by subject. Two subjects account for 11 of the 12 points.
Result
The question has narrowed from “why are grades down” to “what changed in these two subjects”, in two splits and no research.
Where to go next