Overview
An issue tree decomposes a large problem into a branching hierarchy of sub-problems, each of which is smaller and more tractable. Done well, the branches are MECE — mutually exclusive (no overlap) and collectively exhaustive (nothing missing) — so the tree covers the whole problem without double-counting.
There are two flavours: a ‘why’ tree that diagnoses causes, and a ‘how’ tree that structures solutions. Either way, the tree turns an overwhelming question into a set of testable leaves you can assign, investigate and knock out one by one.
When to use it
A problem is too big to attack whole — revenue decline, cost overrun, any ‘why is this happening / how do we fix it’ question.
How to use it
State the problem as the root
Put the core question at the top of the tree.
Break into MECE branches
Split it into a few sub-questions that don’t overlap and together cover everything.
Decompose each branch
Repeat the split on every branch until the leaves are concrete and testable.
Check MECE at each level
Confirm no overlap and no gaps before going deeper.
Prioritise and investigate leaves
Attack the branches most likely to matter first, using data.
Worked example
‘Profit is falling.’ The tree splits into revenue and cost. Revenue splits into volume and price; cost into fixed and variable. Volume splits into new and returning customers. The leaf ‘returning-customer volume’ turns out to be the culprit — a churn problem hiding inside a vague ‘profit’ panic. The tree localised it in minutes.
Common pitfalls
- Branches that overlap, causing you to double-count or argue in circles.
- Missing branches, so the real cause hides in a gap.
- Decomposing forever instead of stopping at testable leaves.
Frequently asked questions
What does MECE mean?
Mutually Exclusive, Collectively Exhaustive — branches that don’t overlap and together cover everything.
Why vs how tree — which do I use?
A ‘why’ tree to diagnose a cause; a ‘how’ tree to structure a solution. Often you use both in sequence.
How deep should the tree go?
Until each leaf is specific enough to test with data or assign to someone.