Foundations
Issue trees: how to break down any business problem
Updated 4 min readBy the MECE Editorial Team
Short answer
An issue tree is a diagram that breaks one business question into smaller questions, level by level, so each branch can be analyzed on its own. Every level should be MECE: no overlaps and no gaps. Diagnostic trees ask why something happened. Solution trees ask how to fix it or reach a goal.
Key takeaways
- Start from one precise question, not a topic. "Why did revenue fall 12% this year?" beats "revenue".
- Diagnostic trees find causes; solution trees generate options. Know which one you are drawing.
- Each level is MECE. Three or four branches per level is plenty.
- Stop splitting when a leaf is testable with data you could actually get.
- Lead with a hypothesis and test the most likely branch first.
What is an issue tree?
An issue tree starts with one question on the left (or top) and splits it into sub-questions that together fully answer it. Each sub-question splits again until the ends of the branches are things you can check. Consultants draw one at the start of almost every problem, because it turns a vague worry into a work plan.
Why did a meal-kit company's revenue fall 12%?
Number of orders
- Orders from new subscribers
- Orders from existing subscribers (retention × order frequency)
Revenue per order
- List price
- Discounts and promotions
- Mix of meal plans
What is the difference between a diagnostic and a solution issue tree?
| Diagnostic tree | Solution tree | |
|---|---|---|
| Question type | Why? ("Why did margin fall?") | How? ("How can we raise margin 3 points?") |
| Branches are | Possible causes | Possible actions |
| Best first split | An equation or process that must explain the change | Levers the client controls |
| You are done when | Data confirms or rules out each cause | Options are sized and ranked |
How can a regional airline raise profit per flight?
Raise revenue per flight
- Fill more seats
- Raise the average fare
- Sell more extras (bags, seat selection)
Cut cost per flight
- Fuel
- Crew
- Airport fees and maintenance
- Overhead
How do you build an issue tree?
Write the key question
Make it specific and measurable: what, how much, by when. "How can the chain restore a 10% operating margin within two years?"
Pick the first split
Use an equation or a process if one exists. They are MECE by construction. Otherwise use opposites or a single segmentation dimension.
Split each branch again
Keep going until each leaf is something you can test with data you could realistically get.
Check every level
Run the overlap, gap and sum tests on each level. See What is MECE.
Prioritize with a hypothesis
Say which branch you think explains most of the problem, and why. Test it first.
Turn leaves into analyses
For each leaf, write the data you need and what result would prove or disprove it.
What is a hypothesis-driven issue tree?
A hypothesis tree starts from your best guess at the answer and lists what would have to be true for it to hold. It is faster than exploring every branch, because you only test the conditions that matter. If one condition fails, you revise the hypothesis.
Hypothesis: the meal-kit company should stop its 60%-off first box
Discount customers churn much faster
- Compare 90-day retention of discount vs. full-price cohorts
Sign-ups will not fall more than churn savings
- Test a smaller discount in two markets
Competitors will not use it against us
- Check rivals' offers and how customers compare them
What mistakes make issue trees fail?
- A topic instead of a question at the root, which gives you branches that describe rather than answer.
- Mixing causes and actions in the same tree.
- Too many branches at one level. Group seven ideas into three buckets.
- Branches that overlap, so the same cost or customer shows up twice.
- Stopping too early. "Operations" is not testable; "pick rate per worker-hour" is.
When should you use an issue tree outside interviews?
Any time a problem is fuzzy or a team disagrees about where to look. Issue trees scope projects, split work across people without overlap, structure root-cause reviews after an incident, and organize a memo before you write it. A tree on one page is also a fast way to get a manager to agree on what you will and will not analyze.
Practice this with a live case
Reading builds recognition; solving builds skill. Each case below runs with MECE's AI interviewer, which answers your clarifying questions, pushes back and scores you out of 100.
Profitability · B2B software · medium
A SaaS company's gross margin is sliding Practice liveProfitability · Travel & leisure · medium
A Colorado ski resort's profits have halved Practice liveGrowth strategy · Telecom · hard
A cable company is losing broadband customers to 5G home internet Practice live
Browse all 50 case interview examples and 50 market sizing questions.
Frequently asked questions
What is the difference between an issue tree, a logic tree and a driver tree?
They are close cousins. Logic tree is the general term. A driver tree is a numeric version where branches add or multiply (revenue = price × volume). An issue tree is the question-based version consultants use to plan an analysis.
How many levels should an issue tree have?
Two or three levels in an interview. In a real project, keep going until each leaf is a specific analysis you can run.
Is an issue tree the same as a case framework?
A framework is a reusable starting tree, such as the profitability framework. An issue tree is the version you build for a specific question, often starting from a framework and then tailoring it.
Keep learning
- FoundationsWhat is MECE? The principle behind structured problem solvingRead the guide
- Case frameworksThe profitability framework for case interviewsRead the guide
- FoundationsThe Pyramid Principle: how to give a clear recommendationRead the guide
- FoundationsCase interviews: how they work and how to prepareRead the guide
MECE (mece.in) is an AI practice platform for case interviews and business problem solving, named after the consulting principle Mutually Exclusive, Collectively Exhaustive. It is not affiliated with McKinsey or any other consulting firm. Companies in worked examples are fictional and their figures are illustrative.
