Issue tree, with worked examples
An issue tree breaks one problem into the smaller questions that answer it, branch by branch, so the branches cover the whole problem with no overlap. In a case interview it is the structure you lay out before asking for data.
Why tree: find the cause
A chain of 40 bakeries made $4.0M of profit last year and $3.2M this year. Why?
Profit is revenue minus costs, so the first split can't overlap or leave anything out.
Why did profit fall 20%?
$4.0M to $3.2M
Revenue
Down 2.5%
$40.0M to $39.0M
Number of stores
Flat
40 both years
Transactions per store
Down 2.5%
100,000 to 97,500 on average
30 stores
Flat
100,000 both years
10 stores near a new competitor
The cause
100,000 to 90,000
Average ticket
Flat
$10.00 both years
Costs
Barely moved
$36.0M to $35.8M
Ingredients
Fell with sales
$8.0M to $7.8M, 20% of sales
Labor
Flat
$16.0M both years
Rent and other
Flat
$12.0M both years
The answer
Ten stores lost a tenth of their customers to a new competitor. That is $1.0M of sales, and because only ingredients fell with it, $0.8M of profit: a 2.5% dip in sales became a 20% fall in profit.
Three kinds of issue tree
Same rules, different first question. The bakery above is a why tree. The other two are worked below.
Why tree
Why did this happen?
Splits a result into its parts until one part explains it. Use it for profit drops, churn and delays.
How tree
How could we get there?
Splits a goal into the levers that reach it. Use it for growth targets and cost cuts.
Hypothesis tree
Is this answer right?
Starts from the answer you expect and lists what must be true for it to hold. Use it for yes or no decisions such as entering a market.
How tree: find the options
A gym chain wants to double its revenue in five years. How?
Revenue is clubs times members per club times revenue per member. Each branch is a lever you can size.
How do we double revenue?
About 15% a year for five years
More clubs
Open new clubs
Payback per new club
Buy a regional chain
Price per member bought
More members per club
Sign more people up
Sign-ups per month, cost per sign-up
Keep members longer
Months the average member stays
More revenue per member
Membership price
Price against nearby gyms
Extras
Share who buy training or classes
The answer
Size each lever, then check that they add up to double. If price and retention give about 40%, the club count has to carry the rest.
Hypothesis tree: test an answer
Should a European e-bike maker enter the US through independent bike dealers?
Start from the answer you expect. Each branch is something that must be true for it to hold.
Yes, through dealers, if all three hold
The market is worth it
Big enough
Premium commuter sales by city
Growing
Sales growth over three years
We can win
On product
Price and features against US brands
On rules
Changes needed for US e-bike classes
Dealers pay back
Dealers want us
Interest from 20 dealers
Margin after costs
Tariffs, shipping and dealer margin
The answer
Test the branch most likely to break first. If dealers won't carry the bike, the rest doesn't matter.
How to build an issue tree
Five steps, with the bakery example.
1. Write the question
One sentence the client would sign off on, with a number and a time: why profit fell from $4.0 million to $3.2 million this year.
2. Split the top cleanly
Use a formula (revenue minus costs) or a clean either-or (existing customers or new ones). The first split decides whether the whole tree can be MECE.
3. Go down until you can test
Keep splitting until each branch can be answered with one piece of data. Two or three levels is usually enough in an interview.
4. Say where you would start
Mark the branch you think holds the answer and why. An interviewer wants a tree that points somewhere, not a list of everything.
5. Prune as the data comes in
When a branch comes back flat, say so and drop it. In the bakery example costs barely moved, so the work goes into revenue.
Mistakes that break a tree
What interviewers push on, and the fix.
Branches that overlap: price, competition and marketing
Competition shows up in price and in volume. Split on the formula first, then ask what moved each part.
A memorized framework pasted on top
Interviewers notice. Build the split from this client's own numbers, then borrow buckets from a framework if they fit.
Ten branches at the first level
Two to four at each level. More is a list, not a structure.
Leaves you can't check: customer sentiment
Make each leaf a question with data behind it: repeat visits per month.
No starting point
End with the branch you'd test first and the data you'd ask for.
What the firms say
From McKinsey and Bain's own pages, read October 15, 2026.
McKinsey
A McKinsey consultant's post on its careers site calls an issue tree a tool the firm uses to structure problem solving, breaking the problem into mutually exclusive and collectively exhaustive parts.
McKinsey careers blogMcKinsey Insights
In McKinsey's seven-step problem-solving method, step two is to disaggregate the problem with logic trees, and the classic example is the profit tree. Step three is to prioritize the branches.
McKinsey, September 2019Bain
Bain's own sample case tells candidates to take one to two minutes to think about how they would approach the problem, and to always structure their thinking.
Bain FashionCo sample caseBain
Bain says it is less interested in a textbook structure than in how candidates adapt their thinking to the case.
Bain careers blog
Lay out your tree out loud
Drawing a tree alone is the easy half. Practice a case with an AI interviewer that asks for your structure, pushes on gaps and overlaps, and scores it.
Questions people ask
What is the difference between an issue tree and a framework?
A framework is a ready-made set of buckets, such as the 4Ps. An issue tree is the structure you build for one problem. A good tree often borrows a framework's buckets, but it starts from the client's question.
Is an issue tree the same as a logic tree?
Mostly. Logic tree is the general name for breaking a problem into parts. Issue tree, driver tree and hypothesis tree are names for particular kinds.
How does MECE relate to an issue tree?
MECE is the rule each split follows: no overlap, nothing left out. The tree is what you get when you apply it level after level.
How many branches should each level have?
Two to four. Fewer usually means the split hides a second idea, more means it is a list. A formula split often gives exactly two or three.
Should I draw the tree in the interview?
Yes, on paper, and turn it so the interviewer can see it. Then talk through the top level first and only go deeper where your hypothesis points.
How long should I take to build it?
Bain's sample case suggests one to two minutes to think before you present. Then give the top level and the branch you would test first. A tree that takes several minutes to explain is too big for the first pass.