For consultants structuring a client problem
Issue Tree Template and Examples
Break a client problem into a SMART key question and a three-level issue tree, checked for gaps and overlaps, with a hypothesis for every leaf.
Building the tree
Four ways to split a branch of an issue tree
Once the key question is agreed, each level of the tree splits a branch into parts that don't overlap and together cover it. Charles Conn, a former McKinsey partner, starts with a profit tree in nearly every business he looks at1, which is a split by arithmetic. He also likes to try two or three different cuts, because each gives a different insight.
| Split | When it fits | How to check it |
|---|---|---|
| By arithmetic | The branch is a number with a formula behind it: profit is revenue minus costs, and revenue is price times volume. | The parts add or multiply back to the number above, with nothing left over. |
| By process step | Work passes through the same stages in order, as a hire moves from approval to sourcing, interviews and an offer. | Every day of delay, or every case that drops out, sits in exactly one stage. |
| By segment | Results may differ by customer, product, region or channel, such as chain accounts against independent stores. | Split on one attribute that each case has exactly one of, and check that the segments sum to the total. |
| By opposites | Nothing adds up or flows in order, but a clean either-or splits the branch: inside or outside the client's control, can't or won't. | Ask whether anything could be both or neither. If reasons overlap, check them in a fixed order and count each case under the first that applies. |
Use one kind of split per level, though the next level down can switch: a margin tree might split by arithmetic, then by customer segment. For a yes-or-no decision, the branches are the conditions that must all hold for a yes, so test the one most likely to fail first. The examples are illustrative.
How it works
What is an issue tree, and how do you build one?
An issue tree breaks one client question into smaller questions until a single analysis can answer each. Agree that key question with the client first: in a 2019 McKinsey podcast, senior partner Hugo Sarrazin called defining the problem the most important step1. Make it SMART (specific, measurable, action-oriented, relevant and time-bound), as Umbrex, a network of independent consultants, advises in its guide to McKinsey's problem-solving approach2. Then note who will act on the answer, how soon they need it and what is off limits.
A diagnostic tree asks why something happened, and a solution tree asks how to reach a goal. Arnaud Chevallier, a strategy professor at IMD business school, gives the general rule: know the why before the how3, so diagnose first if the cause is unknown. Split each level one way, as in the table above, and write every branch as a short question or statement, never a single word. Consultants call a split with no overlaps and no gaps MECE, for mutually exclusive and collectively exhaustive, an abbreviation Barbara Minto introduced at McKinsey4.
Give each leaf a hypothesis, the result that would prove it wrong, the analysis that tests it and the data it needs. Rank the leaves as Charles Conn does, by how much each matters and how much you can move it1. Keep what he calls a one-day answer, the answer you would give today, and rewrite it as each result comes in. River can draft the first tree from your client's problem, and the pyramid principle template turns your findings into a storyline.
How it works
From client problem to workplan
Download it or open it in River
Download both Word documents free, or click Edit with AI to open them in River, with or without your client's problem pasted above.
River drafts a first tree
If you pasted the problem, River writes the key question and a three-level tree, with a hypothesis per leaf and missing facts in brackets.
Check the split
Run the six overlap and gap checks, including three real cases only you can test, and redraw any branch that fails.
Plan the work
Add owners and dates, then ask River to turn the hypotheses table into a workplan or a data request for the client.
| Leaf | Hypothesis | Analysis and data |
|---|---|---|
| 1.1.1 Deeper discounts to large accounts | Discounts to the top 20 accounts rose after a competitor arrived | Discount by account by quarter · invoice data, finance |
| 1.2.2 Lost supplier rebates | Buying by depot dropped us below rebate tiers | Rebates by supplier, before and after · supplier agreements |
| 2.1 Delivery costs | Two small depots run longer, emptier routes | Cost per drop by depot · fleet costs, route plans |
| 3.1 One-off charges | A stock write-down explains part of the fall | Margin bridge with one-offs shown · general ledger |
Checked: every cost sits in one branch, and the branches add back to the margin.
What you'll get
Key question, branches, hypotheses and data, plus six examples
- A SMART key question and three levels of branches
- A plain-words check for overlaps and gaps
- A hypothesis, analysis and data source for every leaf
Before and after
Your client's problem in. A first tree out.
Margins at our client, a regional food distributor, dropped from about 6% to just over 4% this year. Sales are still growing. The CFO thinks it's the new chain accounts. The board meets in March.
Key question: Why has operating margin fallen from about 6% to just over 4% in [period], while sales kept growing? Tree type: Diagnostic (why) 1. Are we earning less on each sale? 1.1 Are we selling at lower prices? 1.1.1 Have discounts to the new chain accounts deepened? Hypothesis: chain accounts get deeper discounts than other customers (the CFO's view, per your notes). Data: invoice-level sales by account (finance). To confirm with the client: exact margin figures and period; any one-off charges this year.
Brackets and the to-confirm list mark what only your client knows. River never fills them with guesses.
What lands in your doc
What's inside the issue tree template
- Issue Tree Template
The fill-in doc in eight sections, from the key question to the answer so far and what to confirm. - A SMART key question
One question with a five-part check, plus context, what's off limits and what a good answer looks like. - A why or how tree
A diagnostic or solution tree, three levels deep, with four ways to split each branch. - An overlap and gap check
MECE in plain words: six checks to run at every level, including three real cases. - A hypothesis for every leaf
What you expect, what would prove you wrong, the analysis, the data and who holds it, and an owner. - Issue Tree Examples
Six worked trees, from falling margin and slow hiring to a yes-or-no decision, plus a broken tree fixed.
Questions consultants ask
Common questions
What is an issue tree, and how is it different from a hypothesis tree?
An issue tree is a diagram that splits a key question into smaller questions, level by level, until one analysis can answer each. A hypothesis tree splits it into statements to prove or disprove. McKinsey's approach, as Umbrex describes it, prefers hypotheses because they focus the analysis2, but starts with an issue tree when the team knows too little to state them.
How do I write the key question for an issue tree?
Write one question whose answer tells the decision owner what to do, with a number and a date. Umbrex's guide to McKinsey's method2 gives the example “How can XYZ Bank close the $100 million profitability gap in two years?” In a 2019 podcast, McKinsey's Hugo Sarrazin sharpened “Can we grow in Japan?”1 into a question about one product, segment or channel there.
Should I build a diagnostic or a solution tree?
Build a diagnostic (why) tree when the client knows something is wrong but not why, such as falling margin. Build a solution (how) tree when the goal is clear and the question is how to reach it. Many projects need both, in that order. For a single failure, such as a recurring defect, a root cause analysis fits better.
How many levels should an issue tree have?
Usually three. Stop splitting when one analysis, done in days, could answer a leaf. If a leaf would still take weeks, split it again. If it is too small to need an analysis of its own, fold it into a neighbor. Most branches split into two to five parts; more than that often means two different splits are mixed together.
How do I turn an issue tree into a workplan?
Pick the leaves that could change the answer most and that the client can act on, and park the rest. Give each one you keep an analysis, end product, data source, owner and date. Former McKinsey partner Charles Conn says a 50-page work plan stretching out to three months1 is always wrong, because what the team learns soon outdates it.
Is this issue tree template free?
Yes. Download both Word documents, the fill-in template and a second with six worked examples, with no account. Edit with AI opens them in a free River workspace, where River can draft a first tree from your client's problem. Paid plans add more AI use and access to the most advanced models.
What does Edit with AI do?
It opens both documents in River, ready to fill. If you pasted your client's problem, River drafts the key question, a three-level tree and a hypothesis for every leaf, using only your notes. Facts your notes don't give stay in brackets and go on a list to confirm with the client. Under River's privacy policy, nothing you paste is used to train AI models.
Your next step