From McKinsey & Company problem-solving practice, given its discipline by Barbara Minto (The Pyramid Principle, 1978 edition) and documented by Ethan M. Rasiel (The McKinsey Way, 1999).
Breaks a large question into branches that do not overlap and together cover it, down to pieces small enough to investigate, producing a work plan for the analysis.
Why operating profit fell $3.6M, split until every branch can be answered
Each level splits the box before it into parts that do not overlap and leave nothing out; select a box to read the check that proves it, or a leaf to read the analysis it calls for. Then mark the three leaves you would run first. Illustrative.
Links between branches:
0 of 3 marked1 of 3 marked2 of 3 marked3 of 3 marked
An example
A specialty chemicals distributor saw operating profit fall from $18.0 million to $14.4 million in a year when sales were described internally as steady. The finance director had two weeks and three analysts before the board met.
She wrote the question at the top of a page: why did operating profit fall $3.6 million? Below it went revenue and cost, then volume and price under revenue, then customers and order size under volume, and so on until each box could be answered by one analysis with data the company had. At every level the team asked two questions: could any cause sit in two boxes, and was any possible cause missing a box. The second question added mix of products under price, which nobody had raised.
The analysts split the leaves among them, starting with the three the director judged likeliest. Within a week the answer was clear: discounts had grown from 6 to 11 percent of list price, and a drift toward small orders had raised freight per ton. Those two leaves explained $3.1 million of the fall.
Why it works
The tree turns a question too big to study into a set of questions small enough to answer, and its rules guarantee that the answers add back up to the original. If the branches at each level do not overlap, no effort is spent twice and no dollar is counted twice. If together they leave nothing out, the real cause has a branch to sit on even when nobody suspects it.
The structure also makes the work visible. Every leaf names an analysis, so the tree is the work plan, the task list, and the progress report at once, and a team can see which branches are answered and which are still open. When a finding fits no branch, that is a signal that the split was wrong, and redrawing it is part of the method.
The thinking behind it
Barbara Minto Barbara Minto, the first woman MBA hired by McKinsey, taught its consultants to write with the main point first and the supporting points grouped beneath it. She required each group to be mutually exclusive and collectively exhaustive, MECE for short, which is the test every level of an issue tree must pass. She published the method herself through Minto International as The Pyramid Principle, in a revised second edition by 1978. Her pyramid holds an argument; the issue tree uses the same grouping rule to hold a question.
Ethan M. Rasiel Ethan M. Rasiel, a former McKinsey associate, described in The McKinsey Way (1999) how the firm's consultants structure a problem: start from a hypothesis, break the problem into a MECE issue tree, and test each branch with facts. His book is a practitioner's account of firm practice. The issue tree has no single inventor; it grew up in consulting work, and Minto's grouping rules gave it its discipline.
How to run it
- Write the question as one exact sentence. Name the number, the period, and the business: why did operating profit fall $3.6M this year, not why are we less profitable.
- Split by an identity where one exists. Profit is revenue minus cost and revenue is volume times price; a split that follows an identity cannot overlap or leave a gap.
- Test every level for overlap and gaps. Ask whether any cause could sit in two boxes, and whether any possible cause has no box.
- Stop when a leaf can be answered. A leaf is finished when one named analysis, with data you can get, would answer it.
- Write the analysis on each leaf. Name the data, the cut, and the owner, so the tree becomes the work plan.
- Run the likeliest leaves first. Put the most effort where most of the answer probably lies, and check the rest quickly.
- Redraw when a finding breaks the tree. A finding that fits no branch means the split was wrong; change the tree.
Where it goes wrong
Branches that look separate share a cause. Smaller orders show up as lower revenue per order and as higher freight per ton, and two analysts end up sizing the same dollars. Spot it when the leaves add up to more than the gap, and size the shared cause once.
Depth follows interest. One branch gets split five levels down because someone finds it fascinating, while a larger branch stays at one level. Let the dollars at stake set the depth.
The tree hardens. Findings get forced into boxes drawn on the first day, and the answer comes out wrong. Treat the first tree as a guess and redraw it when the facts disagree.
When to use something else
- Pyramid Principle. When the analysis is done and the findings need to be presented as an argument, conclusion first.
- Fishbone Diagram. When the problem is a defect or failure with no clean identity to split by, and causes need gathering across categories.
- Five Whys. When the problem is small enough that one chain of causes, followed down, is likely to find the answer.
Where it sits in SOLVE eX
Go to the source
- The Pyramid Principle: Logic in Writing and Thinking, Minto International, revised second edition; the first edition is undated in library catalogs, 1978
- The McKinsey Way, 1999