MECE in 2026: Who Invented It, And What Overlap Costs

Who invented MECE, and what Barbara Minto actually claims

The acronym has an author, and she is quoted on it exactly once in a place you can reach. McKinsey's own alumni publication carries Barbara Minto saying: "I invented it, so I get to say it how to pronounce it."

Then she amends it in the next breath. "Well, it wasn't really I, actually, it was Aristotle," she says, "but I was the first one to abbreviate it and apply it to analyzing groups of ideas."

Read what that claim covers and what it leaves out. She claims two things: the abbreviation, and its application to groups of ideas. She disclaims the logic itself, which is the law of the excluded middle and the principle of exhaustive division, both much older than consulting. Her own claim is therefore narrower than almost every use the term is now put to.

The same page gives the chronology. Minto was at the firm from 1963 to 1973, first in Cleveland and then in London. Her first manual was titled "Skillful Writing through Structured Thinking". The firm records The Pyramid Principle as published in 1985, and a self-published edition catalogued by the Internet Archive is dated 1981, under the imprint Minto International.

The odd part is where the word is absent. Her own website states the pyramid concept in full and never uses the words MECE, mutually exclusive or collectively exhaustive, on the live page and in captures from 1999, 2008 and 2011. The book text itself sits under controlled digital lending and returns a refusal, so which edition first prints the acronym cannot be checked from a primary source. Checked 6 October 2026.

So the honest statement is this. The author of the acronym is known and quotable, the logic predates her by about two millennia by her own account, and the first printed appearance of the five letters is not verifiable from anything we could open.

The two words do different jobs

Mutually exclusive is a property of pairs. Take any two buckets and ask whether a single item could belong in both. If yes, the cut leaks, and every total built on it is inflated.

Collectively exhaustive is a property of the set against its parent. Add the buckets and compare the sum with the thing you started from. If they differ, something is missing, and the missing piece is often where the answer lives.

Two checks, two different gestures.

1. Exclusivity check. Take one real item and try to file it in two buckets at once. One example that fits twice is a proof of failure.
2. Exhaustiveness check. Sum the buckets. The sum has to equal the parent quantity, in the parent's unit.
3. Residual check. If the sum falls short, name the gap as its own bucket and label it unexplained. A named residual is honest and a hidden one is not.
4. Dimension check. Confirm every bucket was cut on the same dimension. Mixing geography with product line produces a set that leaks on both sides.

The third habit is the one that separates a working structure from a tidy one. Calling a residual "other" hides it; calling it "12 percent unexplained, to be resolved by Thursday" puts it on the work plan.

Eight published sources, eight different objects

Once you start reading who writes the acronym down, the interesting variable is what each one applies it to. These are published sources, read on 6 October 2026, with the object each one attaches the test to.

Minto's object, groups of ideas, is the narrowest on that list. The widest is the Deloitte reading, which detaches the two halves: mutually exclusive becomes a property of the tasks you split a problem into, and collectively exhaustive becomes a property of the solution they add up to. That is a different test from the one Minto described, because a set of tasks can be non-overlapping while the solution they produce still misses something.

The practical consequence for a case interview is that nobody can tell you what your interviewer means by the word. If they mean the categories of your answer, they are checking your partition. If they mean the solution, they are checking your coverage of the client's problem, which is a harder standard and the one our own rubric weights heaviest.

One more pattern in that table, and it is about credit. Of the eight, exactly one names an author, and it is the firm she worked for. Columbia's page even quotes Minto's pyramid sentence verbatim and attributes it to "a McKinsey consultant", unnamed. The idea has outrun its author inside a single generation of institutional retelling.

How strict collectively exhaustive really is

Mutually exclusive is unambiguous: an item appears in one bucket or the structure is broken. Collectively exhaustive is where the published definitions split, and the split changes what a reader is allowed to leave out.

Three readings, all from published sources read on 6 October 2026.

1. Strict partition. Dartmouth's career guidance says to "slot every possible type of patron into exactly one category without any exclusion or overlap". The VCU textbook makes it a rule: "every potential consumer in the market must belong to a group, and only one group." Under this reading there is no residual category and no judgment about relevance.
2. Scoped to the problem. The University of Minnesota version asks that the pieces "cover the whole scope of the problem". BCG's data governance paper asks that domains "collectively encompass all possible data that can be collected". Both bound exhaustiveness by a stated universe rather than by everything conceivable.
3. Hedged openly. Bain states the two conditions in full and qualifies them: the structure "ideally should cover all the relevant areas without any overlap". Two hedges in one sentence, and the firm never uses the acronym at all.

The third reading is the one worth carrying into an interview, because it is the only one that survives a real problem. A revenue tree that is exhaustive over everything conceivable includes buckets nobody will ever act on. A cut that is exhaustive over the relevant universe, with the universe stated out loud, is both finishable and checkable.

So say the universe. "I am splitting the lost accounts into five reasons, and by lost accounts I mean the 1,200 that cancelled in the twelve months to June." That one clause converts an unverifiable claim about exhaustiveness into a claim a client can check, and it is the move the next three sections depend on.

A churn cut that sums to 145 percent

A business software company lost 1,200 accounts last year. Average revenue per account is 8,000 dollars, so 9.6 million dollars of annual recurring revenue walked out. The account team records a reason on every loss, from a five-option list.

The five reasons sum to 1,740 against 1,200 accounts lost. That is 540 more entries than there are accounts, so 45 percent of the volume is counted more than once. The cause is visible in the labels: an account that left for a cheaper competitor is recorded under price and under competitor won, because both are true.

The set fails exclusivity badly and it is not obviously exhaustive either. An account that signed and never deployed appears nowhere in the list, and those accounts do exist.

Recutting it so the pieces add up

Fix it by choosing one dimension and one only. Lifecycle stage works here, because every lost account sits at exactly one point in its own history.

Add them: 180 plus 395 is 575, plus 310 is 885, plus 240 is 1,125, plus 75 is 1,200. The cut closes on the parent.

Notice what the recut surfaces. The second bucket, 395 accounts that lapsed without anyone having a renewal conversation, was invisible in the reason-code cut because no reason code covers "nobody called". It is now the largest single bucket at 32.9 percent, and it is the cheapest to act on.

What the overlap cost, in dollars

Both cuts describe the same 1,200 accounts. They price the price problem differently, and the difference is what an overlapping cut costs a client.

Under the reason-code cut, price is 480 of 1,200 accounts, which is 40 percent, so 40 percent of 9.6 million dollars is 3.84 million dollars of revenue attributed to price. Under the lifecycle cut, price is the commercial-terms bucket, 310 accounts, which is 25.8 percent, so 25.8 percent of 9.6 million is 2.48 million dollars.

The gap is 1.36 million dollars. The overlapping cut overstates the price problem by 3.84 divided by 2.48, which is a factor of 1.55, so 55 percent.

That number has consequences a client feels. A 3.84 million dollar price problem justifies a discount policy review and probably a list price change. A 2.48 million dollar price problem, sitting next to a 3.16 million dollar problem of accounts nobody called, justifies hiring renewal coverage instead. The 395 lapsed accounts are worth 395 times 8,000, which is 3.16 million dollars, and they were not on the original list at all.

So the overlap did two kinds of damage. It inflated one bucket by 55 percent, and it concealed a larger one entirely. The first error is arithmetic and the second is exhaustiveness.

Where a firm writes the word down for candidates

The three firm documents in the table above are client-facing publications. On recruiting pages, which is where a candidate goes looking, exactly one firm writes the word down, and where it puts it is more interesting than the fact that it does. On the recruiting pages read here, McKinsey is not that firm. MECE and McKinsey get quoted as a single phrase, and the only reachable Minto words on the acronym sit in an alumni interview rather than in anything the firm tells candidates.

Deloitte's LeadAuto case, published on its own recruiting site, splits every model answer into two tiers: "A good answer may include" and "In addition to the above, a great answer may also include". A MECE internal and external split appears in the second tier of question one, checked 30 September 2026 at usrecruiting.deloitte.com. The firm treats a clean cut as an upgrade above a passing answer.

Its general case guidance never mentions a framework at all. It publishes a five-step approach and writes that candidates are "not expected to provide the "right" answer, but rather to clearly convey your logic and thought process", checked 30 September 2026 at deloitte.com.

Kearney takes the opposite line on imported structure while describing the same discipline. Under a tip it calls thinking top down, the firm writes: "Understand the overriding issues, and use them to prioritize and organize issues that have lesser consequences. Try to avoid forcing your solution to fit a standard framework", checked 30 September 2026 at kearney.com. Prioritizing and organizing issues is what a clean cut delivers. The firm wants the property and rejects the template, which is a distinction worth holding onto. Framework mistakes works through the rest of that disagreement.

Oliver Wyman publishes six evaluation questions and a six-step method in which breaking the problem down is a named step, checked 30 September 2026 at oliverwyman.com. None of the six criteria mentions a framework either.

Cuts that pass MECE and still fail

A MECE structure is necessary and nowhere near sufficient. Exclusivity and exhaustiveness are properties of a partition, and a partition can be clean and useless at the same time. Four cuts that pass both tests and still produce nothing.

1. Accounts by first letter of company name. Perfectly exclusive, perfectly exhaustive, and no action attaches to any bucket.
2. Accounts by contract number range. Same problem, with a false air of rigor because the buckets are numeric.
3. Accounts by whether they churned. Two buckets, closed and complete, and it restates the question.
4. Accounts split 80 buckets deep. Exhaustive, exclusive, and impossible to hold in a conversation or a slide.

The missing test is whether the buckets have different economics. The lifecycle cut above works because each bucket has a different fix, a different owner and a different cost to address. A cut whose buckets all respond to the same action has partitioned the data and not the problem, which is the difference issue trees turns on.

Where separation is scored

The four weights in the block below belong to us. They are lifted from the rubric code our grader runs, and no consulting firm publishes a marking scheme of its own, so nothing here is Deloitte's view of a clean cut or Kearney's.