MECE stands for mutually exclusive, collectively exhaustive. The phrase came out of McKinsey and has since spread to every part of finance and strategy, usually as a slogan. Its real use is narrower and more valuable: it is the property that lets the pieces of an analysis add up to the whole. When a bridge reconciles to the pound, the split was MECE. When it does not, the split is where to look.
What the two halves mean
Mutually exclusive means that any single pound, customer or event belongs in exactly one box. Collectively exhaustive means that every pound, customer or event belongs in some box. A split that has both properties is a partition: the parts sum to the total, with nothing left over and nothing counted twice.
The distinction matters because the two failures damage an analysis in different ways, and you will meet both before you have finished your first project.
The two failures, with numbers
Overlap double counts. Suppose a software company reports £40m of revenue from enterprise customers and £25m from financial services customers. Adding the two gives £65m, but total revenue is £52m. The difference is £13m from enterprise financial services clients, which sit in both boxes: 40 + 25 - 13 = 52. Any analysis that treats the two segments as separate will overstate the business by a quarter.
A gap hides the answer. A retailer splits sales by region and reports the United Kingdom at £30m, Europe at £18m and the United States at £12m. That is £60m, but the ledger says £66m. The missing £6m is online sales that are not assigned to any region. If online has been growing while the three regions are flat, the whole story of the year sits in the box nobody drew.
Overlap is usually caught by arithmetic, because the total comes out too high. A gap is caught only if you compare the parts with a total you trust. That is why the first habit of a careful analyst is to reconcile every split to a known figure before drawing conclusions from it.
Three reliable ways to cut
Almost every MECE split in practice is one of three kinds. They differ in how much you have to trust your own judgement.
| Cut | Example | Why it is MECE | Risk |
|---|---|---|---|
| An algebraic identity | Revenue = volume x price | True by definition, so the parts must multiply to the whole | Almost none |
| A process or sequence | Cash: operations, then investment, then financing | Each pound arises at one stage only | A stage you forgot, such as interest or tax |
| A segmentation | Customers by size: small, mid-market, large, other | Defined by one rule, with an explicit remainder | Overlapping definitions, and an oversized "other" |
Why an identity is the safest cut
An identity cannot overlap and cannot leave a gap, because the equation itself is the proof. Revenue is volume multiplied by price. Profit is revenue minus cost. Enterprise value is equity value plus net debt. If you split along the terms of such an equation, any error shows up as a number that fails to reconcile.
A process or a segmentation depends on you having thought of everything. A process can omit a stage, and a segmentation can use categories that blur at the edges. Both are legitimate, and both should be tested against a total. When you have the choice, start with the identity and use the other two to go deeper inside one of its terms.
A revenue decline, decomposed to the pound
A subscription box business sells two products. In year 1 it sold 60,000 Standard boxes at £50 and 20,000 Premium boxes at £120, for revenue of £5,400,000. In year 2 it sold 52,000 Standard at £52 and 17,000 Premium at £118, for revenue of £4,710,000. Revenue fell by £690,000. The question is why.
Start with the identity, revenue = volume x price, and add the one piece it hides: the mix between a cheap product and an expensive one. Total volume fell from 80,000 to 69,000, a drop of 11,000 boxes. The year 1 average price was £5,400,000 divided by 80,000, which is £67.50.
The volume effect is the change in boxes at the old average price: -11,000 x £67.50 = -£742,500. The mix effect asks what year 2 volume would have earned at year 1 prices, compared with year 2 volume at the old average. At year 1 prices, year 2 sales are 52,000 x £50 + 17,000 x £120 = £2,600,000 + £2,040,000 = £4,640,000. At the old average they would have been 69,000 x £67.50 = £4,657,500. The mix effect is the difference: -£17,500, because the business sold slightly fewer Premium boxes as a share of the total. The price effect is year 2 volume times each price change: 52,000 x £2 + 17,000 x (-£2) = £104,000 - £34,000 = +£70,000.
The three effects sum to the total change: -742,500 - 17,500 + 70,000 = -690,000. As a check, year 2 revenue at year 2 prices is £4,710,000, which is £70,000 above the £4,640,000 at year 1 prices, matching the price effect exactly.
Now go one level deeper inside volume, using a second identity: boxes = customers x boxes per customer. Year 1 had 16,000 customers taking 5.0 boxes each, which is 80,000. Year 2 had 15,000 customers taking 4.6 boxes each, which is 69,000. Losing 1,000 customers at 5.0 boxes removed 5,000 boxes. The remaining 15,000 customers ordering 0.4 fewer boxes removed 6,000. Together that is 11,000, as before. At £67.50 a box, customers account for -£337,500 and frequency for -£405,000.
| Step | Calculation | £ |
|---|---|---|
| Year 1 revenue | 60,000 x £50 + 20,000 x £120 | 5,400,000 |
| Fewer customers | -1,000 x 5.0 boxes x £67.50 | -337,500 |
| Lower frequency | 15,000 x -0.4 boxes x £67.50 | -405,000 |
| Mix | £4,640,000 - £4,657,500 | -17,500 |
| Price | 52,000 x £2 + 17,000 x -£2 | +70,000 |
| Year 2 revenue | 52,000 x £52 + 17,000 x £118 | 4,710,000 |
What the bridge says
Price was not the problem; it added £70,000. Mix is a rounding error. The decline is a volume story, and within volume it is weighted towards frequency: 55% of the lost volume came from existing customers ordering less (405,000 of 742,500), not from customers leaving. That redirects the investigation. The question becomes why subscribers skipped boxes, which is a product and retention question rather than an acquisition one.
None of that was visible in the headline figure of a 12.8% fall. It appeared only because each cut was exhaustive and the parts were forced to reconcile.
An EBITDA driver tree
Apply the same method one line further down the income statement. Take the same business with a gross margin of 60% in both years, and operating costs of £2,150,000 in year 1 and £2,200,000 in year 2. EBITDA is gross profit less operating costs, so year 1 is £3,240,000 - £2,150,000 = £1,090,000, and year 2 is £2,826,000 - £2,200,000 = £626,000. EBITDA fell by £464,000.
The first cut is the identity: change in EBITDA = change in gross profit - change in operating costs. Gross profit fell by £414,000, which is the revenue fall of £690,000 at a 60% margin. Costs rose by £50,000. So -414,000 - 50,000 = -464,000. Gross profit then splits further along the revenue bridge, and costs split by line: staff, marketing and other fixed costs.
The insight is proportion. EBITDA margin fell from 20.2% to 13.3%, but almost 90% of the cash impact (414,000 of 464,000) traces to the lost volume, and costs that did not flex made up the rest. A driver tree is simply the MECE split taken to the point where each leaf is something a person could act on.
An issue tree for an investment question
Questions about whether to invest are harder to split because the top is a decision rather than a number. The fix is to find the identity underneath. Equity value gained equals the growth in EBITDA at the entry multiple, plus any change in the multiple, plus the debt repaid. A buyer paying 8.0x for £25m of EBITDA, with £100m of net debt, who sells five years later at 8.0x with EBITDA of £32m and net debt of £70m, makes: 7 x 8.0 = £56m from growth, nothing from the multiple, and £30m from deleveraging. That is £86m, and it ties to the change in equity from £100m to £186m.
The identity gives the tree its main branches: can earnings grow, will the multiple hold, and will cash be generated to repay debt. A fourth branch, what is the price paid, sets the starting equity against which all three are measured. Each branch can then be tested with its own evidence, and a weak answer on one of them tells you where the recommendation is fragile.
Market sizing as a MECE exercise
Sizing a market is a chain of identities, and the top-down and bottom-up routes are two independent partitions of the same quantity. Take the annual UK spend on professional dog walking.
Top-down: 28 million households, of which 30% own a dog, gives 8.4 million. Assume 10% of owners pay for walking, which is 840,000 households. If each buys three walks a week for 48 weeks, that is 144 walks a year, or 120.96 million walks. At £14 a walk, the market is £1.69bn.
Bottom-up: assume 30,000 people walk dogs for a living. Each handles ten dogs a day across five days for 48 weeks, which is 2,400 dog-walks a year, and 72 million in total. At £14, the market is £1.01bn.
The two answers differ by a factor of 1.7. Neither is right, and the gap is useful. It says that either the share of owners who pay is nearer 6% than 10%, or the number of walkers is nearer 50,000 than 30,000. Within a factor of two, with the gap traced to a named assumption, is a good result. The segmentation is MECE because every household either owns a dog or does not, and every owner either pays or does not.
Framing an investment thesis in three or four reasons
A thesis is a MECE list of reasons the investment should work. The reasons must be independent enough that you can lose one without the others collapsing. For a buyout, a defensible set is entry price, earnings growth and cash generation, which map directly onto the value identity above.
A set such as "great management, strong market, attractive valuation, differentiated product" fails the test. Management and product both explain growth, so they overlap, while nothing addresses how much debt can be repaid. Three or four reasons is the working range because each needs its own evidence, and a fifth rarely earns its place. The guide on moving from a financial model to a recommendation shows how to write them as headlines with the numbers beneath.
Common mistakes
- Mixing levels. A list that holds "pricing", "the north of England" and "a competitor entering" puts a lever, a place and a cause side by side. Each answers a different question, so they cannot be mutually exclusive.
- An oversized "other". If the residual box is larger than 10% to 15% of the total, the real categories have not been found. A third in "other" is a gap with a label on it.
- Defining categories by two rules at once. Splitting customers into "London", "online" and "large corporates" uses location, channel and size together, and a large London corporate buying online belongs in all three.
- Forcing three buckets. Three is a habit, not a law. If the identity gives two terms or five, use two or five.
- Never reconciling. A split that has not been summed back to the total is an assertion. Check it before you build on it.
Five exercises, with the answers
Work each before reading the answer that follows it.
Exercise 1. A sales director splits revenue into "UK", "Europe" and "online". What is wrong? Online is a channel and the others are places, so a UK customer ordering online fits two boxes. Cut by one rule only: channel first (stores, online), then geography inside each.
Exercise 2. A cafe group's revenue fell from £8.0m to £7.6m while costs fell from £7.2m to £7.1m. Split the profit change. Profit fell from £0.8m to £0.5m, a change of -£0.3m. Revenue contributed -£0.4m and cost savings +£0.1m: -0.4 + 0.1 = -0.3.
Exercise 3. A product sold 10,000 units at £20 and then 9,000 at £22. Revenue went from £200,000 to £198,000. Volume effect: -1,000 x £20 = -£20,000. Price effect on the new volume: 9,000 x £2 = +£18,000. Together -£2,000, which equals the change. Revenue barely moved, but it hid a 10% loss of volume.
Exercise 4. A company lists the reasons customers left as price, service, a competitor, moving house and product quality. Why is this not MECE? A competitor is where the customer went, not why, and price and quality can both be true of the same customer. Cut by who initiated the departure: caused by the company (price, service, quality), caused by the customer's circumstances (moving house), and other, which should be small.
Exercise 5. A company's cash fell from £40m to £28m. EBITDA was £30m, capex £18m, working capital absorbed £9m and tax was £7m. Does the split explain it? 30 - 18 - 9 - 7 = -£4m, but cash fell by £12m. The £8m gap is a missing branch, here interest paid. Process cuts fail by omission, and the reconciliation is what exposed it.
Next
Practise cutting a problem, with the answer checked
The structure-the-problem lab gives you a messy question, asks for your split, and tests it for overlaps and gaps before it shows you a cleaner one.
Get told when the programmes you care about open
Preparation pays when it lands before the window does. Leave your email and we will tell you the day the programmes you care about open, so this is done in the first fortnight rather than the last.
One email when applications open, plus a short weekly round-up in season. Free, and you can leave at any point.
Frequently asked questions
What does MECE mean?
Mutually exclusive and collectively exhaustive. Each item belongs in exactly one category, and the categories together cover everything. A split with both properties adds up to the total, with no overlaps double counting and no gaps hiding part of the answer.
How do you check that a split is MECE?
Reconcile it to a known total. If the parts sum to more than the whole, two categories overlap. If they sum to less, something is missing. An algebraic identity such as revenue = volume x price passes automatically, which is why it is the safest first cut.
How big can the "other" category be?
Keep it below roughly 10% to 15% of the total. A larger residual means the real categories have not been found, and the interesting part of the answer is probably sitting inside it.
What is a price, volume and mix bridge?
A decomposition of a revenue change into the effect of selling more or fewer units, the effect of the mix between cheaper and dearer products, and the effect of changed prices. The three effects must sum exactly to the total change in revenue, which is the test that the split is complete.
Do I always need three branches?
No. Use as many as the logic of the problem gives. Two terms in an identity means two branches, and five stages in a process means five. Forcing a split into three makes a category blur or a branch overlap.

