MECE is a list. It is a list of qualities lists should have. According to MECE, any list should be
- Mutually exclusive – the members of the list should “exclude” each other, i.e. be distinct
- Collectively exhaustive – the members of the list should “exhaust” the relevant field, i.e., contain everything that belongs on the list.
In plain English, MECE says that a list should have
- No overlaps
- No gaps
MECE is widely used by management consultants. In fact it seems to have acquired the status of holy writ. According to Ethan Rasiel in The McKinsey Way:
MECE…is a sine qua non of the problem-solving process at McKinsey. MECE gets pounded into every new associate’s head from the moment of entering the Firm. Every document (including internal memos), every presentation, every email and voice mail produced by a McKinsey-ite is supposed to be MECE. Ask any number of McKinsey alumni what they remember most about the way the Firm solves problems and they will tell you, “MECE, MECE, MECE.
It is therefore interesting to ask whether MECE is adequate, even by its own lights. So we should ask:
- Does MECE have any overlaps?
- Does MECE have any gaps?
It seems clear that gaps and overlaps are very different, so MECE looks to be ME. But is it CE? We need to ask whether there are any properties that a properly formed list should have over and above ME and CE.
And obviously there are. For example, what’s wrong with the following list?
- Bashful
- Doc
- Dopey
- Grumpy
- Hannibal Lecter
- Happy
- Sleepy
- Sneezy
The list is ME – all items are genuinely distinct. And it is CE – it “exhausts” the relevant field, which is the seven dwarfs. But it is clearly a stupid list. It violates the commonsense principle that lists should not include things that don’t belong on the list.
And it is not too hard to think of other desiderata which are missing from MECE.
So MECE is not merely not MECE. MECE is obviously not MECE. MECE does not live up to its own standards.
Tim van Gelder 
MECE, for all it’s usefulness, is not and never was meant as the be all and end all when judging work product. I only had two years of Latin in middle school, but I still know that something can be “sine qua non” yet also “necessary but not sufficient.”
No consulting rubric, no framework, no template is worth a wooden nickel if the consultants wielding them don’t apply their own intelligence and common sense.
Ethan, Thanks for that. You’re quite right that every guideline for thinking has to be applied with intelligence and commonsense. Still, I think you are agreeing that MECE is not MECE, which it should be, at least within McKinsey, if everything should be MECE. Also, the very next sentence in your book says “MECE structures your thinking with maximum clarity (hence minimum confusion) and maximum completeness.” If MECE is merely a “sine qua non” then it alone can at best *help* thinking have maximum clarity and completeness. Other principles are needed as well – some of which should have been included in MECE if MECE was MECE.
Tim, you’re the philosopher, not I, but it seems to me that you have made a category error. MECE is a rule for work product (awful term that, though I’ve now used it twice), not a work product (arrgh, three times!) itself. One could, I suppose, come up with some sort of Grand Universal Theorem for documents, but the acronym would be ever so unwieldy.
As to my statement on clarity and completeness, I think it stands: lack of overlap means no confusion among elements; no gaps means nothing essential left out. If you then want to add something about, say, “Nothing Extraneous”, thus weeding out Dr. Lecter, and, perhaps, “High Value Only”, and, for good measure, “Fully Implementable”, then you end up with MECENEHVOFI. I don’t think this, or any other acronymic construct I could come up with after a few hours of deeper thought, would be all that useful to practitioners.
Ethan, as is often the case, there’s a purely semantic or definitional aspect to this. I took MECE to be “a list of qualities lists should have,” whereas you I think are taking it to be something like “a handy list of a couple of the most important qualities lists should have, handy in part because it has a useful acronym.” Fair enough. MECE-hood certainly is relative to the nature or definition of the list, or what we would call the “grouping principle.” The really substantive point is not the MECE-hood of MECE, but what are the criteria for well-formedness of lists or groups? ME and CE belong, as does NE (Nothing Extraneous), and you’ve added some more. Another important one I think is that a list should be Ordered, i.e. the elements listed in some natural or sensible order. What others?
The acronym’s almost there…
O rderred
N othing
E xtraneous
L inked
I nclusive
S separated
T hemed
well… almost :)
Craig, not bad at all. A lot of those things (ordered, linked, themed) would be subsumed under Minto’s Pyramid Principle, which is part of McKinsey’s “basic training”. I don’t have time to check (heading to the airport in a few minutes), but I’m pretty sure MECE comes from Minto. It’s certainly in there, even if she doesn’t use the term explicitly.
There’s no doubt that MECE is a shorthand. It’s most useful as a very quick check of document structure. It’s usually easy to spot if a document is not MECE, which means it gets sent back for reworking and — if necessary — rethinking.
Barbara Minto refers to MECE in ‘The Pyramid Principal’ (1987). In chapter 7, in the section on Structural Order, she writes:
“When you divide a whole into its parts – whether it be a physical whole or a conceptual one – you must make sure that the pieces you produce are:
– Mutually exclusive of each other
– Collectively exhaustive in terms of the whole.
I abbreviate this mouthful to MECE, but it is a concept you no doubt apply automatically every time you create an organization chart”
She continues with further examples and explanation.
Minto may have put a nice acronym on it, but the MECE principles are a lot older than that. They come up a lot in my area of ontology and are clearly formulated at least as far back as the 13th Century by John Duns Scotus.
Have a look at paragraph 1.4 of “A Treatise on God as first principle”, here: http://www.ewtn.com/library/theology/godasfir.htm
Speaking of which, the ridicule of lists (categories) is also a long running practice. Foucault of course made famous Borges’ satire of “a certain Chinese encyclopedia” which states “animals are divided into: (a) belonging to the Emperor, (b) embalmed, (c) tame, (d) suckling pigs, (e) sirens, (f) fabulous…”
cheers,
RdR
Thanks Richard for digging up the Scotus reference. I was aware that these principles went a long way back but didn’t have a specific text to point to.
Interestingly, Scotus goes beyond MECE in listing not two but three qualities lists or “divisions” should have:
“1.4 For a division to be clear it is necessary (1) that the members resulting from the division be indicated and thus be shown to be contained in what is divided, (2) that the mutually exclusive character of the parts be manifest, and (3) that the classification exhaust the subject matter to be divided.”
Obviously (2) is ME and (3) is CE – indeed he uses that same language! [as translated, of course]
Principle (1) is a little obscure but I think it might be, in fact, the “Nothing extraneous” point which I raised in the original post.
In other words, to point to another fundamental principle: There’s nothing new under the sun!
This is nonsense. A partition of the sample space is one thing. The attributes of the partition is another.
MECE does not partition a sample space, so it cannot succeed ot fail as a partition of a sample space.
Michael,
Could you clarify a bit more how you see the difference between “partitioning” and “the attributes of a partition”?
I take it from the “one thing… another” phrasing that you see some fundamental distinction in the context of a collection of entities, but I’m not entirely sure what it is.
When I was taught the Pyramid Principle, we learnt that the items in the list needed to be not only MECE, but also of the same type.
It is this that sorts out the cannibal from the dwarves, rather than MECE itself. It is the essence of Scotus’ first test.
Creating a MECE list becomes harder when dealing with more nebulous ideas, as becomes necessary when doing Issue Analysis on topics such as whether a company should invest in sales in a certain territory. Interminable discussions on flip charts and whiteboards can ensue.
MECE is only a small part of the tool set. It just checks whether the conceptual space has been partitioned in a way that leaves no gaps and has no overlaps. It doesn’t tell you anything about what should be in that space, or where the dividing lines should go. That is left to the intelligence of the users, applying other methods.
Hello,
Thank you for the interesting discussion.
If I may offer a further point of view (from a former quant lost in business), MECE can be viewed also as the partition concept of probability theory.
The McKinsey definition may not be completely sound but it works well enough, which pragmatically is all I care in a business context.
On the other hand, I do share and appreciate the point of view of a sound skepticism on this kind of tools.
Going back to the far more interesting and elegant domain of probability theory, there, partition (MECE) is the cornerstone of the whole cathedral.
Probability is how we think (even more in the bayesian and de Finetti sense, or Savage sense for an Anglo-saxon reader). Probability is driving the latest successes in narrow AI (from translation to IBM’s Watson system)
It is a wonderful building, so partitions must have some special and deep connections to what we are as thinking beings and (future) creators of thinking machines.
Great blog!
DD
Too funny.
I just read some of the comments above (I should have read them ALL before posting).
Just two additional points:
1 – I’m not the voice-mail leaving guy
2 – I agree with Michael. Partition and MECE share a form but they’re “orthogonal” in relation to the attributes.
…yet again, partitions can be “colored” and this starts a link with logic, graphs and hierarchies which is again running deeply in both the way we think and computer science.
DD
Dimittrij,
I think we all empathise with your pragmatism — at least as far as having experienced the need for it. However, you probably don’t mean stopping to look for improvements to theory or its application.
“Partitions” are a cornerstone to conceptual thinking beyond probability. In ontology, partitions have a long history under the heading “mereology” (the theory of parts and wholes) and stretch even further back to Aristotle, especially in his piece “Categories”, but also in the “Physics” and “Metaphysics”. One of the important aspects is that almost any predicate creates a category or partition. So, partitions or categories ends up being a study of the nature of things — often in terms of bundles of predicates. “Carving up the universe at its joints” as ontologists put it.
Note that even pragmatically, MECE is incomplete. Tim pointed out “nothing extraneous”, but Scotus also has an antecedent — namely that something has to be “indicated”. What Scotus points to is that you can have a false MECE: it may look exhaustive and nicely partitioned, but the subjects are not indicated by the labels… the predicates don’t apply. Think “phlogiston”, “phrenology”, “therapeutic touch” and the like. Or, in applied probability (i.e. risk management), dare we talk about “expert score cards”?
Tim it seems to me you went too easy on the ME portion of MECE … without getting into the discussion of whether not being ME and CE somehow compromise MECE’s value, can’t one argue that the slightest possibility of something being both mutually exclusive AND collectively exhaustive preclude ME and CE being mutually exclusive? In fact you could almost argue that they are *by definition* not mutually exclusive, otherwise the concept as a whole would be useless … or am I oversimplifying this??
btw I have an interview for an Associate position with McKinsey coming up next week, let’s see how far this kind of thinking gets me ha!
Hi Amir, A provocative comment, but I think I’m having difficulty grasping your point. Certainly if MECE is to be MECE, then ME and CE ought to be mutually exclusive of each other. It seems to me that they are, but just what it is to be ME is, generally, not very clear. In some simple cases it is clear enough (are Switzerland and Germany ME, geographically? presumably they are) but often in my travels I’ve looked at items in some kind of more abstract or conceptual list, and wondered what it would mean for those items to be ME. It is probably not a question that has any kind definitive answer. Good luck with the interview. – Tim
Hmmm I wasn’t very articulate … that doesn’t bode well for the interview.
What I was trying to say is, while you concede that ME and CE are mutually exclusive but argue that they are not collectively exhaustive (for MECE’s purpose), I’m wondering how they’re even mutually exclusive. To say that they are suggests that there is no overlap between ME and CE, yet from one perspective the reason for MECE’s being is to guide people to achieve maximum overlap in this regard, i.e. make lists/subcategories that are both ME and CE.
Essentially what you argued is that if one were to make a list of what qualities good lists should have, it would need to have more qualities listed than just ME and CE in order to be considered complete i.e. exhaustive (CE). I’m contending that, as items on that list, ME and CE aren’t even mutually exclusive … if they were, it would negate the very purpose of the MECE mantra.
I do however agree with you that there is some degree of subjectivity involved in defining what it means to truly be mutually exclusive, and we are quite possibly just defining it differently.
It’s just a guideline, and a very solid one. All guidelines require common sense and intelligence to determine the quality of the output. The McKinsey approach to brief engagements (developing 3 solutions that are MECE) is simple and effective, rather than complete and extensive.
Quite frankly I don’t understand this… Mutually exclusive it means that you cannot exclude any of the items, without compromising the information (CE). Hannibal Lecter doesn’t fit the bill. Sorry. Your list is not MECE
Hi Adrian, I think you’re working with a different interpretation of “mutually exclusive.” My understanding of ME is that it means that each item in the list “mutually excludes” each other item, i.e. it is a pairwise (“mutual”) relationship. Your interpretation has to do with how all items relate to each other and other potential items, and this is more properly called a “collective” property. My interpretation (or rather, the interpretation I went with, since I think is the standard interpretation “out there”) appears to have been accepted by the other readers and respondents to this list. However I’d be interested if you can provide evidence that your interpretation is correct. Can you find some existing quote (e.g. in Minto’s work, or in the management consulting literature) clearly supporting your interpretation?
Adrian, that’s not the interpretation that Minto gives of MECE. She gives an example of departments in a company: “Mutually exclusive means that what goes on in the Tire Division is not duplicated in Housewares… In other words, no overlaps.” *The Minto Pyramid Principle*, page 82
MECE needs to face the MuSIC. Let me explain…
It’s actually not that hard to come up with MECE lists. Almost any categories will do, as long as you follow some simple rules. Let’s say that you have some coloured balls and you make some categories for them: Red, Blue and Yellow. Now, if you have balls of primary colours, then your categories are MECE. But let’s say a purple ball comes along. Now the categories no longer seem Mutually Exclusive.
To solve this, firstly identify some primary categories for the entities of interest. A primary category is a predicate: a property (a feature) that your objects either have or don’t have. Now ensure that the primary categories are “multi-selectable”. That is, for any object you will select all the primary categories that are applicable. We can now build general categories from the evaluation of the bundles of primary categories. For example, treat Red, Blue and Yellow as primary categories. A general category is then the evaluation of the bundle of the three primary colour categories. One such general category is (Red=yes, Blue=no, Yellow=no). Another general category, known as “purple”, is (Red=yes, Blue=yes,Yellow=no). The three primary colours lead to 8 categories in total.
These categories are mutually exclusive, due to the binary nature of predicates. Of course, proper interpretation of the primary categories is necessary. For example, if a ball is Red today and Blue tomorrow, then you have to decide whether your categories mean colour on all days, or a specific day… or perhaps decide to add some more predicates to handle time.
These categories are also collectively exhaustive, as long as you include all possible evaluations of the predicates. In other words, all yes/no combinations across all primary categories.
However, we need to add one more item: what Scotus referred to as “indicating”. That is, our categories need to be relevant to the entities that we’re categorising and there need to be such entitites in the first place. So, our example categorisation is appropriate for coloured balls, but not for, say, types of numbers (nor for certain types of logical puzzles, such as Russell’s paradox… but that’s another story).
So,the recipe for better MECE lists is to use Multi-Selectable, Indicating Categories (MuSIC).
MECE needs to face the MuSIC. Let me explain…
It’s actually not that hard to come up with MECE lists. Almost any categories will do, as long as you follow some simple rules. Let’s say that you have some coloured balls and you make some categories for them: Red, Blue and Yellow. Now, if you have balls of primary colours, then your categories are MECE. But let’s say a purple ball comes along. Now the categories no longer seem Mutually Exclusive.
To solve this, firstly identify some primary categories for the entities of interest. A primary category is a predicate: a property (a feature) that your objects either have or don’t have. Now ensure that the primary categories are “multi-selectable”. That is, for any object you will select all the primary categories that are applicable. We can now build general categories from the evaluation of the bundles of primary categories. For example, treat Red, Blue and Yellow as primary categories. A general category is then the evaluation of the bundle of the three primary colour categories. One such general category is (Red=yes, Blue=no, Yellow=no). Another general category, known as “purple”, is (Red=yes, Blue=yes,Yellow=no). The three primary colours lead to 8 categories in total.
These categories are mutually exclusive, due to the binary nature of predicates. Of course, proper interpretation of the primary categories is necessary. For example, if a ball is Red today and Blue tomorrow, then you have to decide whether your categories mean colour on all days, or a specific day… or perhaps decide to add some more predicates to handle time.
These categories are also collectively exhaustive, as long as you include all possible evaluations of the predicates. In other words, all yes/no combinations across all primary categories.
However, we need to add one more item: what Scotus referred to as “indicating”. That is, our categories need to be relevant to the entities that we’re categorising and there need to be such entitites in the first place. So, our example categorisation is appropriate for coloured balls, but not for, say, types of numbers (nor for certain types of logical puzzles, such as Russell’s paradox… but that’s another story).
So,the recipe for better MECE lists is to use Multi-Selectable, Indicating Categories (MuSIC).
Might I suggest MECEFR or MECEFU with FR standing for Fully Relevant OR FU standing for Fully Useful
Yeah technically MECE just means the above two things (No overlap and no gaps). But the implications should go beyond that. We explain that in detail here! (VIDEO included)
http://mconsultingprep.com/case-interview-mece/
Hi Tim. Thought provoking, however you have made a category error. MECE is not an attribute of a *list*…. MECE is an attribute of a *grouping* applied to a list. MECE has *nothing* to say about the contents of a list. The contents of a list are defined elsewhere, and then MECE is used to group them.
For example, when attempting to group the items in the list you provide, a MECE grouping scheme might be “Fairytale Fictional Characters”, and “Contemporary Fictional Characters”. A non-MECE grouping scheme might be “Characters whose name begins with S” and “Characters whose name begins with H”, because that doesn’t include everyone and violates CE. Another non-MECE grouping scheme would be “Characters whose name begin with H”, “Serial Killers”, and “Dwarfs”, because some items fit in multiple categories, and so this fails the ME test.
TL/DR: Actually, MECE is a grouping principle for separating a set of items into subsets, not for validating which items are allowed in the original set.
I’m neither a philosopher nor one who knows anything about formal logic, but I notice that the starting of Tim’s original posting and the underlying assumption in the subsequent discussions is that we are given a list of items to group.
As far as practice goes, I believe MECE was introduced not to group existing items, but rather to ensure the breakdown into appropriate parts of one initial question that is itself an implicit grouping of multiple notions and assumptions.
In that context, MECE is both applicable and MECE. In Tim’s original posting, MECE is not applicable
Thank you for this post. I’ve been struggling with this concept since a friend introduced it to me (albeit in a context that I think was too natural, subjective, or nuanced to be applicable). This may be why I was having such a hard time. I also think that the simplicity of MECE is the source of its flaws. For example, how could the division of fictional characters into fairy-tale and contemporary (as set out above) be ME or CE? Are there no contemporary fairy-tales or is that a semantic consensus that must be reached? Are there no other choices or are we allowed to simply assert the dichotomy? I think this flaw is rooted in language and the fact that, for many domains, labels are inherently too vague to be completely mutually exclusive. Thus MECE isn’t CE as you’ve already pointed out (there must be other qualities). If the tool can be improved, then why not improve it regardless of how complex it may seem to become (why do we need to stop at MECE)?
Additionally, as I think Amir touched on, if we’re allowed to simply assert that “common sense” should exclude unrelated items, why can’t we assert that “common sense” will ensure all related items will be included? Doesn’t this weaken the tool?
If we’re allowed to assume that the items are originally from a complete “whole”, then we can also assume that the mutually exclusive groupings of the items (assuming that every item must go into at least one grouping) will either automatically result in collectively exhaustive lists or the “whole” was actually missing items to begin with. Thus MECE isn’t even ME, as the set of ME sets becomes a subset of CE sets. If we are to consider the set of ME sets that aren’t CE, then we’re considering “wholes” that began with missing items (remember, extra or unrelated items don’t technically make the list any less exhaustive) and we are now thinking about what to include in our list (something others have said is not the job of MECE).
I think we have to care about where the items came from (require items belong in addition to ME and CE). If we don’t do so then we either (a) don’t need lists to be CE because we know that the lists begin with related items only (thus assuming hypothetical lists are the same as the concrete lists created by humans) or (b) can allow unrelated items as a possibility only checked by “common sense” which is disastrous. Simply claiming that the act of partitioning a whole automatically excludes unrelated items from the resulting set doesn’t make it any less of a necessary step of the tool. Without it, the other two indicators are drained of their power. At the end of the day, we do not really partition hypothetical “wholes” into sets. Instead we construct lists, based on expertise, which may be incomplete, compared to the hypothetical “whole”, for practical (we don’t know the remaining items) or semantic (we just arbitrarily define our way out of the problem) reasons.
By leaving the tool so simple, we reintroduce “grey areas” into a system designed to remove “grey areas”.
Thank you for this post. I’ve been struggling with this concept since a friend introduced it to me (albeit in a context that I think was too natural, subjective, or nuanced to be applicable). This may be why I was having such a hard time. I also think that the simplicity of MECE is the source of its flaws. For example, how could the division of fictional characters into fairy-tale and contemporary (as set out above) be ME or CE? Are there no contemporary fairy-tales or is that a semantic consensus that must be reached? Are there no other choices or are we allowed to simply assert the dichotomy? I think this flaw is rooted in language and the fact that, for many domains, labels are inherently too vague to be completely mutually exclusive. Thus MECE isn’t CE as you’ve already pointed out (there must be other qualities of lists). If the tool can be improved, then why not improve it regardless of how complex it may seem to become (why do we need to stop at MECE)?
Additionally, as I think Amir touched on, if we’re allowed to simply assert that “common sense” should exclude unrelated items, why can’t we assert that “common sense” will ensure all related items will be included? Doesn’t this weaken the tool?
If we’re allowed to assume that the items are originally from a complete “whole”, then we can also assume that the mutually exclusive groupings of the items (assuming that every item must go into at least one grouping) will either automatically result in collectively exhaustive lists or the “whole” was actually missing items to begin with. Thus MECE isn’t even ME, as the set of ME sets becomes a subset of CE sets. If we are to consider the set of ME sets that aren’t CE, then we’re considering “wholes” that began with missing items (remember, extra or unrelated items don’t technically make the list any less exhaustive) and we are now thinking about what to include in our list (something others have said is not the job of MECE).
I think we have to care about where the items came from (require items belong in addition to ME and CE). If we don’t do so then we either (a) don’t need lists to be CE because we know that the lists begin with related items only (thus assuming hypothetical lists are the same as the concrete lists created by humans) or (b) can allow unrelated items as a possibility only checked by “common sense” which is disastrous. Simply claiming that the act of partitioning a whole automatically excludes unrelated items from the resulting set doesn’t make it any less of a necessary step of the tool. Without it, the other two indicators are drained of their power. At the end of the day, we do not really partition hypothetical “wholes” into sets. Instead we construct lists, based on expertise, which may be incomplete, compared to the hypothetical “whole”, for practical (we don’t know the remaining items) or semantic (we just arbitrarily define our way out of the problem) reasons.
By leaving the tool so simple, we reintroduce “grey areas” into a system designed to remove “grey areas”.
As this is a very engaging and instructive post and thread, I thought it’d be a useful exercise to attempt to logically integrate the wide range of ideas presented by the various commentators, to date.
It seems that most of the conflict in this thread seems to arise from an initial logical oversight: general criteria for the “well-formedness of lists or groups” don’t logically exist at the same level as low level criteria, such as ME or CE (or ‘order’, or ‘high value’, etc).
Thus taking MECE to be a “a list of qualities lists should have” (Tim van Gelder) frames the issue at too general a level and leads to both the trivial and imprecise conclusion that MECE is incomplete (since other criteria obviously exist) and needless confusion with commentators debating at cross-purposes.
The real questions would seem to be to what extent MECE is MECE on its own terms and the precise nature of its relationship to the other viable list criteria mentioned (‘nothing extraneous’, ‘ordered’, etc).
The issue of MECE’s MECE-ness can arguably only be settled by
1) identifying a tenable super-ordinate group (at the appropriate ‘Goldilocks’ level of abstraction) to which the ME and CE rules belong i.e. the “subject matter to be divided” in Scotus’ phrase (Richard de Rozario) and
2) determining whether ME and CE are the only members of that group (i.e. whether ME and CE are themselves CE, since their ME-ness seems incontrovertible)
In probability theory, both ME and CE are fundamental criteria used to assess whether a space has been properly logically partitioned (Dimitrij Dugan). Hence, a plausible super-ordinate group is the category of ‘logical partition criteria’ i.e. properties of a well formed logical grouping.
We can then ask the more precise question as to whether ME and CE are the only such logical partition criteria. Since the ME and CE criteria collectively define a partition, it seems that MECE is itself CE.
A potential complication is Scotus’ ‘indicated’ criterion ( Richard de Rozario). However, since a MECE grouping need not be inductively valid in order to be logically valid, this would not appear to be a critical distinction on the level of ME and CE.
I found it useful to generalise this process to organise the collective wisdom of the commentators and map where the MECE rule fits in relation to the other list well-formedness criteria mentioned.
Below is a first pass at integrating and mapping the hierarchical/tree-like ‘deep structure’ of the debate (more abstract criteria were generally obtained by abstracting from the specific low level concrete critera mentioned in the thread):
[‘CRITERIA FOR THE WELL-FORMEDNESS OF LISTS’] // (Tim van Gelder) root
– LOGICAL COHERENCE CRITERIA
— LOGICAL PARTITION CRITERIA (Dimitrij Dugan)
— [ME, CE] // ME == ‘Separated’? (Craig Brown)
—- ‘Nothing Extraneous’ criteria (Ethan M. Rasiel)
— LOGICAL HIERARCHY CRITERIA
— nested MECE criteria (‘Linked’, ‘Themed’) // (Craig Brown)
— LOGICAL SEQUENCING CRITERIA
— ‘Ordered’ // (Tim van Gelder)
– VALUE CRITERIA
— ‘High Value Only” // Pareto principle? (Ethan M. Rasiel)
– UTILITY CRITERIA
— PRACTICALITY
— ‘Fully implementable’ // (Ethan M. Rasiel)
– REALISM CRITERIA
— ‘indicated’ // (Richard de Rozario)
With this framework, any list can be explicitly and precisely assessed against several high-level “well-formedness” criteria: logical coherence, value, utility, and realism, and their subtypes.
This framework is clearly a work in progress (the relationships between the partition and hierarchy criteria could probably do with some re-analysis, for instance) but already it helps us to think more clearly about the issues and direct our mental efforts more productively:
For example, to return to the original question motivating the post, it doesn’t appear to make sense to attempt to assess MECE’s MECE-ness without first specifying an appropriate reference group/level. So MECE does appear to be MECE within the scope of logical partition criteria but will clearly fail its own test when compared against “well-formedness” criteria from other branches.
Another example, the notion of singular or multi-selectable categories (Richard de Rozario) doesn’t seem to belong within this framework since it evidently refers to a property of the basis of division used to construct the groupings and is not a criteria for well-formedness:
[LIST CHARACTERISTICS]
– BASIS OF DIVISION
— ATTRIBUTE SINGULARITY
— singular
— multi-selectable // (Richard de Rozario)
Relatedly, as a final point, regarding the original ‘seven dwarfs plus Hannibal Lecter’ list, the addition of Lecter creates a list with a different – indeterminate (by me, at least!) – basis of division to that of the seven dwarfs list (e.g. ‘Snow White’s diminutive human friends’?). Without a clear basis of division it’s practically impossible to determine whether a list is CE so it would appear to be an open question as to whether this list is actually CE or not!
Interesting point, David — I tend to agree that in judging the quality of a partitioning approach like MECE, it would help to clarify the criteria for judgment. Given that it is easy to generate a categorisation that is MECE by using predicates instead (see my other reply), it shifts the focus to what those other criteria might be.
For example, one can imagine generating a categorisation for plants like this: “Grue”, “Reen”, “Brue”, “Other”, where every plant ends up in the “Other” category. This would probably not be a very useful categorisation for most practical purposes, because the categorisation doesn’t really sort the items of interest.
Similarly, one can imagine a categorisation of causes of some effect, like illness. For example, we might have categories like “air quality”, “nutrition”, “socio-economic status”. Now let’s say that there are an equal number of cases of illness for each causal category. That actually may not be helpful, because none of the categories is distinctive from the others in predicting the effect.
On the other hand, let’s say we have a categorisation for books in a library and each category ends up with roughly the same number of books. In this case, the categorisation may be useful, because the even distribution over the categories will speed up the categorical search for a particular book.
My point with these examples is that structual criteria like MECE, coherence, hierarchical features, etc. may not be the most important criteria for deciding on categorisations.
I concur. Going back to the less philosophical and instead very applied context for which MECE was concocted, the 2 tests I always teach students to apply when breaking down a business question about what do to into smaller questions are:
1. Is the breakdown MECE?
2. Is the breakdown practical and insightful?
Richard,
“My point with these examples is that structural criteria like MECE, coherence, hierarchical features, etc. may not be the most important criteria for deciding on categorisations.”
I’d completely agree with that assessment – hence the effort to create a framework that could potentially handle all possible criteria, including pragmatic considerations, such as value, utility, and realism (which would seem to subsume the criteria motivating your diagnostic and book search examples).
In the philosophically playful spirit of the original post, I used the MECE rule as a jumping-off point for a structured exploration of potential list criteria without any preconceptions as to their relative importance.