Pareto triage: Difference between revisions

{{a|work|{{image|Pareto triage|jpg|}}}}{{dpn|pəˈreɪtˈəʊ ˈtraɪɪʤ|n|}}

To exercise “[[Pareto triage]]” is to move beyond the ''observation'' that, in a given period, eighty percent of your revenues may be attributed to twenty percent of your customers, and ''vice versa'' — which is just one of those unfortunate, immutable characteristics of any large group of uneven numbers— and to use it as a business action plan to impose order upon the intractably messy universe.

{{Quote|We have observed that one fifth of our customers provide ''four'' fifths of our revenue. We have spreadsheets to prove it. That means the remainder of our customers — eighty percent of the blighters! — generate just twenty percent. These people are hardly worth the bother. We would be much better served just foregoing that marginal revenue — or, at any rate, paying not the blindest bit of attention to keeping it — and instead concentrating on that lovely twenty percent segment who bring in all the rest of our income. }}

Now, unless your customers are materially identical in size, shape and interest for your goods and services — and while the gormless denizens of [[data modernity]] urges us on in that direction, we’re not there yet — it is ''statistically certain'' that one half of your customers will buy more of your goods and services in a given period than than the other.  “Given”, here, means “arbitrary”: however you define that period, and however you assemble your customer roster, it will remain true.

For this is no wondrous insight into the mystic citadel of ''homo sapiens'', but a basic property of a group of different numbers.

You can arrange any group of random numbers into subgroups such that one subgroup will have a greater average than another. The key fact is that ''you ''are arranging them that way. It is a wilful, planned, intellectual choice. Your thumb is on the scale. That, having so arranged your customers, this should hold for their revenues is unremarkable. It would be extraordinary if it did ''not'' hold, for any period and any subgroup of customers you chose. ''It is a property of the group, not of any individual in the group.'' The same individual may be in the majority of one group, but the minority of another. It depends which other individuals you choose.  

This is to say that an average of is an [[emergence| emergent property]] of all the members of the group, and not any individual members. Isolate that single member from the group and the “average” evaporates. The average net worth of the passengers on the Clapham omnibus can be increased by a factor of ten by the simple expedient of having one [[Techbro on the Clapham Omnibus|techbro]] board the bus, without changing the actual income of any of the other passengers.

This is the logical error of [[Pareto triage]]. It is to mistake a mathematical property of a set of data for a hard, determinate, property of members of that set.

It therefore follows that the “eighty” and the “twenty” segments of your customer base — also being uneven numbers — are no more homogeneous then the whole. The Pareto rule will apply equally to each of them.

That Pareto principle is, therefore ''[[fractal]]''. It scales down and up. If you cut off the “bad bit”, you will see, to your horror, your new, concentrated, high-value, but radically down-sized “good bit” ''still'' has a bad bit requiring [[Pareto triage]]: there are ''still'' twenty percent of its population generating eighty percent of the revenue. The revenue pot is just smaller, that’s all.

This is a [[Xeno’s paradox]] for our age: If we chase a Pareto triage to its logical conclusion we will end up with one customer.

And that single customer might not be much chop, either. Just because it fell into “the gilded 20%” in ''one'' period does not mean it would have been there in another. Some of the customers you exited may have come good.  

Nor does Pareto triage tell you anything about the ''forward'' value of your customers. Whatever period it covers, your data is a historical, averaged, a ''snapshot'': it will tell you nothing about the development of customer revenue before, after, or for that matter even during the sample period.

We are prone to [[averagarianism]] here, too. A revenue stream that increased exponentially over a decade may, if averaged over the decade, seem tepid. A customer whose revenue has been stable for years, but recently fell off a cliff, may still look heroic.

And even if you can see that historical trend, of itself ''it doesn’t tell you anything''. There are any number of reasons a customer’s revenue profile might change. They may be internal to the customer, a function of market, a reflection of your product, or a result of poor sales coverage.  It may be harder work than crunching a spreadsheet, but looking into why a specific customer is not buying your services any more it will give much better basis to make the decision to terminate it.

To be sure, loyal, revenue-generating customers do have enormous ongoing value, but that has nothing to do with Pareto triage. You can only find that out the hard way: by patiently building a positive commercial relationship with them.