The minority rule
This is a named pattern with documented cases and a sound logic behind it, not an effect anyone has measured. Treat it as a lens that keeps proving useful rather than as a law.
Taleb set the rule out in Skin in the Game in 2018, in a chapter called The Most Intolerant Wins, after publishing an early version as an essay in 2016. He calls it the mother of all asymmetries.
The structure is always the same. One group is intransigent, meaning it will not accept the alternative under any circumstances. The other group is flexible, meaning it can live with either option. Someone who can drive a manual car can also drive an automatic. The reverse is not true. Someone who keeps to a dietary rule will not eat outside it. Someone with no such rule will eat either. When the two groups mix, the flexible side absorbs the constraint, because doing so costs it very little and refusing costs the other side everything.
Then the cost of separation decides the outcome. If running two versions is expensive relative to running one, the producer converts everything to the stricter standard. Taleb's central example is that manufacturers make drinks compliant with dietary law because the compliance costs almost nothing and running two production lines costs a great deal. He makes the same point about cars. The spread of automatic transmission is not necessarily evidence that most drivers prefer it, since a person who can drive a manual can also drive an automatic while the reverse does not hold, so the flexible option wins without anyone having chosen it.
He borrows a piece of physics to explain how a small number becomes a large effect. Renormalisation is the idea that a rule applied at one scale propagates upward into the next. One person in a household who will not eat something converts the household. That household attending an event converts the caterer's order. Enough such orders convert the supplier. At each level the constraint is absorbed by people who do not mind, and by the top of the chain it looks like everybody chose it.
Which produces what Taleb calls an optical illusion. An observer looking at the average sees a market that appears to prefer the stricter standard, when almost nobody in it has a preference at all.
Two conditions decide whether it runs, and both come from Taleb himself. The first is spatial. The minority has to be spread evenly through the population. If it is concentrated in one area, that area gets its own supply and the rest of the market is untouched. The second is cost. Compliance has to be cheap. He is explicit that if the stricter standard were expensive to produce, the rule would not propagate, and the market would separate into two rather than converting.
The honest limit is what kind of claim this is. It is an argument assembled from cases rather than a measured finding, and the three to four percent threshold is an estimate offered as an illustration rather than a number anyone has tested. The renormalisation analogy is a way of picturing the propagation rather than a derivation of it. The direction of the mechanism is easy to verify in the world and the strength of it is not established.
The version that costs companies money is procurement and compliance. One large customer requires a security certification, a data residency guarantee, an audit standard. Rather than run one compliant operation and one ordinary one, the company certifies everything, because maintaining two tracks costs more than maintaining one. From that point on every customer pays for that certification in the price, including the ones who never asked for it and never would.
The same shape sets internal defaults. One department cannot run the new system, so the company standardises on the version everyone can run, and the capability gap that creates never appears as a line item because it shows up as things nobody does. One client needs reports in a particular format, so every report is built that way. One customer segment will leave if a feature is removed, and everyone else is indifferent, so the feature stays and every release from then on carries the cost of maintaining it. This is a large part of why software accumulates and rarely simplifies.
None of these involve anyone behaving unreasonably. The person who cannot use the new system is not being difficult, and the client with the format requirement is not being unreasonable. The asymmetry does the work on its own, because refusing is free for one side and insisting is expensive for the other.
What makes it worth auditing is that these are permanent decisions produced by temporary conditions. The certification outlasts the client who demanded it. The format outlasts the account. The old system outlasts the person who could not use the new one. Nobody revisits them, because nobody remembers them being chosen in the first place.
So the useful question is not which of your defaults were set this way. It is whether the person or the customer they were set for is still there. On defaults that were never chosen at all, see the default effect. On the asymmetry Taleb builds the rest of his argument from, see skin in the game.
Source: Taleb, Skin in the Game, 2018, chapter two. The earlier essay version is The Most Intolerant Wins: The Dictatorship of the Small Minority, 2016. The three to four percent threshold and the United Kingdom lamb figure are Taleb's own, offered as illustration rather than as measurement.
Nassim Nicholas Taleb, 2018
The chapter on the minority rule sits early in the book and works on its own. The rest is his argument that decisions without consequences attached are not decisions, which is where the intransigence comes from in the first place.
Draw your own card. It does not take long, and it rewards taking your time.