Technical standards

Why does a plug made by one company fit a socket made by another that never consulted it?

Two firms that have never spoken produce a plug and a socket that fit. Nothing physical forced the dimensions; there is no uniquely correct pin spacing waiting in nature to be discovered. So the agreement is entirely conventional — and yet it is more reliable than most contracts. What sort of problem is being solved here, and what actually holds the solution in place?

Look first at the shape of the problem, because it is unusual. Almost nobody has a stake in which convention wins; everybody has an enormous stake in picking the same one. Driving on the left works exactly as well as driving on the right, and what does not work is half and half. Problems like this have many acceptable answers and no way to choose between them on merit, so the entire cost falls on failing to converge rather than on converging badly.

So how does convergence happen without a central authority? Two routes, worth keeping apart. In the first, a body writes the answer down — a standards organisation, an industry consortium, sometimes a regulator with the force of law. Everyone can read the document, conformance is testable, and the standard exists before the products do. Call that the de jure route.

In the second, no one writes anything. One implementation accumulates users until copying it is cheaper than not, and the standard is simply what most things already do. Nobody ratified VHS. The specification, if it exists at all, is written afterwards by people documenting what the market decided.

Now ask why the second route tips rather than settling into a comfortable share for each option. Because the value of a choice here depends on how many others made it — a network effect. Below some threshold, two conventions can coexist; past it, every new adopter makes the leading option more attractive to the next, and that is a self-reinforcing loop, so the market swings hard rather than drifting. The Betamax and VHS contest is the standard illustration: Betamax was widely held to have the better picture early on, VHS had longer recording time and more manufacturers licensed to build it, and once rental shops began stocking one format preferentially, the tapes on the shelf decided the matter for everyone still choosing a machine.

Then a second force locks the result. The installed base is not just devices; it is tooling, spare parts, training, documentation, and every complementary product built to fit. Switching means paying for all of that again, at once, in exchange for a benefit that only arrives if everyone else switches too. That is why an inferior convention can be immovable: the cost of change is concentrated and immediate, the gain diffuse and conditional. Britain settled its railways on the narrower gauge by the Gauge Act of 1846, and the Great Western's broader gauge — which had real engineering arguments in its favour — was converted away in 1892 because compatibility with everything else mattered more than the merits of the track.

Which forces an uncomfortable conclusion. Technical merit is one input among several, and often not the decisive one: timing, licensing terms, the availability of complements, and who already had customers routinely outweigh it. The folk example here is the QWERTY keyboard, and it deserves a caveat — the popular story that it was designed to slow typists down and that a better layout was demonstrably faster rests on evidence that has been seriously disputed, so it is a weaker illustration than its fame suggests. The format wars and the rail gauges make the point on firmer ground.

One more observation ties the two routes together: they are stages more than rivals. A de facto winner is frequently taken up by a standards body afterwards and formalised, converting a company's private decisions into a public document anyone may implement. The traffic runs the other way too — a written standard can be extended by a dominant implementer until conformance to the document no longer guarantees interoperability with what everyone actually uses.

A standard is best read as a coordination device rather than a claim about quality. Its job is to let strangers make interchangeable parts, so its authority comes from expectation — from everyone believing everyone else will follow it — not from any body's endorsement. This is why an unenforced standard that everyone follows is far more powerful than a mandated one that nobody implements.

Two honest qualifications. First, "network effects plus switching costs" explains why a standard persists but does not by itself tell you which candidate wins; the early history is usually contingent, and accounts of any particular format war differ on how much weight to give price, licensing, marketing and luck. Second, lock-in is not permanent, only expensive. Standards do get displaced — but typically when something changes the surrounding technology enough that the old installed base loses its value anyway, rather than because a better alternative simply argued its case.

• How patent licensing terms inside a standards body shape which technologies get written into a specification.
• Why some transitions succeed by running both standards at once for years, and what that costs.

Every term the collection defines is gathered in the glossary.