How often do superior alternatives fail to catch on?

But people don't do a cost benefit analysis or anything rational like that when they select a keyboard or when they select most things. Its a subconscious decision mainly factoring in what other people do (what is most popular) and the rationalizations come after.

There is a drug used to treat a form of eye disease that causes blindness. The product, eylea, has dominated the market despite (until recently) having no evidence that it improves vision vs the competition, provides no significant improvement on other relevant clinical endpoints, had no major safety benefit, and it is priced at a comparable or higher level than all competitors

A development stage product had shown better potential for vision improvement without major safety concerns, but when I interviewed physicians to ask whether they'd still prescribe eylea if this "better" drug was approved. They unanimously, without hesitation, said they'd prescribe eylea

Why would doctors treat eye disease with a product that does a worse job of treating eye disease? Why would patients accept a treatment that does not optimally restore their vision? There was nothing in the clinical literature to suggest this made sense. This must be a case of a large, established pharma company playing dirty, maybe bribing doctors or brainwashing patients with tv ads

The reality is that eylea required fewer doses than the competing products. For this drug, dosage means injection in the eye with a big needle. Further, physicians don't get paid more for doing an additional dose, so for them it's lost revenue. Turns out the incremental vision improvement does not offset the benefits of less frequent dosing

Getting this insight is not a feat of data analysis or scientific brilliance, but simply talking to customers. And getting this right, in this case, means winning a $4-5 billion market

A related thought based on my recent experience.

One of my teams is currently using a mesh colorization system for our game, which we developed in-house. E.g. we have a game level where each mesh is assigned some color ID, such as "primary accent" or "background1" (but not an actual RGB color); and there are color schemes which we can apply to the level according to these color IDs to get different looks.

The first version of the color scheme system was "clearly superior": it offered a lot more possibilities for assigning color schemes to meshes, but it needed about 80 color values for each color scheme.

The second version of the same system was "clearly inferior" -- just 28 color values, and much less creative control over the resulting look of the level.

Guess which system was adopted? Right, the second, simple one. Despite offering less creative possibilities, the second system was smaller and thus could fit more easily into artists's heads, which produced better-looking results overall.

To sum up, "superior" is multidimensional. Dvorak can offer better typing speed, but it offers much less "habit compatibility" with existing keyboards in the real world, and thus is a worse time investment. Or, Haskell offers much better maintainability, refactorability and reliability, but dynamic languages offer faster time-to-market and easier-to-hire coders.

The reason.com article inflates the significance of the QWERTY myth, and then claims that debunking it debunks path dependence, which, of course, it doesn't.

Actually if you research the story of the bicycle, it was a failure that took over one century until people started to realise how it could be useful, without killing themselves in the process.