Some reasons to measure (2021)

There is an interesting thing going on here that the article doesn't quite make explicit - the interaction of the principle-agent problem with the bias tech companies have towards action.

A tech company has to have a bias towards just building something. Nobody knows what is profitable. The industry is littered with low-hit wonders who seem to have stumbled on a good idea then under rare circumstances muscled their way into a few other fields. Google is a poster child for this, one product that may as well be the hand of Midas and a graveyard of ideas (some of the concepts were winners though, just other companies took the profits). So we suspect that anyone who succeeds in tech probably has a bias towards trying a large number of things until they get lucky and strike good. The pattern seems common to me.

Then we have the basic principle-agent problem where employees are all rather incentivised not to rock the boat and draw attention to parts of the tech stack that are poorly designed but not literally a legal liability. Companies don't have the capability to figure out the difference between a badly designed system and a hard problem anyway.

Put these together, and it is no wonder that people don't bother measuring things. It is quite rare to see someone hugely successful who got there through measurement. Managers do themselves no favours proving that systems under their command are unreliable. Cultures likely have a heavy bias toward action.

The problem here is really the primitive management theory in the tech industry. One day we'll have a Deming moment where someone convinces management to focus on quality and has some huge success stories - hopefully inspiring copycat behaviour. I expect we'll see that once hardware stops improving at an exponential rate and the benefits of being first to figure out what to do with it dissipate.

Discussed at the time:

Some Reasons to Measure - https://news.ycombinator.com/item?id=28326275 - Aug 2021 (17 comments)