from Patrick McKenzie | by Patrick McKenzie

There is something like a "general factor of infrastructure" and if you get it, you can see around corners of issues like this without knowing the exact shape of the issue in advance. Seems like it would be an interesting subfield for e.g. operations research. t.co/idkEPdqPUT

I wonder if it is institutionally easy to acknowledge that because I think you can more easily intellectualize e.g. queuing theory, in the sense of "yes, obviously worth a paper", versus intellectualizing "common pitfalls in multi-party software deployments."

Anyhow: if you've been around computers and finance a while, and you see a problem where it would be really helpful to have near real-time understanding of complicated data, you can immediately start speculating on which subsystems will find "real-time" to be... challenging.

Also if you've been around for a while you could speculate pretty confidently about which parts of banks define real-time in which ways. Example: real-time in credit card processing implies probably more than 2 orders of magnitude slower than real-time in equities trading.

And thus you know that at least one party somewhere in the system has said "Ah yes, as a result of a huge technical project in the late 2000s, we have that very important data newly available in real-time: 99.9% available by close of business the following day!"

"Wat." Well it is real-time by comparison to previous system, where it was 80% available by the monthly close, and a mid-level manager thought "real-time" was a good buzzword to get it approved by higher-ups, and now this SLA is that suborg's understanding of what real-time is.

Anyhow, downstream implications of this include "If you want to make data-driven decisions under conditions of extreme uncertainty and urgency you have to have good intuitions for what data that collects quickly is most likely to be upstream of the truth."

Some surprising implications include "I bet the team that counts mobile accesses has a better estimate as to outflows in the last hour than the team that counts actually money" and "Put the stock of ourselves and contras on the dashboard because that will often be early warning."

One upon a time I got a push notification from my brokerage saying that a particular stock was now subject to SEC Rule 201. I called an executive immediately; that was in context an obvious Big Red Button issue raised by a very non-obvious pathway.

To save you a scintillating dive into circuit breakers and not say too much: there are some stocks which one expects to be volatile and there are some stocks which one expects to not be volatile. Sudden volatility in one's upstream providers could conceivably be worth a war room.