Thread by @patio11

A page to watch this week. Historically it updates on Mondays. niid.go.jp/niid/ja/flu-m/…

It would be very bad if Tokyo's updated numbers show early March had a spike of pneumonia deaths of unknown origin.

There is an official comment about this issue.

https://twitter.com/motokorich/status/1253263294376247296?s=20

n.b. Note that due to reporting delays and the challenges of managing data pipelines which are heavily manual, these sort of data “cure” upwards, both in general and with specific regards to this series.

“‘Cure’?”

Consider the challenges of counting all mortgages originated on April 1st via a census method. On April 2nd you’d have some idea, if you had total observability, but since you don’t, you’re going to wait while information goes from banks to local aggregators to you.

So you might publish mortgage numbers for April 1st on April 15th.

But quite plausibly the person for Bank of America was on vacation for a few days and accidentally missed the deadline. Whoops.

So when you report April 8th’s numbers on 22nd, you might revise 1’s upwards.

Plausibly if you really, really cared about the mortgage numbers you’d have everything computerized (not e.g. faxed) and attached to an ops team, such that if Bank of America were an hour late someone would call a named contact there and ask what the story was.

Finance, yay.

Now that would be a difficult decision to make if you were in prompt economic crisis, so you’d hope to have made the correct call the last time you had an economic crisis, such that your observability would be adequate the next time.

But there are lots of tradeoffs.