A developer estimated the story would take 3 days to complete. Turns it in 3 days later.
Returns from QA. Dev estimates 1 day, returns on time.
5x more Q A bounces, returned in estimated time.
Say:do for estimates = 100%
Except that lead time is 9 weeks.

If we think that estimation *causes* predictability, we can double down on that, and might learn to do it well.
But here we need to measure first-time-through in order to pursue more hygienic work practices.
And what of outcomes? Do we perfectly build the wrong things?

Some people need to do time estimation and plannin in a traditional way, and so they should (until they find a way to stop needing it).

In product orgs and project orgs alike, making time estimations shouldn't distract improvement efforts from more valuable work.