He wants people there to be his version of Minitrue, providing the numbers he wants to see, not the real ones:
Reporting unworkers doubleplusun-good, rewrite fullwise upsub antefiling.
So if you want early data it will inherently be of limited accuracy because that involves a lot of extrapolation with whatever incomplete data has been collected by that time. If you want accurate data you will have to wait for it because that data takes longer to be collected. You do want both because you need to make timely decisions, since most times the early numbers don't get revised by much, but you also want to course-correct when later, better data gives a different signal.
Agencies like the BLS publish their methodologies in great detail. Big revisions have always been happening, only they are getting more attention these days because of the heavy politicization.
The initial estimates haven't been accurate recently, because many workers have been completely dropping out of the economy because they're afraid of immigration raids. Information about these workers is just harder to gather and takes longer to verify.
Here, take a look for yourself: https://www.bls.gov/web/empsit/cesnaicsrev.htm#2024
If this is an understood part of the process, why is it such a problem now?
Name some organizations that have "fire employees until we get success". Does that create a culture that prizes success, or just encourage employees to hide failure?