Revisions come in layers: monthly revisions to the prior one or two months, annual benchmark revisions against administrative records, and occasional comprehensive revisions that rewrite decades. The establishment-survey benchmark and the annual gdp-vintages revisions are the ones that matter most.
The trading implication is that the first print moves the market and the truth arrives quietly later. Real-time policy is made on numbers that will be different in a year, which is a large part of why output-gap estimates are so unreliable in the moment.
Example: payrolls are reported at plus 187,000. The following month it is revised to plus 122,000 and the month after to plus 105,000. The originally reported strength was 78,000 jobs that did not exist.
Related: gdp-vintages, establishment-survey, seasonal-adjustment, output-gap, nowcasting