The BLS runs two separate surveys every month. The establishment survey produces the headline NFP number by polling about 119,000 businesses. The household survey polls about 60,000 households and produces a separate employment count, along with the unemployment rate.

They measure different things. The establishment survey counts jobs. The household survey counts employed people. One person working three jobs shows up three times in the establishment count and once in the household count.

Since early 2023, the gap between the two has been unusually wide. The establishment survey showed consistent job growth throughout 2023. The household survey told a more subdued story. Employment by that measure was essentially flat for much of the year. The Fed's research staff tracks both series. When they diverge sharply, it signals that the job market is more complex than the headline suggests.