How we verify a statistic, step by step

By Nan Liang, data editor · 2026-11-13

Here is a real example from our remote work report, walked through the four gates.

The claim: a large share of the U.S. workforce worked remotely in recent years. Blogs quote anything from 20% to 60% for this, which is your first warning sign.

Step 1: find the primary source

The most quoted version traces back to surveys by major research firms, not to blogs. We go to the firm that ran the survey.

Step 2: open it and read the question

Remote work numbers move around because the questions differ. "Worked remotely at least one day last week" gives a much bigger number than "worked remotely full time". We record which question the figure answers.

Step 3: write the claim with its scope attached

On statfern you will see the figure, the institution, the year and the sample next to each other. If any of those four pieces is missing, the entry does not ship.

Step 4: the random re-check

Before publication, one editor re-opens a random 10% of entries at their sources, cold. On our first remote work report, that sample check caught one entry where the year had drifted by one. The whole report went back for a full re-check, and the fix shipped with the correction noted.

That is the whole method. It is not clever. It is just slow, and that is the point.