Data check: can the number be trusted?

What it measures

Rankings of “who grew the most” reward the strangest numbers, and the strangest number is often a single day.
For the last 90 days of each daily series the check marks gaps, isolated spikes, isolated drops and revisions.
The data is never corrected: the day is marked, and the 30-day change is shown with and without it.

Isolated spike: a day above 5 times the median of the 14 days around it, with both neighbouring days below 2 times.
Isolated drop (levels such as value locked): a day below half that median, with both neighbouring days above half.
Gap: a day the source did not publish. A published zero is data, not a gap.
Revision: a past day that changed by more than 1% between two downloads made on different days.
Cleanliness is the share of days with no mark, from 0 to 100.
A marked day is not treated as an error. It may be a real one-day event. Either way it is not a trend. The spike and drop checks require at least eight surrounding observations, both adjacent days and a positive reference median. Recent dates can lack enough context. Revision checks require earlier downloads. No flag does not establish that every check was possible.