Interrogate a dataset before you trust it

Here's a dataset: {{paste sample rows plus column list, row count, and how it was collected}}.

Before any analysis, audit it:
1. For each column: type, unit, expected range, % missing, % suspiciously default (0, 1970-01-01, "unknown"), and cardinality.
2. Name every column whose meaning is ambiguous and say what I need to confirm with whoever built it.
3. Find the joins and duplicates: what's the true grain of one row? Is it what I think it is?
4. Selection bias: given how this was collected, who or what is missing from it? Be specific.
5. Time issues: timezone, changing definitions, backfilled rows, incomplete final period.
6. List the three questions this data can answer well, and three it looks like it can answer but can't.

Do not compute headline metrics yet. Tell me what would make them wrong.