Column NULL Audit — Solution & ExplanationProblems / Column NULL Audit / EditorialApproachThe output rows describe *columns*, so no GROUP BY over data values can produce them — this is a manual unpivot of four independent whole-table aggregates:

For each audited column, one aggregate SELECT: 'email' AS column_name, COUNT(*) - COUNT(email) AS null_count, and ROUND((COUNT(*) - COUNT(email)) * 100.0 / COUNT(*), 2) AS null_pct.

UNION ALL the four blocks and ORDER BY column_name.

The core trick is COUNT(*) vs COUNT(col): COUNT(*) counts rows, COUNT(col) counts non-NULL values, so their difference is the NULL count. The empty-string email in the data is deliberately not counted — '' is a value, and treating it as NULL is a classic profiling bug.Key InsightCOUNT(*) - COUNT(col) is the canonical NULL profiler, and it distinguishes NULL from empty string automatically. source has no NULLs and still reports a row (0 / 0.00) because each UNION ALL branch is a whole-table aggregate that always yields exactly one row — unlike a filtered GROUP BY, whose empty groups simply vanish.Hints

The output rows describe columns, not data groups — that calls for a manual unpivot: one aggregate SELECT per column, stacked with UNION ALL.

COUNT(*) counts all rows while COUNT(email) counts non-NULL emails, so COUNT(*) - COUNT(email) is the null count. Empty strings are non-NULL, so they are (correctly) not counted.

null_pct is ROUND((COUNT(*) - COUNT(col)) * 100.0 / COUNT(*), 2) — multiply by 100.0 to force float arithmetic before rounding.

Solve Column NULL Audit yourself →