1
# Checks by column type
2
3
## Identifiers
4
5
- Distinct count equals row count. If not, the key is not the key.
6
- No nulls. A null id is a broken row.
7
- Consistent format. Mixed formats mean two sources merged.
8
- Check for leading zeros lost to a numeric read. Zip codes and account numbers are the
9
  usual casualties.
10
11
## Numeric
12
13
- Read as numeric, or did one bad value force text?
14
- Min and max physically possible?
15
- Zeros: real measurements, or missing values wearing a disguise?
16
- Negatives: possible for this quantity?
17
- Suspiciously round numbers in quantity suggest manual entry or a default.
18
- Units consistent throughout? Currency, weight and time are the usual mixed columns.
19
20
## Categorical
21
22
- Distinct values, listed. Read them.
23
- Same category, different spelling: case, whitespace, punctuation, abbreviation.
24
- A catch-all value like "Other" or "Unknown" holding a large share hides structure.
25
- Categories present in only part of the date range mean a taxonomy change.
26
27
## Dates and times
28
29
- Parsed as dates, or still text?
30
- Time zone: recorded, assumed, or mixed? Mixed is the common silent disaster.
31
- Any future dates? Any at the epoch, or at 1900-01-01?
32
- Gaps: whole days or weeks missing means a pipeline outage.
33
- Day-first versus month-first ambiguity: check whether any day exceeds 12.
34
35
## Text
36
37
- Length distribution. A cluster at exactly 255 means truncation upstream.
38
- Leading or trailing whitespace.
39
- Encoding damage: mojibake, replacement characters, doubled encoding.
40
- Embedded delimiters that may have shifted columns during a bad parse.
41
42
## Booleans
43
44
- How is it encoded? True/False, 1/0, Y/N, yes/no, or all of them in one column.
45
- Is null a third state with meaning, or just missing?
46
47
## Across columns
48
49
- Do totals equal the sum of their parts?
50
- Do dates order correctly: created before updated, start before end?
51
- Do foreign keys resolve? Count the orphans.
52
- Are mutually exclusive flags actually exclusive?
53