What this checks
This one does not convert anything. It reads each column in your table, works out which unit dominates that column, and then flags every row that carries a different unit, plus every row that carries no unit at all. The output is a per-row verdict and a per-column summary of what was actually found.
That second category, the row with no unit, is the one that causes the expensive
mistakes. A bare 1.250 in a column of inches is almost certainly an inch
value, and a bare 1.250 in a column of millimeters is almost certainly a
millimeter value, but nobody can prove either from the document. Flagging it is the
only honest response.
How to read the verdict
OK means the row carries the unit that dominates its column. AT LIMIT means the row carries no unit, so it is assumed to follow the column but is not stated to. FAIL means the row carries a unit that contradicts the column, and the note names both, in the form col 2 reads mm, column is in.
The per-column summary is worth reading on its own. A column whose histogram is split evenly between two units is not a column with a few typos, it is two different lists that were pasted together, and the right fix is upstream of any conversion.
Worked example
The sample list has one column with seven values. Four carry millimeters, two carry inches, and one carries nothing. The dominant unit is millimeters, so the two inch rows fail and the bare row is flagged as having no unit. If those seven values were converted as a batch assuming the dropdown unit, four of them would come out wrong by a factor of 25.4 and the resulting list would look entirely normal.
That is the failure this page is built for: not a wrong conversion, but a correct conversion applied to a value that was never in the assumed unit.
Why unit mixing survives review
A mixed column usually arrives from a real cause. One supplier quotes in inches and the rest in millimeters. A table was extended by copying rows from an older revision. Someone converted the first page of a drawing and stopped. Each of those produces a list that reads as consistent at a glance, because the numbers themselves are perfectly plausible: 12.5 mm and 0.5 in are both ordinary values for the same feature.
Nothing about the numbers looks wrong, which is why the check has to be mechanical. Reading a column and noticing that row four is 25.4 times smaller than its neighbors is exactly the task human attention is worst at.
What it will not do
It does not fix anything and it does not guess. A bare value is flagged, not assigned. A mixed column is reported, not normalized. Deciding whether the odd rows should be converted or the column heading should be corrected is a document question, and getting it wrong in either direction produces a list that is confidently incorrect.
Convert the column once it is consistent →
Questions
Why flag a bare number instead of assuming the column unit?
Because the assumption is the thing being tested. A bare value in a column of inches is probably inches, but the whole point of the scan is to surface the rows where the document does not actually say. Treating it as a pass would hide precisely the uncertainty you opened the tool to find.
Does a mixed column mean the values are wrong?
No. Every value can be individually correct and the column still be unusable as a list. The scanner reports the inconsistency, not an arithmetic error. What you do about it depends on which supplier or which revision the odd rows came from.
Which units does it recognize?
The eight length units the site uses, in their word and symbol forms: mm, cm, m, km, in, inch, inches, ft, foot, feet, yd, mi, plus the quote marks for feet and inches. A unit written as part of an expression such as 3/8 in is recognized too.
Can it scan more than one column at a time?
Yes. Every column is analyzed independently and gets its own dominant unit, so a table with a millimeter column next to an inch column is read correctly rather than reported as uniformly mixed.
What about a column that has no units anywhere?
It is left alone. When no row in a column states a unit there is no inconsistency to report, and flagging the whole column would be noise. The scan reports contradictions between rows, not the absence of units in general.