mixed types in a column — A column holds numbers and text (no error raised)

Not an error. The column parses, normalizes, and produces a canonical field whose values are inconsistently typed — which breaks the maths downstream rather than at ingest.

Start free All protocols

Diagnosis

When you hit this

A numeric column containing OFF, N/A, or --- for missing readings.
Locale decimal commas, where 23,5 parses as text while 23.5 parses as a number.
Units appended to values, such as 72.1C instead of 72.1.

Resolution

How to fix it

Use empty cells for missing data, not sentinels
An empty cell is unambiguous. OFF and N/A are text and will be carried as text.
Normalize the decimal separator before export
A locale comma is the most common cause and is invisible on inspection.
Strip units from values
Units belong in the header, where the corpus reads them, not appended to each value.

Normalization

What Forge does about it

Forge will not silently coerce these. A value it cannot read as a number stays as text, and downstream the read-time validators are what catch the consequence: a field that stops varying because every reading is now an unparsed string trips the stuck-sensor check after 20 identical readings.