Reading the profile
Open a file and every column gets a card. The badge shows the detected type: whole numbers, decimals (including decimal commas), dates with their day/month order, yes/no values or text.
- Missing counts empty cells and placeholders such as NA, null or a dash.
- Distinct is the number of different values; a column where every value is distinct may be a key.
- Patterns spot emails, URLs, domains, IP addresses, UUIDs, phone numbers and postcodes.
To fix what you find, CSV Lark's CSV cleaner trims spaces and empty rows, and its duplicate remover deals with repeated rows.
Frequently asked questions
What does the profiler check?
For every column it reports the detected type, empty cells, distinct values, the most common values and a small histogram, plus minimum, maximum, mean, median and standard deviation for numbers and dates. It then flags problems such as mixed types, stray spaces, case variants, duplicated IDs and outliers.
How are outliers found?
A number counts as an extreme outlier when it lies more than three interquartile ranges below the first quartile or above the third (Tukey's far-out fences). That is deliberately strict, so the warning points at values worth a second look.
Is the standard deviation the sample or population one?
The sample standard deviation (dividing by n − 1), the same as STDEV.S in Excel and sd() in R.
Does it work with large files?
Yes. Parsing runs in a background worker and every value is counted, so a file with a few hundred thousand rows profiles in a few seconds on a typical laptop.