How to use Dataset Profiler
- Paste CSV or TSV, or choose a CSV, TSV or XLSX file. Select a workbook sheet by name or number when needed.
- Check delimiter and header settings. Choose whether surrounding spaces should be ignored in profiling.
- Optionally include sample values and column histograms, then profile the data.
- Read empty rates, distinct counts, observed types and ranges. Review candidate keys against the source's intended meaning; download the full report if needed.
Example: Dataset Profiler
Check a small order file before importing it, where one quantity is empty and one customer name repeats.
Options
- CSV or TSV text
- Paste the table to describe, or leave this empty and choose a file. The data is read and summarised; nothing is changed.
- Worksheet name or number
- Choose the worksheet used for column profiling. An empty choice uses the first visible sheet; otherwise enter its name or 1-based position.
- The first row holds the column names
- For column profiling, enable this when the first record names the columns. Without headings, refer to columns by their 1-based positions.
- Ignore spaces at the ends of a value
- On by default, so a cell holding only spaces counts as empty and " ABC " counts the same as "ABC".
- Show up to five example values per column
- Sample values are the first few distinct non-empty values in file order.
- Add a histogram for each column
- Number columns get up to 10 equal-width buckets and other columns get their most common values. The on-page list covers the first 20 columns and the download holds every column.
Supported inputs and limits
- Up to 5 MiB, 10,000 rows, 100 columns and 2,000,000 cells.
- A chosen table file replaces pasted text.
- Types describe observed cell syntax, without changing the source.
- Histogram preview covers the first 20 columns; the text download covers all columns.
- Number buckets are descriptive and do not establish a probability distribution or identify every outlier.
Where your input is processed
This tool does not upload your inputs. See the limits for anything saved on your device.
Profiling before an import
Before an import, look for empty required fields, repeated keys, numeric-looking columns with text markers and dates written in mixed formats. Review sample values alongside counts rather than trusting the heading alone. The profile describes the supplied file; it does not decide which source values are correct.
Why a guess is not a schema
The type in this report comes from the characters in the sample only. A column that holds only digits this month can hold a prefix next month, and a date written in short form may arrive in another format in the next export. The report is evidence about the file in front of you, so confirm the intended type against the data source before writing a schema.
Questions about Dataset Profiler
Why is my date column shown as text?
The type guess recognises only a few plain date shapes, such as 2026-09-29 and 29/09/2026, and keeps a written month as text because many formats are ambiguous. Check that column by hand before relying on it.
Why is my ID column called integer when it holds codes?
The guess describes the characters, not their meaning. A column of digits reads as a number even when the digits are a label. Treat the type as a hint and decide the real type for yourself.
Why is a numeric column called mixed?
At least one non-empty value is not numeric, such as n/a or a blank marker written as text. The column is reported as mixed so a numeric summary is never shown for data that partly is text.
Does the profiler change my file or upload it?
No. The table is read and summarised in this browser. Nothing is written, sent to a server, or changed on your device.
What is a candidate key?
A column with no empty cells and a different value in every row is named in the notes. It is a candidate only: a column can look unique in this sample and still repeat in the full dataset.