How to split a string and take one part in QuickSight

split(text, delimiter, position) cuts a string at each delimiter and returns the part at the given position, counting from 1. It is the simplest way to take the domain of an e-mail address or the first level of a category path, with one restriction: AWS documents it for calculated fields added during data preparation, not in an analysis.

Copy and paste

In the dataset (data preparation):

Domain of an e-mail address:

split({Email}, '@', 2)

First level of a path such as Food/Fruit/Apples:

split({Category Path}, '/', 1)

Second level:

split({Category Path}, '/', 2)

In an analysis, without split:

Everything before the first delimiter:

left({Category Path}, locate({Category Path}, '/') - 1)

Everything after the first delimiter:

substring({Category Path}, locate({Category Path}, '/') + 1, strlen({Category Path}))

Example

Data preparation only

Add the field in the dataset editor with Add calculated field. Every analysis that uses the dataset then has the part as a normal column, and for SPICE datasets it is computed once at refresh. AWS also notes that split is not supported in direct queries to Microsoft SQL Server.

When you can’t edit the dataset, use the locate formulas. locate(text, '/') returns the position of the first /, 1-based, or 0 when there is none. left takes the characters before it, and substring the characters after it, for a length that is at least as long as what is left.

Rows without the delimiter

A path with no /, such as Gift cards, has no position 2. With the locate formulas, locate returns 0, so left(..., -1) is not a valid length. Guard the case:

ifelse(
  locate({Category Path}, '/') = 0, {Category Path},
  left({Category Path}, locate({Category Path}, '/') - 1)
)

Check a few such rows after creating a split field as well, before you build filters on it.

Better in the source

If the delimited column is a list of values, such as tags, splitting it into one row per value belongs in the source query or the warehouse. QuickSight’s split returns one part per field, so “all products tagged X” across positions 1 to 5 would need five fields.

Checked against the Amazon Quick Sight function reference (October 2026).