Skip to content

Create custom groups

4 min read · Updated 18 Nov 2026

A custom groups dimension puts the values of a field into groups you define yourself, by rule: sites into regions, status codes into “Open” and “Closed”, countries into sales territories. The result behaves like any other dimension, so every question can break down by the same groups. If you are looking for custom groupings, value mappings or a CASE WHEN dimension, this is it.

A bucket cuts a number into even bands. Custom groups do the job a bucket cannot: you choose which values go together and what each group is called.

Before you start: open a semantic layer and go to Business Context → Dimensions, then select Add → Custom groups.

Create custom groups
  1. The dimension Add menu with Custom groups among the kinds

    In the Dimensions tab, select Add, then Custom groups: it puts a field's values into groups you define yourself rather than the even bands a bucket cuts.

  2. A new custom groups dimension with its editor greyed out below Base field

    Name it and add a Description if it needs one, then choose the base field: every rule reads it, so the editor below stays greyed out until you do. Select field opens the columns these groups can classify.

  3. A custom groups dimension with its first group named Open

    Choosing the base field opens the first group row. Name the group Open and set its condition beneath: Contains pending, read against status. For a condition those boxes cannot express, turn on Advanced. Add Group starts the next rule.

  4. Two group rules with the narrower one below the broader one

    Rules are read from the top down and the first match wins, so a row reading pending_payment would be counted as Open. Move the narrower rule above it with the up arrow, or drag it by its handle.

  5. The Everything else fallback with its value list open

    Set Everything else for rows no rule matched. It arrives as the group name Other, which you can rename; the list beside it can return null instead, or leave those rows out of the dimension.

  6. The saved Order state dimension listed in the Dimensions tab

    After you select Create, Order state sits in the Dimensions list, saved into the local draft, with the column its rules read named under it.

Only a plain, existing column qualifies as the Base field. Calculated columns are not offered, and neither are dimensions carrying a date level or date part, because the group stores the column underneath and that transformation would be silently dropped. Buckets, drill-throughs and other custom groups cannot be a base field either. By default the picker lists table fields; Settings → Custom groups can make it offer dimensions as well.

Once the base field is chosen, set Returns — the data type every group value and the fallback must be. A group whose value is not that type blocks publishing, so choose it before naming many groups.

Rules that already say more than the simple condition boxes can show open in Advanced with the control locked on. Rules read only the base field’s own table. The generated column belongs to that table, so the pickers offer its fields and nothing else. In a rule, Like and Not like use the SQL wildcards % and _, not the * and ? of other filters; see Custom group rules use SQL wildcards.

Order decides the answer. The rules are evaluated top to bottom and the first match wins, so a row that satisfies two rules is counted only under the higher one. Put your most specific rules above your general ones. Reorder with the up and down arrows on each row, or drag a row by its handle. A row that matches no rule takes Everything else, or no value where there is none.

Use Show distribution to count the rows in each group without publishing anything. It shows each rule only the rows no rule above it took, so it is also how you catch a rule that a broader one above has swallowed. The button needs a complete, named group and a live data connection.

Changing the base field later keeps your work: a simple rule follows the new field and its values are retyped for it, so "100" on a text field becomes 100 on a numeric one. A rule built in the advanced editor keeps exactly what it says, and is marked unreachable where it reads a field the new base field’s table does not have.

Publishing is blocked until the dimension has a base field, at least one group, and group values matching Returns. Semantic layer issues lists what is outstanding.

In Settings → Custom groups, choose what the Base field picker offers:

  • Leave Show dimensions as options off to choose from table fields.
  • Turn it on to list eligible dimensions before the table fields.
  • Also turn on Show only dimensions to hide the table fields. This option is available only while Show dimensions as options is on.

The Custom groups section of Settings with Show dimensions as options on and Show only dimensions available below it

Both options are off by default and are saved for the semantic layer. Turning the first option off also clears the second.

Only dimensions backed by a single existing, non-calculated field without a date transformation are offered. Buckets, drill-throughs and other custom groups cannot be used as the base field.

Choosing a dimension stores its underlying table field. The closed picker shows that table and field, so two dimensions backed by the same field produce the same selection. Renaming or deleting the dimension you chose does not change the group’s field binding. These settings control the picker options; they do not restrict document editing or agent authoring.

Custom-group rules preserve physical column names, including case, spaces and punctuation. References use your data source’s identifier escaping.