Dimensions, measures and field settings
How fields work in a dataset: roles, types, aggregates, default formats, descriptions, hiding, and each column's time zone.
A field is one thing people can pick when they build a chart: a column from the source, a calculated field, or a SQL expression. You manage them on a dataset's Fields tab. Settings here are defaults, so getting them right once means every chart over the dataset starts well.
Dimensions and measures
Every field has a role.
A dimension is something you group or filter by: region, channel, order date. Charts put dimensions on an axis or in a legend.
A measure is a number you add up: revenue, orders. A chart's numbers are measures.
When Halo builds a dataset, numbers become summed measures unless they look like keys, years or places, and everything else becomes a dimension. Change a field's role with the Dimension and Measure switch on its row. A store number that came through as a measure, for example, should be a dimension.
Types
The type of a field's values comes from the source column, or from the formula for a calculated field.
| On the field | Means | Example |
|---|---|---|
string | Text | Online |
number | A number, with decimals | 1249.50 |
integer | A whole number | 3400 |
boolean | True or false | Yes |
date | A calendar date | Jan 4, 2026 |
timestamp | A date and time | Jan 4, 9:30 AM |
The type decides what you can do with a field. Only numbers can be summed or averaged, and only dates and timestamps group by a time grain. Timestamps also group by hour and minute.
Aggregates
A measure's aggregate is how its values combine into one number per group. Set the default on the field's row. Authors can pick another for one chart in the chart editor's measure list.
| Aggregate | For | Result |
|---|---|---|
| Sum, Average, Median | Numbers | Adds them, the mean, the middle value |
| Minimum, Maximum | Any type | Lowest and highest, earliest and latest |
| Count | Any type | Rows where the value isn't empty |
| Count distinct | Any type | Different values, such as customers |
| 25th percentile, 75th percentile | Numbers, in the chart editor | The quarter points of the values |
A calculated field that already aggregates, such as SUM([Revenue]) / SUM([Orders]), has no aggregate to pick. Its formula decides.
Default formats
The Format on a numeric field's row sets how its values display everywhere: Automatic, Number, Currency in USD, EUR, GBP or CAD, Percent, Basis points, Duration (hh:mm:ss), or Custom… for a pattern of your own, such as #,##0.00, which you write under Custom format in More settings. A currency field also has Decimals under More settings: automatic, or fixed from 0 to 4. Number formats has the rules.
Labels, names and descriptions
Every field has a label, what people see, and a name, the stable one charts and filters use. Rename labels freely: nothing breaks. Names are made when the field is, and stay.
The description says what the field means, in a sentence: "Order value, before refunds." A new dataset's fields start with the column's description, from the catalog or your dbt project. It shows when people pick the field, and Halo AI reads it when it answers questions and builds charts. Describe with Halo at the top of the list asks Halo AI to draft descriptions for fields that have none, and you choose which to keep.
Hiding fields
The eye icon on a field's row hides it from pickers. Hide keys, ids and raw columns people shouldn't need.
A hidden field keeps working in calculated fields and in charts that already use it.
If published editions read the dataset, saving asks whether to keep them or withdraw them, since their figures can't follow the change. See When a dataset changes under an edition.
Halo AI and members' AI tools can't see it: it isn't listed, and a query that names it is refused. A calculated field that uses it still works for them.
To take a field out altogether, open More settings and choose Remove field. Charts that use it will need another field, so hiding is usually the better choice.
Folders and semantic types
A folder groups related fields in pickers, such as Revenue and Refunds under Money. Datasets built from tables start with a folder per table.
A semantic type says what the values are, beyond their data type: Country, US state, ZIP code, DMA, Latitude, Longitude, Currency, Percent, Email, URL, Identifier and others. Discovery sets it, and you can change it. It matters most for maps, which use it to find places, and for DMA codes, which show as DMA names and filter from the DMA list.
Time zones per column
Timestamps stored without a zone need to say which zone they were recorded in. Set Recorded in time zone under More settings. Halo then converts them to each report's zone before grouping by day or month, so an order placed at 11 p.m. lands on the right day.
Halo starts these columns as UTC when it builds a dataset, so check them if your source records local time. Plain dates and formula fields don't have this setting. Dates, time zones and fiscal calendars explains how zones fit together.