03.1 What a dataset is
A dataset is a saved recipe for the rows you want to report on. It answers: which tables, joined how, which columns, filtered how?
Reports and dashboards don’t read tables directly. They read datasets. Build a dataset once and many reports can use it.
An example
Your database has an orders table and a customers table. You want sales by country.
A dataset could:
- take orders and customers,
- join them on customer id,
- keep order date, amount and country,
- leave out cancelled orders.
Every report built on it starts from those clean rows.
Where the rows come from
When you create a dataset you choose one of two sources:
| Source | What it means | Use it when |
|---|---|---|
| A database | Read live from your database each time a report runs | You need up-to-the-minute data |
| A warehouse cube | Read from a summary Analytics has already built and stored | The data is large and speed matters more than the last few minutes. See 06.1 Making things fast. |
How column names come out
Every column in the result is named after its table: amount from the table orders becomes orders_amount. This keeps two columns called id from clashing. You can rename any column, see 03.4.
What you can do in a dataset
| Guide | Covers |
|---|---|
| 03.2 Create a dataset and add tables | Starting, placing tables, previewing data |
| 03.3 Join tables | Linking orders to customers, and so on |
| 03.4 Choose and rename columns | What appears in the result |
| 03.5 Formula columns | Calculated columns, fixed values, IF/THEN |
| 03.6 Summarise rows | SUM, COUNT… so the database does the heavy work |
| 03.7 Filters and parameters | Keeping only the rows you want |
| 03.8 Stack tables (union) | Putting two sets of rows one under the other |
| 03.9 Ranking columns | Rank, row number, running totals |
| 03.10 Preview and save | Checking the result and saving |
| 03.11 Build a dataset on another dataset | Reuse one dataset’s result inside another |
Next: 03.2 Create a dataset and add tables
Last updated 12 October 2026