Datasets

A dataset is the bridge between your raw data and the widgets that draw it. Create one in the Dataspace, or let a widget's slice write one there. Everything you have is listed in the Datasets panel.

Two kinds of dataset

1. General datasets — created in the dataspace, and ready to bind to a widget as soon as they are ready.
2. Slice datasets — written to the dataspace by a widget's slice. While the widget's slice is on, the dataset updates in real time as people make selections. A slice dataset cannot be bound to a widget directly.

💡 Tip

The usual pattern: start from a slice dataset, use it to pull out the part of the source data that needs work — or to do similar work on related data in the dataspace. Turn the result into a new general dataset, and bind that to your next widget.

What you can do afterwards

Renaming and deleting happen in the Datasets panel.

1. General datasets — rename and delete.
2. Slice datasets — delete.

What changed in v2.0.0

• Background loading — a new dataset appears immediately and fills in as its columns become ready. Binding is unavailable until it is complete, so nothing draws a half-built answer.
• Bindings survive renames — a binding refers to the column itself, not its title. Renaming headers or rebuilding a dataset no longer silently unbinds a chart.
• No collisions — before a dataset is committed, the target area is checked; overlapping an existing dataset or an array-formula spill is reported instead of quietly overwritten.
• Create from spilled array results — when you select a contiguous range to create a dataset, cells an array formula has spilled into that range contribute their computed values; before this version, a selection containing them was refused outright. A header check tells you what was detected.
• Capacity limits — up to 2,000 records of up to 15 fields per dataset, and slice datasets are subject to the same limits.

Best practices

1. One dataset per worksheet when a dataset may grow — leave room in the row or column direction so nothing runs into its neighbours.
2. Keep a slice dataset and its source dataset in the same worksheet — shorter formula chains, easier to check.
3. Give names that carry meaning and location — Sales!Annual_Revenue beats Sheet1!A. These are also the names the assistant uses when it refers to a dataset.

Capacity limits and loading status

A dataset holds at most 2,000 rows across at most 15 columns. Asking for more is refused at creation with the reason, and the suggestion to aggregate first. Binding stays unavailable until loading finishes; the Datasets panel shows progress while it works. See Datasets That Load in the Background.

Where to go next

• Dataspace — preparing and calculating your data
• Data Binding — how widgets consume a dataset
• Data Slice — where slice datasets come from
• Datasets That Load in the Background — progress and capacity limits

---

Datasets are the foundation of everything you visualize. Put them together with widgets and the dashboards take off.

The datasets are ready — now for the visual design! 🚀