Skip to content
Saiku Cloud

The Saiku workbench

The workbench — Query Editor in the top nav — is the screen you land on when you click Open in Saiku on a cube from the Analyze page. Everything you do to get numbers out of a cube happens here.

Three regions: the cube and its fields down the left, the shelves you drop fields onto in the middle, and the result underneath. This page walks each one so the rest of these docs make sense.

The left panel

Cube picker. A dropdown at the top listing every cube your account can reach. Pick one and the panel below fills in. The refresh button beside it re-reads the catalog — worth a click if someone has just published a cube and you can’t see it yet.

Measures, first. These are the numbers the cube knows how to aggregate — sales totals, counts, averages. They’re grouped by measure group, with any Calculated measures in a group of their own underneath.

Dimensions, below. Each dimension opens to its hierarchies, and each hierarchy to its levels. A date dimension typically reads Time › Year › Quarter › Month › Day; you use whichever levels you want and ignore the rest.

The shelves

Four drop targets, stacked:

MEASURES. The numbers you want. Drag a measure here.

COLUMNS. What runs across the top of the result. Drop a dimension level to get one column per member.

ROWS. The same idea, vertically — and where most analyses put the thing they’re slicing by.

FILTER. Dimensions that restrict the query without appearing in the output. “Only 2026” belongs here: drop the Year level in, then choose which years.

Drag from the left panel onto a shelf. Levels land inside a hierarchy chip, so Product Family arrives under a Products chip — that grouping is what lets you drill without re-dragging. The × on a chip removes it.

There’s no submit step. Auto-run is on by default, so the result appears the moment the shelves make sense together. If you’d rather compose a complex pivot before spending a query, turn auto-run off in the Run dropdown and drive it manually.

The toolbar

Left to right:

File — New, Open, Save, Save as. The save icon carries a dot when there are unsaved changes.

Undo / Redo — full history of your query edits, not just the last one.

Run, with a dropdown beside it holding four toggles that change what the query actually does:

  • Auto-run — re-query on every change. On by default.
  • Non-empty — hide rows and columns where every measure is null. Usually what you want; turn it off when the absence of data is the finding.
  • Visual totals — make parent rows sum only the children left after filtering, rather than the whole cube. Reach for this whenever a filtered pivot’s totals look impossibly large.
  • Async — run long queries in the background so the UI stays responsive.

Swap axes — exchange Rows and Columns in one click.

Ask the AI — put the question in English and let Saiku build the query. See Ask the AI.

Tools — MDX…, Drill through…, Drill across…, Report titles…, and fullscreen. The MDX dialog shows the query Saiku generated and lets you hand-edit it; see MDX & export.

Export — XLS, CSV, PDF, and Email (when your server has mail configured).

Viewing the result

Under the shelves, a row of view buttons:

  • Grid — the pivot table. Totals at the edges, and clicking a parent value drills into its children.
  • Chart — the same result drawn. See Charts.
  • More — three compact views: Stats, Sparkline and Sparkbar, for when you want shape rather than precision.

A status line under the result reports the runtime and the result shape — 3 rows × 2 cols. It’s the quickest way to notice that a query returned far more or far less than you meant it to.

Tabs

The strip at the top runs multiple queries at once, each one independent and against any cube. Open a second tab to build a comparison without disturbing the analysis you already have.

Saving and sharing

Save commits the current state — cube, field placement, filters, sorts, the view you were in, the title. Reopening gives you back exactly what you left.

There’s a lighter option too: the whole query lives in the URL. Copy the address bar and send it, and your colleague opens the same query with nothing saved and nothing shared in advance. Save when you want it to persist and be findable; send a link when you just want someone to look at this one thing.

Where to go next