Cube library

The cube library is a set of working Mondrian cubes for domains most companies recognise. Each one comes with its own tables and a synthetic dataset, so provisioning a template gives you something you can query in about ten seconds — before you’ve modelled anything of your own.

Two reasons people use it. The obvious one is a demo: a cube with plausible-looking numbers in it, ready for a screen share. The less obvious one is learning by reading — these are complete, correct Mondrian schemas, so if you’re stuck on how a bridge table or a time hierarchy is meant to look, download one and read it.

What’s in the library

Template	Cubes	What it models
saas-metrics	Subscription Events, Daily MRR Snapshot	Recurring revenue — MRR, ARR, churn, retention cohorts.
ecommerce	Order Lines, Orders	Online retail — revenue, AOV, gross margin, repeat-purchase rate, top SKUs.
hr-analytics	Headcount Snapshot, Compensation Events	Headcount, attrition, comp distribution, promotion velocity.
healthcare-reporting	Encounters, Diagnoses	Provider reporting — visit volume, diagnosis mix, length of stay, 30-day readmits.

All four are Mondrian 4 and Postgres-flavoured. The synthetic data has deliberately realistic distributions — power-law customer sizes, seasonal trends, cohort retention curves — so charts built on it look like a real business rather than uniform noise. Nothing in them is real: every name, email and address is generated, so you can share a template-derived dashboard publicly without a compliance conversation.

Provisioning a template

Templates live on Data sources, not on a gallery page of their own — because a template has to be provisioned into a specific warehouse.

1. Open Data sources and find the connection you want to provision into.

2. Click Use a template →. A panel opens on that row.

3. Pick the template, and name the target schema — the schema inside your warehouse the tables get created in. public is the default. Add a table prefix like saiku_ if you’d rather keep the template’s tables visibly separate from your own.

4. Click Provision & deploy. Saiku creates any tables that aren’t already there, seeds sample rows into any that are empty, rebinds the Mondrian schema to that warehouse and target schema, and saves it as a cube. You land in the cube when it’s done — a few seconds, typically.

Provisioning is idempotent. Run it twice and the second run is a clean no-op: existing tables are left alone, non-empty tables aren’t re-seeded. Nothing you’ve put in those tables gets overwritten.

Caution

This connection needs write privileges. Unlike everything else in Saiku Cloud, provisioning creates tables and inserts rows — so a read-only role will be refused, with a message telling you exactly that. Point it at a schema your credentials own, or use a role that can create tables. If your production warehouse is read-only by policy (and it should be), provision templates into a scratch database instead.

After provisioning

The deployed schema is yours. It shows up in Schemas like anything else — open it on the canvas, edit it, version it, rename it. The template itself is untouched and unchanged for everyone else.

If you want to read the definition rather than run it, the row menu in Schemas will hand you the same cube as Mondrian XML, Mondrian YAML, or canvas JSON.

When a template isn’t the answer

Templates provision their own tables and data. They are not a way to point a ready-made cube at tables you already have — the names and columns won’t line up, and there’s no mapping step that would make them.

For your own data, the right tools are:

• Draft with AI — profile your actual warehouse, describe the cube in a sentence, edit the proposal.
• The visual canvas — build it exactly, when you already know the model you want.

A reasonable middle path: provision the template that’s closest to your domain, read how it models the tricky parts, then build your own with that shape in mind.

Related

• Data sources — where the Use a template button lives.
• Schemas — where the deployed cube lands.
• Schema designer — build against your own tables instead.
• Mondrian schemas — the reference behind everything the templates demonstrate.