Create your first collection (5 min)

By the end of this Tutorial

• You'll have your own collection containing one dataset.
• You'll be able to distinguish collections, datasets, and metadata at the screen level.

This is a self-contained, condensed version of lesson 03 of the Analyst Path. All you need is one CSV (under 10k rows recommended).

1. Create a collection (1 min)

1. In the upper-right of portal, + Create → Collection.
2. Pick a name. Make it expressive of the project context so search finds it (e.g. 2026Q3-retention-analysis).
3. Fill in a one-line description. It shows as a hover preview in the tree and becomes a search match target.
4. Create. It appears in the left tree immediately.

2. A CSV as a dataset (2 min)

Inside the empty collection, Add dataset → CSV / JSON upload.

When the upload finishes, the dataset is registered. Click it and you'll see the preview table with inferred column types.

If you don't have a CSV at hand, grab one from a small scenario in dhub2-examples.

3. A one-line metadata pass (2 min)

This is the lowest-cost work that makes a dataset findable and trustworthy by other people.

• Description — One line. "What data, at what point in time, at what granularity."
• Tags — Biggest effect on search. Domain, team, and cadence are usually enough.
• Owner — Yourself or the data owner.

Next steps

• To put a first widget on top of the same collection, head to lesson 04 of the Analyst Path.
• When you're ready to share with teammates, grant Reader access at the collection level (see the related user docs).