Your first mapping

This walks the core loop end to end: scan → map → publish. If you followed the Quickstart, the demo already has scanned source values waiting, so you can do every step below without connecting a warehouse.

1. Open Review

Review (left nav) is the cross-table inbox of source values that have been scanned but not yet mapped. Values are grouped by table and ordered by frequency — the count next to each one is how many warehouse rows it covers.

Start at the top. The most frequent few values usually cover the vast majority of your rows; the long tail can wait, or stay unmapped indefinitely. You are not expected to map everything.

2. Map a value to a record

Take a value like Brasil. In its row, open Choose record and pick the record it belongs to — Brazil. That creates a draft: a mapping saved in Zug Zug but not yet visible to your warehouse.

A few ways to move faster:

• Not sure which record? Choose Suggest for an optional AI suggestion (bring your own OpenAI/Anthropic key in Preferences). It only proposes — nothing is saved until you confirm it.
• Two records are really the same thing? Select them in the grid and merge them into one.
• Junk value you'll never want? Skip it — it leaves Review without creating a record.

Drafts made by teammates land in Approve teammates' work — an editor reviews them before they go out. Roles decide who can map versus who can publish.

3. Publish

When you've mapped what matters, hit Publish. One publish folds every draft — and any record edits since the last version — into a new numbered version (v17 → v18) and materializes the dim_<x> / map_<x> tables.

Everything before publish is invisible to your warehouse. Publish is the single act that makes it real, and the audit log records who published what, and when. See Publishing for how the result reaches your warehouse (pull-first by default).

4. Use it

The output is two plain tables per list — map_<x> turns every source spelling into a record's key, dim_<x> holds the approved records and their attributes. One join resolves messy input to an agreed value:

select o.*, d.label as country
from raw.orders o
left join map_country m on m.raw = o.shipping_country
left join dim_country d on d.key  = m.key

Next

• Connect a warehouse — point Zug Zug at your own columns.
• Pull API — feed published records into dbt, Fivetran, or any pipeline.
• Publishing — versions, pull-first, and direct write.