Connect to and Prepare Data
Connecting and shaping data
Start on the Data Source page: pick a connector, then choose live or extract. Live queries the source with each interaction; an extract is a local snapshot that can be filtered, aggregated, and refreshed. Understand the logical layer (relationships between tables, cardinality) and the physical layer (joins and unions). Relationships are the modern default and preserve each table's own granularity; joins merge tables into one row-level result and can duplicate rows. Unions stack rows with matching columns; pivots turn crosstab-style columns into rows.
Preparation happens on the data source page: rename fields, set data types and geographic roles, set default formats and aggregation, create aliases, hide unused fields, split delimited strings, and use Data Interpreter to strip titles and footers from spreadsheet-like sources. Extract filters are applied when the extract is created or refreshed, not per worksheet; data source filters apply to every sheet using that source.
Traps: confusing union (more rows) with join (more columns); assuming a relationship filters rows exactly like a join; expecting Data Interpreter on every connector; blending when a relationship would do. Blends need a linking field and have primary and secondary sources.
Worked rule: with Orders (many rows per customer) and Customers (one row per customer), relate on Customer ID so order rows are not multiplied. If a physical join is unavoidable, compare row counts before and after. Always confirm field types before building views, because dates stored as strings break date filters and trend lines.