The import wizard uses one CSV file and lets you select one or more entity types. Selecting several types creates a pipeline automatically; there is no separate mode toggle.
One entity type
Choose one type when each row describes one kind of record, such as contacts or organisations. A single-type import can still match existing records or use supported reference fields; it is not necessarily “create a new record for every row”.
Several entity types
A pipeline applies a separate mapping for each selected type to the same source row. For example, map the employer column to an organisation name, and the person’s columns to contact fields. The organisation is resolved first and linked to that row’s contact. The wizard displays and controls processing order.
Use stable external source IDs and a consistent platform name when matching across repeated imports. Do not assume a blank value is ignored: mapped empty standard fields can clear stored values.
When to use separate files
Use separate completed imports when dependencies are in different source files or the later import needs existing records. In particular, a Connections import requires both sides to exist already; placing it after Contact in the same job does not make newly created contacts available to that lookup.
A pipeline is not a database transaction. Some rows can succeed while others fail, and a failed job does not undo earlier writes. Review the error report and correct affected rows before trying again.