All docs pages (49)
Data Loader
Paste rows from a spreadsheet, map columns to fields and insert, update, upsert, delete or restore records in batches of up to 200.
Updated 2026-11-09
Data Loader runs Insert, Update, Upsert, Delete and Undelete on rows you paste from Excel, Numbers, Google Sheets or a CSV file. Open Data → Data Loader.
How to use it
- Under Operation, pick the operation, then pick the object in Object.
- Optionally change Batch size (1 to 200, default 200) and Threads (1 to 10, default 6), the number of batches sent at the same time.
- For Insert, Update and Upsert, paste your rows into Data. The first row is read as column headers.
- Check the Field mapping grid (Column (from paste), Salesforce field, Sample). Each column is matched to a field by API name or label, ignoring case, spaces and punctuation. A column with no match is set to — Skip this column —. You can change any mapping. On Insert, required fields are marked
*. - For Upsert, pick the key in Match on: the record Id or an external ID field.
- Tick Treat blank cells as null if blank cells should clear the field. Otherwise blank cells are left out and the field is not changed.
- For Delete and Undelete, paste record Ids into Record Ids instead, separated by new lines, commas, tabs or spaces. Duplicates are counted once. For Delete, type DELETE in the confirm box.
- Click the run button, for example Insert 120. For Undelete it says Restore.
The run button stays disabled until at least one column is mapped. Insert also needs every required field (one that can't be blank, has no default and isn't an auto-number), Update needs the Id field, and Upsert needs the Match on field. If two columns map to one field, the rightmost wins and a warning says so.
What you see
While it runs, a progress line counts queued and processing batches and succeeded and failed records. Cancel stops new batches; batches already sent still finish.
At the end, a summary says, for example, "Inserted 118 of 120 Account records · 2 failed". Upsert also counts records created and updated. Failures are grouped by Salesforce error code, such as REQUIRED_FIELD_MISSING ×2. Show N rows opens every row (#, Id, Status, Detail), Show only failures filters it, and Copy failures (CSV) copies the failed rows.
What it reads and writes
It reads the org's object list and the object's field describe. It writes through the REST API's sObject Collections endpoint, one request per batch. Insert creates records, Update changes them by Id, and Upsert creates or updates by the match field. Delete sends records to the Recycle Bin. Undelete restores them through the SOAP API's undelete call. It does not use the Bulk API.
Each batch allows partial success, so one bad row doesn't stop the rest of its batch. The run stops sending new batches after an error that would fail every batch: an expired session, a network error, the org's API request limit, Salesforce unavailable or in maintenance, or too many requests. Rows never sent are marked OPERATION_CANCELLED.
There is no undo, and no copy of the old values is kept. Deleted records can be restored with Undelete while they are in the Recycle Bin.
Limits
- At most 200 records per request.
- Checkbox cells: in version 1.1.13 only the exact text
truecounts as checked. Anything else, includingTRUEfrom Excel,1oryes, is written as unchecked. In the release after 1.1.13, the usual spellings of true and false (TRUE,1,yes,FALSE,0,noand so on) are read correctly, and any other value is rejected for that row instead of clearing the checkbox. - A number cell that isn't a number is sent as text, and Salesforce rejects that row.
- A date-time like
2026-11-09T10:30, with no time zone, is read as your browser's local time and converted to UTC. Other date values are sent as typed. - Address and geolocation fields can't be mapped whole; map their parts, such as street and city.
- A cell containing a line break is split into separate rows, even inside quotes.
- If the run stops on one of the errors above, version 1.1.13's summary marks every row as failed, including rows in batches Salesforce already saved. Check your org before running the load again. In the release after 1.1.13 the summary keeps the saved rows as successes.
Related
- SOQL Builder: query records first, for example to collect Ids to delete
- Data Migrator: copy records between orgs, with a dry run and rollback
- Smart Data Generator: create test records without a spreadsheet