Salesforce CSV over 50,000 records: split it into safe batches

Use a row-count split, not a size guess

Salesforce documents a maximum of 50,000 records for one Data Import Wizard job. Split a larger file into 49,000-row batches so every file stays under the limit even if you edit a few values later.

• More than 50,000 records: choose row-count splitting.
• The job fails with a mapping error: compare headers and required fields first.
• The same Lead, Contact, or Account appears twice: remove duplicates before importing.

• Find CSV rows with inconsistent column counts →Diagnose malformed rows before uploading them to a platform.
• Remove duplicate records before importing →Check duplicate IDs or email addresses before you create a second record.

Check the object and import key

Data Import Wizard supports several standard objects and custom objects, but the required columns depend on the object you select. Decide whether the file is for Leads, Contacts, Accounts, or another object before splitting so every batch uses the same mapping.

• Keep the header row in each output file.
• Use the same external ID or Salesforce ID strategy for every batch.
• Test a small batch first, as Salesforce recommends, before scheduling the full import.

Create batches in LocaCSV

LocaCSV shows the input row count and the output counts so you can reconcile the batches before Salesforce sees them.

1. 1Save the Salesforce export unchanged and open the row-count split workspace.
2. 2Select the CSV and enter 49,000 data rows per file (or a lower limit required by your runbook).
3. 3Run the split and download every part with its header row.
4. 4Name the files in order, such as leads-001.csv, leads-002.csv, and keep the original export for rollback.

Reconcile the batches before Data Import Wizard

• The row counts add back to the original record count.
• All batches contain the same headers and column order.
• The chosen ID or external ID is present and unique where your mapping requires it.
• The first test batch completes without creating unexpected records.

• Remove duplicate records before importing →Check duplicate IDs or email addresses before you create a second record.
• Find CSV rows with inconsistent column counts →Diagnose malformed rows before uploading them to a platform.

Handle errors without losing the source

If Salesforce returns failed rows, export the error file, fix only the reported columns, and retry that batch. Do not merge successful and failed rows blindly; keep the batch number in your import log.

• Find CSV rows with inconsistent column counts →Diagnose malformed rows before uploading them to a platform.
• Check or convert the CSV character encoding →Use this when Japanese, Korean, or accented text is garbled after export.

Make the next Salesforce import predictable

Once each batch is below 50,000 records and reconciled, continue in Salesforce Data Import Wizard and verify the result count after every job.