Salesforce Data Migration Checklist: From Legacy System to Org

Data migrations are where Salesforce projects fail. Not in design, not in development — in the moment when 50,000 Accounts land in production with the wrong owner, blank required fields, or duplicated records. This checklist covers what actually needs to happen before, during, and after a migration.

Prerequisites

Salesforce org with the target schema ready
Source data exported (CSV, SQL dump, or API)
Admin access and a sandbox to test against

Phase 1 — Analyze and Map

Before touching any tool, answer these questions:

Object inventory:

What objects need to be migrated? (Accounts, Contacts, Opportunities, custom objects?)
What are the relationships? (lookup, master-detail, many-to-many via junction?)
Are there objects with external IDs from the source system?

Data quality:

What percentage of records have missing required fields?
Are there duplicates? What's the deduplication key?
What encoding is the source data? (UTF-8 vs. latin-1 can break accented characters)

Field mapping table:

| Source Field | Salesforce Field | Transform Needed | |-------------|-----------------|-----------------| | customer_name | Account.Name | None | | email | Contact.Email | Lowercase | | created_date | Account.CreatedDate | ISO 8601 format | | account_type | Account.Type | Enum mapping |

Phase 2 — Prepare the Org

// Temporarily disable validation rules and triggers for bulk load
// In Setup: deactivate validation rules for migrated objects
// In Apex triggers: add a bypass mechanism
 
public class TriggerUtils {
    public static Boolean isMigrationMode = false;
}
 
trigger AccountTrigger on Account (before insert, before update) {
    if (TriggerUtils.isMigrationMode) return;
    // Normal trigger logic...
}

Also disable:

Workflow rules that send emails
Process Builder / Flow automation that creates child records
Duplicate rules (temporarily, with logging)

Phase 3 — Load Order Matters

Always load in dependency order:

1. Users (if migrating owners — match by email)
2. Accounts (parent objects)
3. Contacts (requires Account lookup)
4. Opportunities (requires Account lookup)
5. Opportunity Contact Roles (requires both Opp and Contact)
6. Custom junction objects last

Phase 4 — Data Loader Strategy

# Use Salesforce CLI for large volumes
sf data import bulk \
  --sobject Account \
  --file accounts.csv \
  --external-id External_Id__c \
  --wait 10
 
# Or Data Loader with upsert mode (recommended over insert)
# Upsert: insert if new, update if external ID matches

Key settings:

Use Upsert not Insert (idempotent — safe to re-run)
Set a meaningful External ID field for every object (store source system ID)
Batch size: 200 for most objects, 10 for complex objects with many lookups
Run during off-hours if possible

Phase 5 — Validate

-- Check for records with missing required lookups
SELECT Id, Name, AccountId FROM Contact WHERE AccountId = null
 
-- Count by owner to spot assignment issues
SELECT Owner.Name, COUNT(Id) cnt FROM Account GROUP BY Owner.Name ORDER BY cnt DESC
 
-- Verify total counts match source
SELECT COUNT(Id) FROM Account  -- should match source system count

Post-load checks:

[ ] Record counts match source system (±0)
[ ] Spot-check 10–20 records manually against source
[ ] Relationships resolve correctly (no broken lookups)
[ ] Required fields populated on all records
[ ] Page layouts display as expected

Phase 6 — Re-enable and Go Live

1. Re-enable validation rules
2. Re-enable triggers (remove migration bypass)
3. Re-enable workflow/flows
4. Run duplicate detection scan
5. Assign record ownership if batch-loaded as admin
6. Notify users and run smoke tests

Rollback Plan

Keep the source system read-only for 30 days post-migration
Export all migrated records with their new Salesforce IDs before go-live
Document the delete query in case you need to wipe and re-run:

// Delete all records with External_Id__c populated (migrated records)
List<Account> toDelete = [SELECT Id FROM Account WHERE External_Id__c != null];
delete toDelete;

Common Pitfalls

Loading without External IDs: you can't safely re-run the migration, every re-run creates duplicates
Ignoring timezones: date/datetime fields shift based on org locale — always use UTC in your source files
Forgetting record types: records loaded without a RecordTypeId get the default, which may not be correct
Owner mapping: Salesforce user IDs differ between orgs — always map by email, not ID