Salesforce
Salesforce Data Migration Testing: Four Layers to Check

Rishabh Kumar
Published On

Table of contents
Lorem ipsum
Lorem ipsum
Lorem ipsum
Lorem ipsum
Lorem ipsum
How Do You Test a Salesforce Data Migration?
Test it in four layers. Reconcile counts and totals between source and target, check field values and relationships on a carefully designed sample, run real business journeys on the migrated data as the people who’ll use it, and confirm that each persona sees exactly the records they should. Rehearse the full migration at least twice in a production-like sandbox before cutover, and keep the evidence.
Matching row counts are the start of the story, not the end. A migration can load every record and still leave quotes attached to the wrong accounts, cases owned by nobody and customers who can see each other’s data.
What Goes Wrong
Data migrations happen more often than teams expect. A first Salesforce implementation replaces a legacy CRM. Two orgs merge after an acquisition.
A business moves from one product to another, such as CPQ to Revenue Cloud, or changes its data model, as in an OmniStudio migration. The failure patterns are remarkably consistent.
Changing the data model brings risks of its own. Dig into one of the most common in OmniStudio migration testing.
Failure | What it looks like after go-live |
|---|---|
Broken relationships | Contacts on the wrong account, opportunities with no account, cases detached from their customer |
Ownership errors | Records owned by an integration user or an inactive user, invisible to the team that needs them |
Record type and picklist mapping | Values that don’t exist in the target, silently defaulted or rejected |
Dates and time zones | Close dates and SLAs shifted by hours, or a whole day |
Duplicates | The same customer loaded from two sources, or loaded twice on a rerun |
Missing history and files | Activities, notes, emails and attachments left behind |
Automation firing on load | Flows sending welcome emails to long-standing customers, assignment rules reassigning thousands of cases |
Sharing and access | Records visible to the wrong teams, or to nobody, after sharing recalculates |
Org merge collisions | Conflicting configuration, picklist values, users and duplicate accounts from the two orgs |
Almost none of these show up in a row count.
The Four Layers
Layer | What it checks | How |
|---|---|---|
1. Reconciliation | Every record arrived, once, and totals match | Counts per object and record type, sums of key amounts, duplicate checks, rejected-record logs |
2. Integrity | Values and relationships are right | A designed sample per object, compared field by field, plus orphan and lookup checks |
3. Business journeys | The data works for the processes that rely on it | Real journeys on migrated records, such as renewing a migrated contract, updating a migrated case or quoting an existing customer |
4. Access | Each persona sees what they should, and nothing else | Sign in as each role and check visibility, including records that should be hidden |
Layer 3 is the one most migrations skip, and the one that finds the problems users actually report. A migrated contract that can’t be renewed because a required field wasn’t mapped slips through reconciliation and integrity checks. It’s obvious the first time a salesperson tries.
Business users are the best judges of whether migrated data works for them, which is why layer 3 overlaps so closely with Salesforce user acceptance testing.
Designing the Integrity Sample
A random sample of migrated records mostly finds ordinary records, which were never the risk. Design the sample instead, so every risky shape of record is checked at least once.
One record per record type, per object.
One record per owner type, including records owned by queues, integration users and users who’ve left.
The oldest and newest records, where date and time zone problems tend to show.
The largest and smallest amounts, and any negative or zero values.
Records with the most related records, such as an account with hundreds of contacts and cases.
Records that hit a mapping rule, like a picklist value that was translated or a field that was split.
For org merges, records from both source orgs, including any the merge rules treated as duplicates.
Fifty well-chosen records, checked field by field, tell you more than five thousand picked at random.
How Many Rehearsals You Need
Run the full migration at least twice in a production-like sandbox before cutover, timing every step and running all four layers each time.
First rehearsal. Finds the mapping, sequencing and volume problems.
Second rehearsal. Proves the fixes, confirms the cutover fits its window, and tests the rollback plan.
Cutover. The same steps and the same tests. Switch off automation that shouldn’t fire on load, switch it back on afterwards, and run critical journeys before users return.
For org merges, compare configuration between the orgs before the first rehearsal, and test journeys as users from both original organisations.
That last check before users return is a smoke test. Keep it short and decisive with smoke testing.
Testing a Migration with Virtuoso QA
Virtuoso QA covers layers 3 and 4, where migrations most often let users down, and runs exactly the same checks at every rehearsal.
Plain English journeys work on migrated records, such as renewals, amendments and case updates, through the browser and API steps. Journeys and visibility checks run as each role, including users from both sides of a merger.
Because checks are built from reusable library checkpoints, each rehearsal and the cutover itself run identical tests, so results compare directly. Touchstone can turn your mapping and cutover documents into requirements that cite their source, and propose journeys for a named person to approve.
Virtuoso QA's Salesforce test automation runs identical, approved journeys at every rehearsal and at cutover, as every persona, so results compare directly and nothing slips between runs.
Related Reads
Frequently Asked Questions
Is a Matching Row Count Enough to Sign Off a Migration?
How Do You Test a Salesforce Org Merge?
Should Automation Run During a Data Load?
Who Should Sign Off a Salesforce Data Migration?
Written by

Rishabh Kumar
AI Testing & Quality Engineering Contributor
Rishabh leads business development and marketing at Virtuoso QA. He writes about AI-powered test automation, end-to-end testing, self-healing tests, and quality engineering, exploring how these capabilities can improve test coverage, reduce test maintenance, and support reliable software releases.










