How do you perform a test conversion?

Required tools

• A sandbox environment — Never test-convert in production.
• Complete dataset — Try it with everything, not just a selection.

Run the test

1. 1.Use the entire dataset, not a sample

   The point is to uncover problems that only arise at scale: timeouts, unsupported characters, fields that are too short. A small sample will not reveal these.

2. 2.Time it

   Note how long each step takes. This forms the basis for planning the actual production deployment and determining whether you need a weekend or an evening.

3. 3.Log everything that fails

   Errors must be corrected in the scripts or source data, not manually in the sandbox. If you correct them manually, the error will reappear the next time the script is run.

Verify the result

1. 4.Reconcile the count with the source

   Count the rows in each table in the source and in Dataverse. Discrepancies must be explained, not accepted—a lower count usually means that a filter has removed something you needed.

2. 5.Spot-check relationships

   Select twenty customers you know and verify that contacts, sales opportunities, and addresses are linked correctly. The right number with incorrect links is worse than an obviously incorrect number.

3. 6.Let users test with real data

   Ask some experienced users to perform common tasks in the sandbox. They will spot things no checklist can catch because they know what the data should look like.

4. 7.Write a deployment plan

   Sequence, timing, who does what, and what triggers you to abort. The trial conversion provides the figures that make the plan realistic.