What is a data migration? Definition and risks
A data migration is any operation that changes the structure or location of existing data: adding a column, splitting a table, converting a format or moving from one system to another. In applications, schema changes are usually written as versioned scripts and applied in the same order on every environment. A well-designed migration can be replayed, verified and, where possible, reversed.
Definition written by Alex Dionisio, Senior Product Manager. Updated November 13, 2026.
Why it matters for a PM
Many features require changing data that already exists: renaming a status, splitting first and last names, merging accounts. That work is often invisible in estimates even though it carries the heaviest risks: lost information, downtime, or inconsistency between the old and new versions of the app. The PM should make it visible in the spec and the plan.
Example
An invoicing app wants to support several addresses per customer, but the schema only allows one. The team works in three steps: create the new table, copy existing addresses while keeping the old column, then switch the app over and drop the old column once checks pass. The migration is rehearsed on staging first.
Key points
- Schema migrations are versioned with the code and applied in a fixed order.
- Working in backward-compatible steps lets the schema change without stopping the service.
- A tested backup and a rollback plan should exist before any risky migration.
- Record counts before and after help detect data loss.
Common mistakes
- Leaving the conversion of existing data out of a feature estimate.
- Editing the production database by hand instead of through a versioned script.
- Removing the old structure in the same deployment as the new one.
Go further with Module
The courses and lessons that cover this concept:
Guides on this topic
- MCP vs API: the difference
When to expose an API, an MCP server or both, and what it changes for your product.
- SQL for Product Managers
The queries that let you decide on your own, the traps to avoid and AI for writing SQL, with a cheat sheet.