Adding a new column sounds simple. In practice, it can stall operations, lock writes, and break downstream systems if done without care. The change must be atomic, explicit, and backward compatible until every part of the stack is ready.
Start by defining the column in your migration script with the correct type and constraints. Avoid default values on huge tables unless you can backfill in batches. Use NULL initially to prevent a full table rewrite in engines like MySQL and Postgres. Then, run background jobs to populate the data before making the column NOT NULL.
If the new column will be indexed, create the index after the data is backfilled. This reduces contention. For high-traffic workloads, consider online schema change tools to avoid blocking queries. Every change should be tested in an environment with production-scale data to detect hidden costs.