Engineering
Change the foundation, carefully
The database is the one place you cannot just roll back.
We make schema changes in safe, backward-compatible steps, test them against realistic data, and keep a path back, so evolving the data model never risks the live system.
How we change schemas without fear
Backward-compatible
Each step works with both the old and new code.
Small increments
Add, backfill, then remove, never all at once.
Reversible path
Every step has a tested way back before we take the next.
Where teams get burned
The database is the one place you can't just roll back
Code reverts in seconds; data doesn't forgive as easily.
A bad deploy can be undone by shipping the old binary. A bad schema change can leave data half-transformed, with no clean way back. That asymmetry is why confident teams treat migrations with extra care.
We change schemas in small, backward-compatible steps, add before remove, migrate data separately, and keep the old and new shapes working together until the switch is truly safe.
Highlights
How we stay safe
Backward compatible
Old and new code both work mid-change.
Tested on real data
No surprises in production.
Reversible
A clear path back at every step.
Deep dive
Migrations are deployments in disguise
The mental shift that prevents disasters: a migration is not a script, it is a deployment with a data payload. It deserves the same review, staging rehearsal and rollback plan as any code release, and the rehearsal must run against production-sized data, where the ten-second ALTER becomes a ten-minute lock.
Expand-and-contract is the pattern that pays the bills: add the new column or table, dual-write, backfill in throttled batches, verify, then remove the old shape in a later release. Each step is individually reversible, which is the whole point.
Long operations get instrumentation, progress logged per batch, resumable checkpoints, and a kill switch that stops cleanly. A migration you cannot pause is a hostage situation with your database.
We build software for teams who want their tools to fit the way they actually work — web, mobile, AI and the systems that tie them together. We write here about what we learn shipping it.