Why we still reach for Postgres first

Most business data is relational: customers place orders, orders hold items, items come from suppliers. Pretending otherwise to save a schema decision on day one costs far more on day three hundred.

Boring is a feature

PostgreSQL has been in production for decades. Its failure modes are documented, its tooling is mature, and every managed cloud runs it. When a client's team takes the system over, they can hire for it.

• Transactions that keep money and stock consistent
• Constraints that reject bad data before it spreads
• Managed options such as Amazon RDS for backups and failover

One database, many shapes

The flexible parts fit too. A jsonb column holds the settings that differ per customer, full-text search covers the first version of a search box, and a queue table carries background jobs until volume says otherwise.

select id, settings->>'locale' as locale
from customers
where settings @> '{"newsletter": true}';

When we pick something else

Some workloads do outgrow it: very high-volume event streams, global low-latency key lookups, or analytics across billions of rows. We add a specialised store next to Postgres for that one job, and keep the business records where they are.

Start with the database you can reason about at 2 a.m., and add the clever one when the numbers ask for it.