Managed Postgres or your own: an honest comparison
Managed database services are excellent and cost several times what the same workload costs on a server you administer. Both facts matter.
This decision gets made on reflex in both directions - either "managed, obviously" or "we can run a database" - and both reflexes are wrong about half the time.
What managed actually buys
Automated backups you have some reason to trust. Point-in-time recovery. Failover you did not build. Patching. Metrics out of the box.
Reframe that as: someone else is on call for your database. For a team without an operator, that is worth a lot, and paying for it is a rational decision rather than a lazy one.
What it costs
Typically three to five times the raw compute price for equivalent resources, plus storage and I/O charges that are easy to underestimate. At small scale the difference is noise. At larger scale it becomes one of the bigger lines on the bill.
You also give up some control: extension availability, configuration parameters, and major-version upgrade timing on someone else's schedule.
When self-hosting is defensible
- Someone genuinely owns operations and is on call
- Backups are automated **and** restores are tested on a schedule
- Monitoring and alerting exist before an incident, not after
- Your availability requirement tolerates the failover story you actually have
That third point is where most self-hosted setups fail. A backup nobody has restored is a hypothesis, not a backup.
What we usually recommend
Managed for client production systems, unless cost genuinely dominates and the team has the operational maturity. The premium buys a reduction in the class of 3am problem that ends worst.
Self-hosted for internal tools, staging, and workloads where a few hours of downtime is survivable - which is more systems than people admit.
The wrong version of this decision is not either option. It is self-hosting because it looked cheaper, without the operational work that makes it safe.