Database
Where your product keeps its records, such as users, bookings and payments, along with the rules for how they are stored, found and changed.
Updated October 2026 · with examples from our own builds
What does database mean in practice?
Most products rest on one main database. Document databases such as MongoDB store flexible records; relational databases such as PostgreSQL store tables with strict relationships. Either can run a serious product, and the choice follows the shape of the data and the team's experience.
Of everything in the stack, the database is the hardest to change later. Moving providers, restructuring records or splitting a database after launch means migrating live data, so its shape deserves more thought than any screen.
Ongoing costs are backups, indexes as the data grows, and access control: which services and which staff can read or write which parts.
Database at a glance
- Common choices
- MongoDB (documents) or PostgreSQL (tables)
- Hardest to change
- The shape of the data after launch
- Ongoing work
- Backups, indexes and access rules
Where has database come up in our builds?
MongoDB is the record for every chat room and message, Redis holds what must be fast or shared between server processes, and Socket.IO carries the live events; most of the chat work was keeping the three in agreement.
For a business running two legal entities with staff in six cities on one system, we recommended PostgreSQL with company and branch on every row. A recommendation, not a decision.
The admin console shares the app's database through a narrow MongoDB role that can read user collections and write only its own.
Common questions about database
MongoDB or PostgreSQL for my product?
Both run serious products. Relational tables suit data with many strict relationships, such as accounting or multi-branch records; documents suit records whose shape varies. Choose by the data and the team, and be wary of anyone who answers before asking about either.
Can two apps share one database?
Yes, and several of our builds do, but both must agree exactly on what each record means. Pakka and EverybodysPet keep the shared models byte-identical in both backends and check them with a tool.
Related terms
Sources
Planning a product and not sure where database fits? Ask us, or browse all 116 terms in the glossary.