I’ve found that good continuous delivery pipelines tend to look like the following:

1. Compile
2. Run tests (unit, integration, checks, linters)
3. Package (create release candidate, e.g., Docker image, executable, gem, …)
4. Run smoke tests
5. Deploy to staging (promote)
6. Run end-to-end tests
7. Deploy to production (promote)

Plenty of variation is possible. For example:

• Some steps could be manual (e.g. deploy to production). Ideally though, they are automated.
• The compile stage could be missing (it’s not relevant for Ruby for example).
• After a production deployment, it might make sense to run another smoke test.
• There could be manual tests
• There could be performance tests
• There could be compliance tests, security audits, etc.

Note that in the list of stages above, the odd ones are building/deploying, while the even ones are testing. Move to the next step, verify that it worked, move to the next step, verify that it worked, … etc.