Count the release responsibilities, not just the automation jobs.

How repeatable is the current build?

A documented build with a reviewed lockfile has a different starting point from one that works only on an old laptop. Missing dependencies and unidentified live code may need a bounded assessment first.

How many environments and services are involved?

Include the web process, workers, schedules, database and integrations relevant to the application. Each can need configuration checks and a place to rehearse.

What changes the data?

Database migrations and imports can make recovery more involved than returning to older code. Include compatibility checks, rehearsals and the limits after new records arrive.

Which behavior needs proving?

Choose representative tasks and nearby failures. A small number of meaningful end-to-end checks may explain the release better than many jobs that only repeat installation steps.

Who will operate the result?

Include usable instructions, access ownership, training where agreed and support boundaries. A more complex platform can add work after launch, so compare its operating needs with a simpler improvement.

Compare provider proposals or discuss a bounded first scope.