I tend to evaluate how well engineered a system is based on these pillars.

* Reliability - Includes HA, fault-tolerance on single node, reconciliation or reliable rollback of failed transactions, ability to manually intervene if something is wrong, and observability

* Security - privilege separation and defense in depth

* Performance - Are basic operations fast? Are there worst case performance pathologies? Are batch interfaces available for efficient processing? How many concurrent users can you handle (if server based).

If we consider all these and perfectly implement, it's already pretty rigorous as an engineering discipline. I think companies at Google scale can do that, on projects with budget. But everywhere else, we make lots and lots of compromises. So there's it.

For example, building high performance network systems or DB engines is more "engineering" than building line-of-business application. Because more of these "engineering" concerns are the part of specification.

I am intentionally leaving out UX, because its a product design problem.