Three principles I decide by.
1. Control over the essentials
Own and understand the pieces that hold the operation up. If I do not understand a critical dependency, I do not control it.
Relying on what is not understood.
2. Evidence before scale
First something works at small scale and can be measured. Then, and only then, it grows.
Scaling a hunch.
3. Quality that survives operations
A design is judged in daily use: with errors, with load and with real people. That is where you learn whether it was done well.
Designing for the demo.