You can hack together a solution. You can overengineer something and never finish. How do you balance the two extremes? What principles help you find the best simple system for now?
The purpose of testing is to increase confidence for stakeholders through evidence.