Everything I know about judging a service I learned between nine and ten on a weekday evening, and everything I got wrong I got wrong by testing at lunchtime.
Networks are not the same at different times of day. This is obvious when you say it and completely absent from how most people evaluate anything. A service tested at two in the afternoon is tested under the easiest possible conditions, and it will look excellent, and it will tell you nothing about the Wednesday you actually care about.
So the rule in this house is that nothing gets judged outside peak hours.
In that hour I start with the thing we watch most and leave it running for ten minutes without touching anything, because problems are periodic and thirty seconds catches none of them. Then I change channel fifteen or twenty times in a row, jumping around rather than going up and down, because the delay on a jump is the one you will live with. Then I open the guide and scroll it, because a guide that takes four seconds to load is a guide you will stop using.
Then I put something on and go and do something else for twenty minutes, and come back to see whether it is still going. Half my failures were failures of endurance rather than of starting.
That is about an hour, at the right hour, and it beats a week at the wrong one.
What a trial should give you, and how long you should expect, is set out here.
One hour, at nine, beats a week of afternoons.