If you have a way to segment your free users then come up with an idea and test it on 10% of them. See what happens. If you made a bad choice and piss of 10% of

If you have a way to segment your free users then come up with an idea and test it on 10% of them. See what happens. If you made a bad choice and piss of 10% of them, then you can turn off the test and leave the other 90% of the free users blissfully unaware. If your idea works, then you can expand it to a larger test cell, or to all free users. View it as an A/B test, but don't just default to a 50/50 split.