Best thing to do is test all reasonable trial expiration scenarious, including those tied to user behavior instead of time. You never know what will be optimal,

Best thing to do is test all reasonable trial expiration scenarious, including those tied to user behavior instead of time. You never know what will be optimal, you need to test because each product and audience is unique. I know of a number of software publishers who had near zero conversion with a 30 or 15 day trial but made a lot of money when they expired trials based on other factors such as usage.