I worked at a software company where we should have run an experiment like this. There's an upgrade on user accounts that we charged a one-time $10 for, and people buy it all the time. We decided to try it as a subscription.

We spent a month getting the subscription code ready, then deployed it to some % of users, only to discover that nobody wanted it. What we should have done was build the front-end only, get people to buy subscriptions, and then just grant them the one-time permanent version.