I have not implemented it, as I am a Splunker (all feeling expressed are my own from my experience).
However, I have worked with many clients on deciding to opt-in or not. I would summarize by saying that the vast majority of customers that I work with have seen very good pay off in making the shift. By that, I mean that they now have access to a greater amount of data on which they can make better decisions. This is allowing them to add new use cases and business outcomes at a faster rate.
The biggest downside is that it is forcing them to deal with a bit more ambiguity, learn a new model, and pay attention to resource utilization. My perspective on that last one is that this is something that we should all be doing all along, but the point is the same.
I will also add a few things which I think are helpful in making the transition: (1) truly make the transition from "data volume" to "resources needed to conduct workloads". The nuance here is that I see clients often start to use techniques like "GB / SVC ratio" too strictly. I personally think is a good quick and dirty way to estimate SVC needs, but if taken too strictly, you will quickly get to the point where you are really just talking about GB / day again. The point here is that there is a lot more flexibility in workload pricing and it's important to recognize that and start thinking differently; (2) while watching SVC consumption is important it is not the only thing that you should be watching. Here I am saying that I want clients to watch "health" in addition to, but not by virtue of SVC consumption. Look at things like data volume in, search count out, search health statistics, etc. in addition to SVC consumption; (3) make sure you start watching the volume of data at rest - hopefully you will be wildly successful and bring in a lot more data, etc. Just make sure that you are watching how that impacts your storage.
I'm eager to hear how things progress. Feel free to chime in with more questions.