 With Power BI, which is a service that I'm responsible for, we deliver it as a service. There are really four questions that someone has to ask when they're operating software as a service, when they're operating a service. Question number one, if users were to come show up for my service right now, is it ready for them to arrive? Monitoring whether the service is alive. The second question is, for the users who have actually shown up and are using the service, are they having a good experience? Is the performance what I expected to be? The third question is, if the answer to number one or two is no, why? What's wrong and how do I fix it as quickly as I can? Then the last question is, great, everything is working really, really well. What are customers doing? What are they using the service to accomplish? Where can we go invest in order to give them even more value? We have tools for each of those questions, and ultimately bringing the right tools to bear that are optimized for taking the right data and answering those questions is how we deal with the problem. I've got a pretty large team. We've got 450 developers across three global development centers, and so we're working on a lot. Only part of that team is actually working on the service. So let's just focus on them for a moment. So that service team that's building the software and actually operating the service, operates on a very fast cadence. So we will always have in process four trains, and there's a train that is the software system that we are delivering the service to our customers with. There's a train that we use inside of Microsoft only. So there's a version of Power BI that only Microsoft employees use. So they're testing it before that train actually rolls into the production station. There's a version of Power BI that we use only within my team, and then there's a test cluster that only the engineers on my team use, and that rolls out every single day. So we've got this continuous process, continuous delivery of software into this multi-train model. So we operate on a backlog where we've got a long list of things that we know we want to accomplish. We staff those backlog items with small teams, we call them squads. They've got a very clear mission. They know exactly what the customer value is they're trying to create and deliver into the service. That work gets done, it gets checked into the system, and eventually deployed in that first train, and eventually that rolls through the other stations and ends up in public. So on a monthly basis, work that was done by the engineering squad that was creating value, it will roll out to our end customers. So it's a staged approach and it's a continuous approach.