Using the machine-learning-driven alerts in Capacity Planner has helped us plan and catch errors before they become service-affecting. We've been able to stand up to the extra demand. We've seen ROI in terms of the number of outages that we've prevented proactively. The other ROI is obviously in terms of right-sizing in the estate. We've been able to run more VMs on our estate by right-sizing the capacity on those hosts.