Hey there. Welcome to module three of our course on Measuring Service Reliability, choosing a good SLI. Now that we've convinced you that you need SLOs for your services, and shown you how you can use them to align incentives across your organization, let's talk about how to specify meaningful SLIs that can drive these SLOs. In this module, we'll start off by taking a look at some characteristics of monitoring metrics that can make them useful as SLIs and contrast these against other metrics that are less useful. Because the choice of where to measure an SLI is a key variable, we'll cover the five main ways you can measure an SLI and compare the pros and cons. Most concerns about service level reliability fall into a few broad categories of SLI depending on the interactions a user has with that service. We'll discuss each of these categories of SLI in some detail, then give example specifications for each of these SLIs that follow our suggested formula, the SLI equation. Then, we'll show you how deciding where to measure your SLI, what events do include, and what makes an event good, turns an SLI specification into something you can implement. One of the major challenges of specifying SLIs in the real world is service complexity. Your service may be composed of tens of microservices and have hundreds of HTTP or RPC endpoints, but having hundreds of SLIs can lead to operational paralysis and important signals being lost in the sea of noise. We'll go into some strategies you can use to manage this complexity and avoid having hundreds of SLIs. Once you've got a good set of SLIs specified for your service, you'll need to consider what targets you want to set for those SLIs. The best way to arrive at a reasonable target is to have historical monitoring data tell you what's achievable, but some cases this data won't exist, or these targets will diverge from those your business aspires to. Iterating toward alignment of achievable and aspirational targets is the focus of our last lesson.