The last two lessons have hopefully given you the tools to figure out which of your monitoring metrics might be useful as SLIs and what types of SLIs you might want to start measuring for your services. It might be tempting to rush off and get started writing SLI specifications for your services, but we've got one more recommendation for you first. You should aim to only have one to three SLIs covering each user journey, even if your system and user journeys are relatively complex. As the complexity of your system increases, this can become harder and harder to stick to. So, in this lesson, we're going to present a couple of techniques for managing complex user journeys that can help you out. I'm sure you're thinking, why do we recommend so few SLIs? Surely, if your system is complex, it's okay to have a lot of SLIs so it can accurately quantify its reliability. Well, it's important to remember that not all of your monitoring metrics make useful SLIs, even if they have reasonable correlations to outages. The more SLIs you have, the more context your operations folks have to learn and retain about the services they are responsible for, to understand the signals that are driving their operational response. What's worse, having a large number of SLIs increases the probability of these SLIs giving conflicting signals about the state of the system, which will drive up your time to resolution as the incident responder chases down red herrings. It's important to stress here that we're not recommending you ditch all of your other monitoring and observability systems once you've started measuring SLIs for your services. The way we think about it is that a deterioration in your SLIs is an indication that something is wrong. But once the deterioration has become bad enough to provoke some form of incident response, you'll need those other systems to help you figure out what that is. Even if you manage to stick to the recommended limits for each of your user journeys, you can still end up with lots of SLIs. If you have lots of journeys, you shouldn't be afraid to ruthlessly prioritize only those journeys whose reliability has significant impact on your users or your business. It's not a problem to decide that a journey simply isn't important enough to cover with SLIs and SLOs if you have the data to back that decision up. In the next video, we'll talk about another way of coping with lots of user journeys, by aggregating similar ones together to give a higher-level view of a class of interactions.