Service Level Objectives and Customers in the Fog
What to do when there is no end customer
Defining service level objectives (SLOs) for customers is a difficult process. It takes effort, time, and dedication to define your end customer, debate what it is they need, and try to understand how you might better meet their needs. How does a customer feel when they visit your website or interact with your service? These questions aren't easily answered, so you can imagine how tricky it may feel defining SLOs when you're not even sure who or what your customer is.
A conversation about SLOs typically beings with a question: Who are you serving? But, for infrastructure and shared service teams, this question often leads to responses like, "We don't have end customers," or, "Our customers are everyone." The discussion can quickly spiral into a frustrating exercise in definition, making SLOs feel irrelevant and too ambiguous to pursue.
Yes, SLOs are all about the end customer... but they're not, really. They're more nuanced than that. They provide more than that. They're about setting a shared understanding of service quality. A team doing SLOs well is benefiting their customer—whoever or whatever that may be—and themselves. By intentionally choosing metrics that reflect key functionality or user experience, and deciding to iterate on them, teams gain a huge advantage. Engineers, product leads, and anyone else involved in the service align around what's been deemed important, and that alignment on its own, regardless of anything else, has its own power.
As an engineer on a shared infrastructure team, I saw firsthand what SLOs can do to an on-call crew. Our alerts became more focused, discussions between engineers and managers improved, and it became clearer what we needed to work on and when. These are the common benefits we hear about when talking about SLOs. We experienced all of that, and on a piece of shared infrastructure that was so broadly used at the company our team could not meaningfully define a single set of users with distinct needs. Instead, we set SLOs around core service goals, iterated on them, and saw improvement. Looking back now years later, I'm not even certain they were the right metrics, but they didn't need to be. Nothing ever is, we shouldn't strive for that, and you don't need a precise customer definition or the perfect metric to see benefits.
When your customer is in the fog, it can be easy hear all of this talk of end user experience and customer buy in that you need in order to define service level objectives and decide it isn't worth it. But I'd urge you to push onward, go a level deeper on what/who your customer might be. If your team can't quite see how SLOs might be worth the effort or you can't get a grasp on who that end customer is, start with just one metric that you think reflects some facet of your service reliability. Track it, set a goal around it, discuss it as a team. Whether or not you hit the bullseye, you'll likely see improvments for your team and, by extension, for your customer.