Course contentsShow
← System Design
Lesson 1284 of 1,91932. Monitoring and AlertingFree lesson

Iterating on SLIs and SLOs

Understand that SLIs and SLOs evolve with product maturity, user needs, and operational learnings over time.

Iterating on SLIs and SLOs

What you'll learn: How to evolve your SLIs and SLOs as your product, users, and operational understanding mature over time.

SLIs and SLOs Are Living Documents

Your first attempt at defining Service Level Indicators and Objectives is rarely perfect. What you initially thought users cared about most may differ from reality. A 99.9% availability target might prove too generous—or impossibly strict. As your system grows, user expectations shift, and your operational experience deepens, your SLIs and SLOs must evolve accordingly.

Why Iteration Matters

Early stages: You might start with basic SLIs like overall HTTP success rate because you lack granular data or understanding of user journeys.

As you mature: You discover that "success rate" hides important nuances. Users care more about checkout page latency than blog post latency. Your SLIs should reflect this—perhaps splitting into separate SLIs per critical user path.

Operational learnings: After a few incidents, you realize your 99.95% SLO is routinely violated by a poorly-understood dependency. You might temporarily relax the SLO while fixing the underlying issue, then tighten it again once stable.

User feedback: Direct customer complaints or support tickets reveal what actually frustrates users. Maybe they tolerate slow page loads but can't accept failed transactions. Your SLOs should mirror real user pain, not abstract technical ideals.

The Iteration Cycle

Think of it like tuning a recipe: your first batch teaches you what's missing, the second is better, and eventually you perfect it. Regularly review your SLIs and SLOs—quarterly is common—asking: "Are we measuring what matters? Are our targets realistic yet ambitious? Has our system or user base changed?"

Key Takeaway: SLIs and SLOs are not set-and-forget metrics—they require continuous refinement based on product maturity, actual user behavior, incident learnings, and evolving business priorities.