“Our system is 99.9% up” sounds good, but unless which operation counts as successful, what the latency is, and what the user experiences are stated, this number is not enough to decide on.
The SLA, SLO and SLI difference
SLI is the signal you measure: successful request rate, latency or error percentage. SLO is your target. SLA is the commitment given to a customer or partner. Mixing these three leads to reading a technical goal like a commercial contract.
Start from the user journey
Pick critical flows like login, payment, order creation or report fetching. For each flow define success, acceptable latency and the measurement source. Counting all endpoints equally important leaves important signals inside noise.
What is an error budget for?
The small accepted error margin for reaching the goal opens a shared decision space between the team and product development. If the budget runs out, reliability work can take priority over new features.
A good metric changes the discussion
When you start asking “can the user complete their task?” instead of “is the server up?”, monitoring, operations and product decisions move toward the same language.