Philippines staffing research ·
How the observation window changes an API error-budget result
Compare short and long windows while keeping traffic mix visible.
The same incident can dominate a one-hour result and barely move a monthly ratio. The denominator, traffic mix, and window answer different questions.
Google SRE material defines indicators, objectives, and error budgets. OpenTelemetry specifies HTTP metric attributes. Neither chooses an objective for a particular API.
Methodology: define eligible requests and success before querying telemetry. For one service revision, calculate requests and failures over hourly, daily, and 28-day windows. Break out routes, status classes, retries, regions, and low-volume groups. Record exclusions.
Scope and inference boundaries: the analysis describes captured requests. It can show sensitivity to window and traffic mix. It cannot measure unobserved failures, user satisfaction, compliance, or cause.
Limitations: sampling, missing spans, retries, bots, cancellations, and routing changes alter the ratio. Small groups are unstable. Service owners approve objectives, exclusions, alerts, and remediation.
Sources
Related Research
FAQ
What should happen first?
Begin with a bounded ticket, approved access, and a named reviewer.
Who approves production changes?
The company’s technical owner keeps final merge and release authority.