A live-sports streaming platform reports 99.95% availability every month, yet app reviews and support tickets during big matches are dominated by 'takes forever to start', 'keeps buffering' and 'push notification showed the wicket before I saw it'. Ten teams own pieces of the path from venue to player — encoding, packaging, origin, CDN, playback API, DRM and the client apps — and nobody owns the end-to-end experience.
The interviewer asks“Our dashboards are green on availability but viewers complain about slow starts and buffering. How would you build a culture where latency SLOs are defined, owned and enforced like uptime? Tell me about a time you did something like this.”
I treat latency as availability that users feel, so I define it from the player, not the server: video start time, rebuffer ratio and glass-to-glass delay, each as a good-events-over-total-events SLI with an error budget. Then I make it ownable by splitting the end-to-end budget into per-hop budgets that each team controls, gate releases on those SLIs through canaries, and alert on burn rate rather than thresholds. Culture comes from rituals and consequences: a weekly QoE review, a signed error-budget policy, and the numbers sitting next to revenue in the business review. When I did this before, start time p90 went from 3.8 to 1.9 seconds in two quarters, and the key was running it report-only first so teams trusted the measurement before it had teeth.
A free account unlocks the rest of it, and saves your progress across every track.