2 min lesson
Tie SLAs to severity, not to a flat clock
Use the lesson to respond to this: "A hiring manager asks, 'How would you know whether Cursor's quality is actually improving over time?' Sketch your answer." Keep the answer plain.
Step 1 of 2
Tie SLAs to severity, not to a flat clockthe worst issues get the fastest clocks
A flat “respond within 24h” SLA is malpractice for a quality program - it gives a cosmetic typo the same clock as a data-loss bug. Anchor each SLA to the severity tier so urgency and the rubric stay aligned.
- Severity
- S0 - Critical
- First response
- Minutes - page on call
- Resolution / mitigation target
- Mitigate same day; rollback if it's a regression
- Severity
- S1 - Major
- First response
- Within hours
- Resolution / mitigation target
- Fix or solid workaround within days
- Severity
- S2 - Degraded
- First response
- Within a day
- Resolution / mitigation target
- Scheduled into the backlog by priority score
- Severity
- S3 - Cosmetic
- First response
- Acknowledged
- Resolution / mitigation target
- Batched; no individual clock
| Severity | First response | Resolution / mitigation target |
|---|---|---|
| S0 - Critical | Minutes - page on call | Mitigate same day; rollback if it's a regression |
| S1 - Major | Within hours | Fix or solid workaround within days |
| S2 - Degraded | Within a day | Scheduled into the backlog by priority score |
| S3 - Cosmetic | Acknowledged | Batched; no individual clock |
SLAs inherit the severity tier so the rubric drives the calendar.
Report trends, not snapshots. “We have 312 open bugs” is noise - leadership can't act on a single number. “Top theme is down 40% over three weeks, but regressions per release ticked up” is a decision. A snapshot flatters or panics; a trend tells you whether the program is working.
When asked what you'd measure, split it cleanly: operational metrics for the response machine, quality metrics for the product, SLAs tied to severity and everything reported as a trend. Then add the leading-indicator point - “a theme spike warns me of a wave before resolution time slips.” That structure signals you've run a program, not just answered tickets.
“I'd track two layers. Operational - first-response, resolution, escalation, reopen - to know the machine works. Quality - top themes by volume and ARR, regressions per release, repeat-contact - to know the product's improving. SLAs hang off severity tiers and I report trends so we can see whether quality is actually moving, not just a snapshot.”