Meridian Field Services: instructional example, not project evidence
Every row is a scenario (stimulus, environment, response, numeric measure) with a fit criterion naming the method that proves it. "Fast" is rejected at authoring.
Omitted categories, noted once: portability and compatibility carry no requirements (single-platform pilot, no co-existence constraint); maintainability carries one row (NFR-P3).
| Id | Scenario (stimulus → response) | Measure | Fit criterion | Driver | Stage | Architecture |
|---|---|---|---|---|---|---|
| NFR-P1 | A job reaches ready during peak dispatch (pilot load: Toronto, 1,500 jobs/day) → the job is visible on the board | ≤ 5 s p95 | The board-liveness run on the perf harness (PS-1[1]) | The dispatcher decision cadence the value case rests on (objectives[2]) | pilot | committed (board liveness[3]) |
| NFR-P2 | A dispatcher opens a job's suggestions → ranked candidates render | ≤ 1 s p95 | PS-1's suggestion series | Suggestion-first only works if it is faster than memory | pilot | committed, with the 2-s degradation boundary (FLW-01[4]) |
| NFR-P3 | A sixth hub onboards → configuration only, no code change (maintainability's one row) | 0 deployments required | The hub-six onboarding drill | Rollout economics (BD-2[5]) | scale | GAP: no committed scenario yet; open question §6 |
A percentage without its window is theater. Degradation rows cite the designed fallback; invented behavior appears nowhere.
| Id | Scenario | Measure (windowed) | Fit criterion | Cites |
|---|---|---|---|---|
| NFR-A1 | The board is available during operations | 99.5% over dispatch hours (06:00 to 22:00 local); a continuous number would buy nothing the business needs | The platform uptime series | committed |
| NFR-A2 | The position feed stops → positions mark stale and manual assignment proceeds | 0 blocked assignments through a 30-min gap | The feed-loss drill (DV-5[6]) | feed-loss resilience[3] · suggestions_unavailable[7] |
| NFR-A3 | A restore is required | Assignment records: zero loss (recovery point) · 4 h (recovery time) | The restore drill (checks[6]) | Contract clause 14 driver (BR-06[8]) |
Pilot and scale are distinct rows; one number retrofitted across stages is the anti-pattern these rows exist to prevent.
| Id | Stage | Frame | Promotion trigger | Architecture |
|---|---|---|---|---|
| NFR-S1 | pilot | 400 positions/min · 1,500 jobs/day | none (active) | committed (scale frame[3]) |
| NFR-S2 | scale (future) | 5 hubs · 2,000 positions/min, same latency rows | The hub-two go decision: an event, not a date (BD-2[5]) | Headroom: measured breakpoint 2,150/min gives 7.5% over the hypothesis; re-run at production sizing before promotion (PS-3 results; open questions[1]) |
Pointer rows only: the security requirements artifact owns every word; a row restating an SR is a defect.
| Quality | Points at |
|---|---|
| Confidentiality: technician location | SR-033, SR-034[9] |
| Integrity: position feed | SR-021[9] |
| Accountability: overrides | SR-009, SR-031[9] |
| Authenticity: staff sessions | SR-001[9] + the assurance table (assurance levels[10]) |
Non-repudiation carries no requirement beyond SR-009's audit row; omitted with note.
"Intuitive" is not a requirement. Conflicts are decided, not discovered in code.
| Id | Scenario | Measure | Fit criterion |
|---|---|---|---|
| NFR-U1 | A first-shift dispatcher (Dana[11]) completes an assignment through suggestions without help | ≤ 90 s by her third shift | Observed pilot sessions, weeks one and two (observations[12]) |
| NFR-U2 | Override entry during assignment | ≤ 1 added step | Click count on the board (SCR-001[13]) |
| NFR-U3 | The board serves assistive technology | WCAG 2.1 AA | The accessibility audit at pilot exit |
BR-03[8] requires a reason; usability wants zero friction. The reason field wins: accountability outranks one click. Decided by M. Chen with N. Duval, logged (decision index[5]); the row carries the decision chip. Silent resolution in code is the anti-pattern this callout prevents.
| Question | Owner | Answer by | Blocks |
|---|---|---|---|
| NFR-P3's hub-onboarding scenario has no committed architecture answer. Does hub onboarding become a quality scenario at the next architecture regeneration? | E. Sandoval | Sep 30, 2026 | The maintainability row's reconciliation only |