Phase 5 — how should a churny shift be drawn?
Same real prod day (2026‑06‑04, ), rendered three ways on normalized spans. Pick the rendering philosophy; layout / drill‑down / live all follow from it. Times in America/Chicago — business timezone is itself a design decision.
READY (idle / waiting)
IN_CALL (on a dial)
IN_ACW (after‑call work)
OFFLINE / gap
A True‑to‑scale + zoom
Every span proportional to real elapsed time. Honest. Dense flip‑stretches collapse into near‑solid bands — you’d zoom into a window to read individual transitions.
Good
- Never lies about time
- Solid block = genuinely sustained state
- Gaps / OFFLINE jump out
Bad
- Individual flips invisible without zoom
- Needs a zoom/pan interaction
B Utilization buckets (adherence strip)
Slice the shift into 10‑min buckets; each shows the share of time per state (gaps stay empty). Reads like a call‑center occupancy strip. Whole shift legible at a glance; trades away transition detail.
Good
- Whole shift legible, zero zoom
- “How occupied, when” instantly
- Comparable agent‑to‑agent
Bad
- Hides exact transition sequence
- “When was call X” needs drill‑down
C Min‑width segments (every flip visible)
Draw every span but force a min width so 2‑second flips stay clickable. Time axis goes non‑linear (distorts duration). Surfaces transition count/sequence.
Good
- Every transition visible + hoverable
- Churn is obvious
Bad
- Lies about duration
- No real time axis
My lean: B as the default overview (legible whole‑shift, comparable across agents), with A as a “zoom to raw” drill for exact transitions. C is a niche debug view — probably YAGNI.
Reply in the terminal: which is the primary view — A, B, C, or a combo? And does “zoom to raw” as a drill feel necessary or like scope creep?