Woods Cross
LIVE · SAT 11:42 AM · UPDATED 0:08 AGO
Forecourt · demo background

Production surface — everything below is what the live dashboard would show. This fold is for the presenter.

Talk track. Pump 6 blocked and Checkout 1 unmanned at the same moment — one trip outside killed the counter. That is the one-system story: a long line inside holds cars on the pumps, and the overlay chart shows the peaks moving together. Caution from the RB's validation morning: the two peaks did NOT move together there — pitch the overlay as a diagnostic ("do yours move together?"), not a promised correlation. Store 8 same hour (0 blocked, 2 min waiting) is the store-to-store comparison. MPD down is their named pain. We do not read dispenser diagnostics. We flag a pump when its traffic drops against its own baseline, whatever the cause. The bagged timer is their exec-visit story: a bag that outlives the repair; the persistence alert exists for that. Forecourt-full minutes prove bottlenecks even where no camera covers the approach lane.

Claims. Continuous presence tracking + sampled AI review. Numbers are illustrative, on real mechanism. Waiting is presence-measured, not a POS timestamp. Timers tick only on exceptions. Pump health is utilization against the pump’s own baseline — we do not diagnose dispensers. Bagged is a sampled photo check on a slow cadence; the duration is time since the bag was first confirmed. Excess pump-minutes count only dwell past a normal fill. Waiting-car counts depend on approach-lane camera coverage; forecourt-full minutes do not.

Build readiness (validated on RB's footage 2026-08-30; see support/tasks/completed/forecourt-video-validation.md). Per-pump occupancy, dwell timers, excess pump-minutes, forecourt-full minutes, long-dwell flags, and the pump/queue overlay are measured on real footage. Blocked-vs-fueling classification works via the pump_activity VLM type (golden cases pass on both models) but is not production-wired yet. The waiting-car chip needs a per-site pump-queue zone. Pump health needs weeks of baseline plus coverage awareness. Bagged-pump duration is not built. This page carries only what has a lever now; the per-pump stats table and demand-shape analytics live on PathCI Forecourt (forecourt-week.html).

4 need attention.

Pump 6 blocked · Checkout 1 waiting · Pump 4 health · Pump 8 bagged
Pump 6
Blocked
Dwell past a normal fill. One car waiting behind.
→ Ask them to re-park
8m 14s
excess
Checkout 1
Customer waiting
Customer at the counter, no employee.
→ Get to the counter
1m 42s
waiting
Pump 4
Health
Traffic at 31% of this pump’s own baseline — cars avoiding it. Flagged 6 of the last 30 days.
→ Walk Pump 4 — check the MPD card reader and screen
47m
since last fill
Pump 8
Bagged
Out of service since Tuesday. No fill attempts seen — the bag is still on.
→ Test Pump 8 — pull the bag if it passes
3d 4h
bagged
Store 8 · same hour · 0 blocked pumps · 2 min waiting
Pumps now
Island A · street
Pump 1
Occupied 3m
Pump 2
Occupied 5m
Pump 3
Occupied 2m
Pump 4
Empty
6 / 30 days
drive aisle
Island B · store
Pump 5
Occupied 4m
Pump 6
Blocked 8m 14s
1 waiting
Pump 7
Occupied 6m
Pump 8
Bagged 3d 4h
test + pull the bag
Today at peak
Excess pump-minutes i
47min
Saturday 11–1 · only dwell past a normal fill counts
Patterns (per-pump table, demand shape) are PathCI Forecourt →
Forecourt full i
12min
Every pump occupied at once · longest run 7 min at 12:10
Entrance cam 2 cars waiting now
Waiting-car count needs approach-lane coverage.
Waiting with no employee
14min
Customer-present time today · CX lens
serving 66 min waiting 14 min
Checkout 1 1m 42s customer waiting
One store
Queue inside vs blocked pumps outside
excess pump-min queue length