Recommendation types are separated: sequence improvement is active,
road-choice scoring uses our cascade plus available live signals
(bus AVL, weather, closures, zone congestion). Cascade-validated.
Compare against:
Point-by-point local resequence
Each card compares the stop's old two-leg context with its suggested
new context. The optimizer keeps moves only when the full cascade
re-run confirms an improvement.
Customer time-window suggestions
These stops are still projected late after optimization. The TW is
binding — the optimizer can't make the close on its own. Operator
action required: extend the customer's window, reschedule earlier in
the day, or split the delivery to another route.
Routific plan vs our smart-cascade ETA
hover a row to highlight the leg · click to zoom · times in EDT (local)
#
id
Routific arr
Our P50
Our P95
pkg
svc
TW (EDT)
click "Load CSV" in the right panel to load the Routific result
Routific stop
⚠N% = our cascade's prob_late
our P50/P95 includes drive + service + curb + building