How the early warning is produced
The system runs on a bundled operational dataset covering the full network. Detection is deterministic: no black-box model sits between the operational record and the exposure number an executive is asked to act on.
Detection pipeline
Five deterministic stages from raw event to owned action
Windowed comparison
For each metric and each entity, a 21-day rolling window is compared with the preceding 60-day baseline. A deviation is only recorded when it clears the metric's noise floor, so ordinary variance is not reported as risk.
Trend projection
A least-squares slope over the recent window is extrapolated 30, 60 and 90 days forward. The projection is deterministic and reproducible — no model weights, no randomness, and the same input always yields the same forecast.
Financial translation
Each metric carries its own unit economics: fuel gallons and price, detention rate per hour, loaded driver hour, expedite premium, contractual penalty rate, or margin per load. The projected gap is multiplied through that formula to produce exposure in dollars.
Causal attribution
Candidate causes are tested with lagged correlation at the same site. The strongest leading indicator becomes the root cause, and the ordered sequence becomes a causal chain across modules.
Routing and action
The responsibility matrix maps every metric to exactly one accountable role, which receives the slippage together with the single intervention that addresses the identified cause.
Source tables
Bundled CSV extracts served with the application
| Table | Rows | Size | Contents |
|---|---|---|---|
| dim_locations.csv | 9 | 1 KB | Terminals, distribution centers and warehouses in the network. |
| dim_customers.csv | 80 | 7 KB | Customer master with SLA targets, penalty terms and contracted revenue. |
| dim_trucks.csv | 220 | 13 KB | Tractor and trailer master with domicile and preventive-maintenance interval. |
| dim_drivers.csv | 240 | 13 KB | Driver master with domicile, hire date and employment type. |
| dim_lanes.csv | 420 | 19 KB | Origin/destination lanes with distance, target rate and mode. |
| dim_owners.csv | 12 | 1 KB | Accountable owners and the responsibility matrix used to route slippages. |
| fact_loads.csv | 84,801 | 14.6 MB | Load-level shipments with revenue, cost components, miles, transit and OTIF outcome. |
| fact_dock_events.csv | 111,971 | 11.6 MB | Inbound and outbound dock events with appointment, wait, load and detention minutes. |
| fact_fuel.csv | 87,922 | 4.6 MB | Daily per-truck miles, gallons, MPG, idle hours and fuel cost. |
| fact_maintenance.csv | 3,515 | 249 KB | Preventive and unplanned work orders with downtime and repair cost. |
| fact_driver_hours.csv | 92,938 | 4.6 MB | Daily driver drive, on-duty, idle, detention and overtime hours. |
| fact_warehouse_ops.csv | 3,288 | 328 KB | Daily warehouse operating metrics per distribution centre and warehouse. |
| fact_sla_events.csv | 6,841 | 493 KB | Service exceptions, SLA breaches, claims and penalties by customer. |
| fact_cost_ledger.csv | 6,318 | 297 KB | Daily cost leakage postings by category with preventable share. |
Metric coverage
Monitored metrics per surveillance module
Trucking & Fleet
- Fleet Utilizationpct
- Empty Mile %pct
- Fleet MPGmpg
- Idle Time Sharepct
- PM Compliancepct
- Unplanned Downtimehrs
- Driver Overtime Sharepct
- Driver Detention Hourshrs
- Lane Margin %pct
Warehouse & Distribution
- Inventory Accuracypct
- Order Backlogcount
- Pick Cycle Timemin
- Pick Accuracypct
- Pack Cycle Timemin
- Order Cycle Timehrs
- Units / Labor Hourratio
- Warehouse Overtime Sharepct
- Space Utilizationpct
- Damage Ratepct
- Fulfillment Ratepct
Dock & Loading
- Average Dock Dwellmin
- Wait Before Dockmin
- Load Timemin
- Unload Timemin
- Appointment Compliancepct
- Dock Door Utilizationpct
- Detention Minutes / Truckmin
- Load Ready on Arrivalpct
- Units / Dock Labor Hourratio
Delivery & Customer SLA
- OTIFpct
- On-Time Deliverypct
- Average Delivery Delaymin
- Failed Delivery Ratepct
- POD Turnaroundhrs
- Claims Valueusd
- Customer Complaintscount
Cost & Margin Leakage
Interpreting exposure
What the numbers do and do not claim
Exposure is conditional
Every dollar figure answers one question: what does this cost if the current trend continues unchecked for the stated horizon? It is not a booked loss and not a forecast of the P&L.
Avoidable share is explicit
Each slippage carries a preventable percentage reflecting how much of the exposure responds to operational intervention rather than market or structural conditions such as fuel price.
Severity is relative
Severity is assigned by rank across the current board, so the list stays triaged. A watch item is still real exposure — it is simply behind twelve larger ones.