FreightLens — Daily Freight Tracker
Real-time visibility across 100+ FCs, eliminating manual spreadsheets and saving 500+ hours per month
Problem Statement
Amazon's inbound freight scheduling was managed through disconnected spreadsheets shared across 100+ fulfillment centers. Each FC independently tracked vendor appointments, standing appointment allocations, and capacity metrics — making it impossible for regional managers to identify scheduling conflicts, over-commitments, or capacity gaps until they manifested as operational disruptions on the dock floor.
FreightLens consolidated all FC scheduling data into a single rolling 21-day matrix view. The system surfaces over-scheduled dates (where committed volume exceeds planned capacity) in real time, provides standing appointment breakdowns by vendor type, and exposes per-FC operational metrics — enabling proactive load balancing and capacity planning across the network.
System Architecture
FreightLens ingests daily scheduling data from FC-level systems through an EventBridge-orchestrated pipeline. Each FC publishes its appointment and capacity data to S3, which triggers a Lambda consolidation function that normalizes, validates, and writes to DynamoDB.
The React frontend queries a REST API (API Gateway + Lambda) to retrieve the rolling 21-day matrix, standing appointment breakdowns, and FC metrics. The Admin Portal writes updates back through the same API layer, with optimistic locking on DynamoDB items to prevent conflicting edits.
CloudWatch dashboards monitor data freshness (each FC must report within its SLA window), Lambda error rates, and API latency. Alarms trigger SNS notifications to the on-call team when an FC misses its reporting window.
Data Model
Rolling 21-Day Capacity
| Column | Type | Notes |
|---|---|---|
fcId | VARCHAR | PK (composite) — FC identifier |
date | DATE | PK (composite) — calendar date |
bmPortal | INT | BM Portal planned capacity (units) |
vendorScheduled | INT | Vendor-confirmed appointment units |
saBlocked | INT | Standing Appointment blocked capacity |
saScheduled | INT | Standing Appointment confirmed units |
saUnitsLeft | INT | 0 = over-scheduled (RED flag) |
totalUnitsLeft | INT | Net remaining capacity |
utilizationPct | DECIMAL | Over 100% = over-capacity |
Standing Appointments
| Column | Type | Notes |
|---|---|---|
fcId | VARCHAR | FK → FC |
vendorType | ENUM | PCP or PSP |
breakdownType | ENUM | Blocked or Scheduled |
date | DATE | Calendar date |
units | INT | Appointment unit count |
FC Operational Metrics
| Column | Type | Notes |
|---|---|---|
fcId | VARCHAR | PK (composite) |
date | DATE | PK (composite) |
plannedCapacity | INT | Target daily capacity |
totalScheduledQty | INT | All confirmed appointments |
ncnsPct | DECIMAL | No Call No Show rate |
vendorReceipts | INT | Actual vendor deliveries |
endBacklog | INT | Outstanding unprocessed volume |
hotPos | INT | Urgent purchase orders count |
dsAllocationPct | DECIMAL | Day-shift allocation |
nsAllocationPct | DECIMAL | Night-shift allocation |
Capacity Utilization Logic
Utilization % = ((vendorScheduled + saScheduled) / bmPortal) x 100
Healthy (0-85%)
AMBERWarning (85-100%)
AMBEROver-scheduled (>100%)
REDLLM Enhancements
Scheduling Risk Analyzer ✦
Analyzes the full Rolling 21 matrix to identify the top 3 over-scheduling risks across selected FCs. Returns FC name, date range, severity (HIGH/MEDIUM), and a recommended operational action for each risk.
Represents how I would build this system today with LLM capabilitiesNatural Language Metric Query ✦
Plain English questions like "Which FCs are over capacity next week?" or "Show me FCs with NCNS above 15%". The LLM interprets the query against FC metric data and highlights matching fulfillment centers.
Represents how I would build this system today with LLM capabilitiesCapacity Forecast Summary ✦
Auto-loads on the Rolling 21 Days page. Evaluates overall network health (HEALTHY / AT_RISK / CRITICAL) based on over-scheduling frequency, identifies the top concern, and recommends a specific operational action.
Represents how I would build this system today with LLM capabilitiesKey Metrics
100+ FCs
Visibility Scope
500 hrs/mo
Manual Work Saved
21 Days
Rolling Window
8 FCs
Demo Deployment