Real-Time Decisioning Demo | Logistics Exception Recovery
Pipeline<15ms
1 / 9
1
Architecture
2
Exception Trigger
3
Ingest
4
Context
5
Feature Serving
6
Ranking
7
Business Impact
8
Outcome
9
Architecture Recap
Stage 1: The Architecture
Unified logistics context for exception recovery
Orders, inventory, telemetry, carrier status, and event streams feed Redis through RDI and Redis Feature Form. Redis RAM handles the hot exception path. Redis Flex holds route history, lane behavior, and recovery patterns so the operation can decide the best next move before the SLA breach becomes a customer issue.
Data Sources

TMS

Orders, route plans, SLA commitments

WMS

Inventory, wave status, pick/pack state

Telematics

Driver, vehicle, and trailer telemetry

Carrier Network

Capacity, tender, and partner response

Kafka / Event Bus

Scan events, weather, traffic, geofence alerts

Ingest Layer

Redis Data Integration (RDI)

Synchronizes shipment, stop, inventory, and customer commitment state from operational systems

Redis Feature Form

Serves ETA risk, exception severity, capacity, and recovery propensity features online

Unified Context Layer

Redis RAM

Hot shipment state, driver ETA, active exception counters

Redis Flex

Warm route history, lane patterns, vectors, and customer service history

Feature Store

Delay risk, recovery cost, SLA, and churn exposure features

Redis Context Retriever

Assembles the Shipment 360 — order state, carrier context, and exception history — and exposes it as structured MCP tools for the decision engine

Decision Engine

Exception Classifier

Late, damaged, inventory short, or missed handoff

NBA Ranker

Reroute, expedite, substitute, or proactively communicate

Policy Engine

SLA, margin guardrails, customer tier, and penalty logic

Vector Search

Match similar past disruptions and proven recovery plays

Output Surfaces

Dispatch Console

Reroute, tender, hold, or expedite instructions

Customer Portal

Revised ETA, substitution, and proactive alerts

Carrier App

Driver and partner instructions

Support Workbench

Case guidance and next best resolution

↺
Learn:  Scan events, customer responses, dispatch actions, and final outcomes feed Redis Streams or Kafka, then retrain recovery policy in Redis.
Decision Target
<15 ms
Surface
Dispatch + portal
Business Goal
Recovery + SLA protection
Stage 2: Exception Trigger
A high-priority shipment goes off-plan in the middle of the route
A refrigerated shipment for a top retailer is due by 6:00 PM. At 2:14 PM, the driver falls behind schedule after a traffic closure, a cross-dock transfer is at risk, and the last good ETA is no longer believable. The operation has minutes to decide whether to reroute, expedite, substitute inventory, or proactively communicate.
Live Exception Event
SR
Shipment #SR-11842
Cold-chain retail delivery | Dallas to Houston | 4-store multi-stop load | Tier-1 customer
AT-RISK
Eventlate_eta_risk_triggered
Current ETA drift+67 min vs committed window
Vehicle stateStopped near closure zone
Cargo conditionTemperature still in range
Next stop priorityHigh-penalty retail DC
Cross-dock backupNearby facility has substitute inventory
Customer exposure$145K order at risk
Why This Moment Matters
Logistics teams rarely fail because they cannot detect a delay. They fail because they cannot decide the best next recovery action before the network cost, SLA cost, and customer cost all grow at once.
This is a classic real-time decisioning problem: one disrupted shipment, many possible moves, each with a different service, margin, and downstream operational impact.
The opportunity: choose the best recovery path before dispatch falls back to manual calls, stale ETAs, and blanket expedite decisions.
Stage 3: Ingest
Shipment, inventory, and telemetry data flow into Redis
RDI synchronizes the order, stop, and inventory state from TMS and WMS. Telematics and scan events stream through Kafka. Redis Feature Form hydrates live ETA-risk and recovery features so the operation is deciding from one shared context layer, not multiple point tools.
Redis Data Integration (RDI)Redis Feature Form
Operational Systems → Redis
TMS
Transportation Management
Order, lane, stop sequence, carrier tender, SLA commitments
→
WMS
Warehouse Management
On-hand substitution inventory, cross-dock readiness, wave state
→
TEL
Telematics + Fleet
GPS, route deviation, dwell time, temperature, fuel, driver hours
→
CRM
Customer Service + SLA
Penalty tiers, customer priority, promised window, issue history
→
KFK
Kafka / Redis Streams
Scans, weather, traffic, geofence alerts, partner acknowledgements
→
Pipeline Status
TMS syncStreaming CDC
Telematics lag<2s
Inventory refreshContinuous
Feature parity100%
Decision dependenciesServed from Redis context layer
Stage 4: Context
The shipment 360 assembles around the disruption
Redis brings together the live disruption, the broader route history, the inventory options, and the customer impact model in the same response path. The decision is not just whether the shipment is late. It is what the best recovery move is right now.
Redis RAMRedis FlexRedis Context Retriever
Operational Context
Delay severityP92 late-arrival risk
Temperature stabilityStill compliant
Driver hours remaining2.1 hours
Alternative route+22 miles, recovers 31 min
Nearby cross-dock12 miles away, substitute stock ready
Carrier backupAvailable but expensive
Customer + Cost Context
Penalty windowBreach starts at 6:15 PM
Order value$145K
Store shelf impactHigh risk of out-of-stock
Customer tierStrategic national account
Historical recovery patternCross-dock swap outperforms blind expedite on this lane
Support case statusNo outreach sent yet
Context signal: Redis Context Retriever assembles the Shipment 360 — order state, carrier context, and exception history — so the decision engine has exactly what it needs. The winning recovery action becomes clear only when live disruption state and historical lane patterns appear in the same response path.
Stage 5: Feature Serving
Recovery features hydrate in milliseconds
Redis Feature Form serves ETA risk, lane volatility, inventory substitution fit, customer penalty exposure, and recovery-cost features from the Redis context tier. That means the ranking layer gets a full operational picture without fan-out to every system at exception time.
Redis Feature FormRedis Feature Store
eta_breach_probability
Probability the committed window will be breached without intervention
0.920.4 ms
substitution_readiness_score
Fit between nearby inventory, order composition, and transfer feasibility
0.870.5 ms
customer_penalty_exposure
Projected SLA cost if the committed window is missed
$18.4K0.3 ms
expedite_cost_ratio
Incremental cost of backup carrier or expedite relative to order margin
0.310.4 ms
lane_recovery_embedding
Vector representation of similar disruption patterns and outcomes
[0.51, -0.18, ...]0.6 ms
proactive_outreach_lift
Expected service-score lift if the customer is informed before the breach
0.730.3 ms
Stage 6: Ranking
Four recovery actions are scored in one decision path
Redis now has the live shipment state, the route alternatives, inventory options, and customer impact in one place. The decision engine can rank the next best action instead of forcing dispatch to choose between speed, service, and cost with partial information.
Redis SearchNBA RankerPolicy Engine
#1 Winner
RECOVERY PLAY
Cross-dock substitution + proactive customer ETA reset
Best SLA recovery with lower cost than blind expedite
Swap the final stop inventory at the nearby cross-dock, recover the most critical shelf impact, and notify the customer before the breach window.
NBA score0.95
#2 Fast Path
EXPEDITE
Backup carrier handoff
Recovers ETA but erodes margin
Viable when service protection matters more than cost, but inferior here because the nearby inventory option is faster and cheaper.
NBA score0.82
#3 Passive
COMMUNICATE ONLY
Keep route and notify customer
Lowest cost, highest SLA risk
Reduces surprise but does not solve the service issue. Good backup path, not the best operational move.
NBA score0.64
Stage 7: Business Impact
The value is not faster alerts. It is better recovery decisions.
Redis helps the operator choose the move that protects SLA, customer trust, and margin at the same time. That is a decisioning story, not a dashboard story.
Recovery Economics
Projected SLA penalty avoided$18.4K
Expedite cost avoided$6.2K
Shelf-out risk reducedHigh to low
Customer effortProactive, not reactive
Operational benefitDispatch standardizes best recovery play
Per-Exception Outcome
manual
Late calls, blanket expedite, stale ETA
→
ranked
Context-aware recovery action in one path
Stage 8: Outcome
Same disruption. Different operating model.
Without Redis, the dispatch team sees disconnected alerts and chooses the loudest response. With Redis, the platform assembles the operational context and recommends the best next move before the exception snowballs.
Reactive Operations
SR
Exception detected, manual triage required
Current status
Late
No ranked action available
Dispatcher workflow
1
Call driver
Check route manually
Delay
2
Open WMS
Look for substitution
Wait
3
Email customer
After breach is likely
Reactive
late
outreach
high
penalty risk
1
default response
Redis-Powered Exception Recovery
SR
Shipment recovery brief, ready now
Next best action
Cross-dock substitution
Notify customer before breach window
Why this wins
SLA
Best service recovery
Protects priority stop
$18.4K
COST
Cheaper than expedite
Backup carrier not required
$6.2K
CX
Proactive ETA reset
Before the customer calls
Now
Unified context assembled in Redis
Recovery confidence: 95%
Operational state, lane history, inventory readiness, and customer exposure all point to the same next move.
95%
confidence
<15ms
decision time
1
ranked play
Stage 9: Architecture Recap
One operational context layer for disruption recovery
TMS, WMS, telematics, carrier, and service systems stay in place. RDI and Redis Feature Form make them operational. Redis RAM and Redis Flex serve the live shipment context. The decisioning stack turns that into the best next recovery action before the SLA breach hardens.
Data Sources

TMS

Orders, route plans, SLA commitments

WMS

Inventory, wave status, pick/pack state

Telematics

Driver, vehicle, and trailer telemetry

Carrier Network

Capacity, tender, and partner response

Kafka / Event Bus

Scan events, weather, traffic, geofence alerts

Ingest Layer

Redis Data Integration (RDI)

Synchronizes shipment, stop, inventory, and customer commitment state from operational systems

Redis Feature Form

Serves ETA risk, exception severity, capacity, and recovery propensity features online

Unified Context Layer

Redis RAM

Hot shipment state, driver ETA, active exception counters

Redis Flex

Warm route history, lane patterns, vectors, and customer service history

Feature Store

Delay risk, recovery cost, SLA, and churn exposure features

Redis Context Retriever

Assembles the Shipment 360 — order state, carrier context, and exception history — and exposes it as structured MCP tools for the decision engine

Decision Engine

Exception Classifier

Late, damaged, inventory short, or missed handoff

NBA Ranker

Reroute, expedite, substitute, or proactively communicate

Policy Engine

SLA, margin guardrails, customer tier, and penalty logic

Vector Search

Match similar past disruptions and proven recovery plays

Output Surfaces

Dispatch Console

Reroute, tender, hold, or expedite instructions

Customer Portal

Revised ETA, substitution, and proactive alerts

Carrier App

Driver and partner instructions

Support Workbench

Case guidance and next best resolution

↺
Learn:  Scan events, customer responses, dispatch actions, and final outcomes feed Redis Streams or Kafka, then retrain recovery policy in Redis.
Decision Target
<15 ms
Surface
Dispatch + portal
Business Goal
Recovery + SLA protection
Press ← → or click to navigate