Redis Real-Time Decisioning Demo | Airline Loyalty Decisioning

Stage 1: The Architecture

Sub-12ms loyalty decisioning at the check-in and booking moment

Miles balance, tier status, seat inventory, partner earn history, and expiration windows must converge before the check-in surface renders. Redis assembles the Passenger 360 so the right loyalty offer reaches the right passenger at the right moment — not a generic upgrade banner for everyone.

Data Sources

Passenger Name Record (PNR)

Booking details, seat assignment, SSR codes, itinerary

Loyalty Management System

Miles balance, tier status, earn/burn history, qualification progress

Seat Inventory System

Real-time seat map, availability, fare class buckets, upgrade inventory

Partner Earn System

Credit card partner transactions, hotel, car rental pending miles

Revenue Management

Dynamic upgrade pricing, load factor, co-pay calculations, yield targets

Kafka / Event Stream

Check-in events, booking events, flight status changes, gate activity

Ingest Layer

Redis Data Integration (RDI)

Syncs PNR, loyalty state, and seat inventory into Redis — always-fresh context

Redis FeatureForm

Serves passenger loyalty features and upgrade propensity scores online

Unified Context Layer

Redis RAM

Hot check-in state, live seat inventory, active upgrade window

Redis Flex

Miles history, tier qualification trajectory, partner earn records

Redis FeatureForm

Feature store — miles balance, upgrade propensity, expiration urgency, tier velocity features

Redis Context Retriever

Assembles the Passenger 360 — miles balance, tier status, upgrade eligibility, expiring miles, and seat availability — and exposes it as structured MCP tools for the decision engine

Decision Engine

NBA Ranker

Ranks upgrade, co-pay, bonus, and tier offers by expected value and acceptance probability

Upgrade Eligibility Rules

Miles threshold, tier requirements, inventory constraints, blackout rules

Revenue Arbitration

Yield trade-off between upgrade revenue, seat fill, and loyalty engagement value

Redis Search

Peer segment matching and historical offer acceptance similarity

Output Actions

Miles Upgrade

Full upgrade offer using loyalty miles

Miles + Co-pay Upgrade

Partial miles plus cash co-pay option

Bonus Miles Offer

Earn bonus miles on this segment — no upgrade

Tier Nudge

Progress notification — miles to next tier status

Decision Target

<12ms

Primary Goal

Loyalty engagement + upgrade revenue

Redis Role

Passenger 360 in real time

Stage 2: Check-in Event

Marcus Webb checks in for AP-847 — six hours before departure, two business seats remain

Marcus is a Gold-tier member with 47,200 miles and 8,200 of them expiring in 23 days. Two business class seats are open on his flight. The upgrade window is active. Redis has under 12 milliseconds to assemble his Passenger 360 and surface the right offer before the check-in confirmation screen renders.

Check-in Event

Marcus Webb

Gold Status | 47,200 miles | AP-847 JFK → LAX

UPGRADE ELIGIBLE

Eventcheckin_initiated

FlightAP-847 JFK → LAX — 06:15 departure, 6h 02m

Current seatEconomy 24C — confirmed

Business class availability2 of 14 seats open

Upgrade windowActive — closes T-4 hours (2 hours remaining)

Miles balance47,200 — 22,200 above upgrade threshold

Expiring miles8,200 miles expire in 23 days

Why This Moment Matters

Marcus has enough miles to upgrade twice over. But he has never used them. He has 8,200 miles expiring in 23 days — a clear urgency signal. Two business class seats are open and the upgrade window closes in two hours. Redis assembles all of this before the check-in confirmation screen renders.

Without Redis, Marcus sees the same generic upgrade banner as every other passenger. With Redis, the offer is personalized: use your miles before they expire, get the upgrade you've earned, and earn the EQM that accelerates your path to Platinum.

Stage 3: Ingest

PNR, loyalty, seat inventory, and partner data flow into Redis

RDI keeps passenger state, miles balance, and seat inventory always fresh. Redis FeatureForm serves the online loyalty features. The Passenger 360 is ready before the first decision request arrives.

Redis Data Integration (RDI)Redis FeatureForm

Systems → Redis

PNR

Passenger Name Record

Booking details, seat assignment, SSR codes, itinerary state

→

LMS

Loyalty Management System

Miles balance, tier status, earn/burn history, EQM qualification progress

→

INV

Seat Inventory System

Real-time seat map, business class availability, fare buckets, upgrade inventory

→

PRT

Partner Earn System

Credit card pending miles (3,400), hotel and car rental transactions

→

REV

Revenue Management

Dynamic upgrade pricing, load factor signal, co-pay yield calculations

→

KFK

Kafka / Event Stream

Check-in events, flight status updates, gate activity feeds

→

Pipeline Status

PNR state syncAlways-fresh — RDI

Miles balance syncSub-second — RDI

Seat inventory syncReal-time — streaming

Partner miles pending3,400 from credit card

Redis FeatureForm parity100%

Decision modeInline check-in context

Stage 4: Context

Expiring miles, upgrade inventory, and tier momentum converge on one offer

The Passenger 360 surfaces three converging signals: Marcus has miles he has never used, 8,200 of them expire in 23 days, two business seats remain, and he is 14,800 EQM from Platinum. The right offer is obvious — if the system can assemble this context fast enough.

Redis RAMRedis FlexRedis Context Retriever

Durable Passenger Context

Loyalty tierGold — 3-year continuous member

Miles balance47,200 — threshold for upgrade: 25,000

Miles above threshold22,200 available after upgrade

Tier qualification14,800 EQM from Platinum — upgrade earns EQM

Partner pending3,400 miles from credit card (not yet posted)

Upgrade history3 upgrade offers received — 0 accepted (never used miles)

Live Flight Context

Miles expiring8,200 expire in 23 days — urgent use signal

Business availability2 seats — upgrade inventory closing

Upgrade window2 hours remaining before window closes

Upgrade price25,000 miles OR 12,000 miles + $89 co-pay

Load factor94% — favorable yield signal for upgrade conversion

EQM on upgradeBusiness earns 1.5x EQM — accelerates Platinum path

Context signal: Redis Context Retriever assembles the Passenger 360 — miles balance, expiration timeline, seat inventory, and tier momentum — so the decision engine has exactly the context it needs. Three signals converge: miles expiring, seats available, Platinum within reach.

Stage 5: Feature Serving

Loyalty features hydrate in milliseconds — expiration urgency, upgrade propensity, tier velocity

Redis FeatureForm serves the online features that drive upgrade offer ranking: miles surplus, expiration urgency, historical upgrade acceptance, tier momentum, seat yield, and co-pay sensitivity.

Redis FeatureForm

miles_surplus_score

Miles above upgrade threshold as a ratio of balance (0 = threshold, 1 = 2x threshold)

0.890.3ms

expiration_urgency_score

Urgency signal based on expiring miles volume and timeline

0.840.4ms

upgrade_acceptance_predictor

Predicted probability of accepting an upgrade offer given context

0.710.5ms

tier_velocity_score

Pace toward next tier relative to qualification window and remaining segments

0.680.3ms

seat_yield_score

Revenue value of converting this seat given load factor and upgrade pricing

0.770.4ms

copay_sensitivity_score

Predicted preference for miles-only vs. miles-plus-cash offer

0.620.3ms

Features Hydrated

1.7ms

P99 Lookup

100%

Train/Serve Parity

Stage 6: Decision

Miles upgrade wins — expiring miles, open seats, and Platinum momentum all align

The NBA ranker evaluates four offer paths against the Passenger 360. The miles-only upgrade wins because all three urgency signals point the same direction.

#1 Winner

MILES UPGRADE

Upgrade to business class — 25,000 miles, 2 hours to decide

Expiring miles + available seats + Platinum path converge on this offer

Marcus has 47,200 miles and 8,200 expiring in 23 days. Two business class seats remain. Upgrading uses miles before expiry, earns 1.5x EQM toward Platinum, and converts an economy seat to a higher-yield booking.

NBA score0.83

MILES + CO-PAY

12,000 miles + $89 co-pay upgrade option

Valid secondary — lowers the miles cost at the expense of cash

Useful if miles_surplus_score were lower. Here, Marcus has ample miles and the urgency signal favors the miles-only path — using more miles now reduces expiration loss.

NBA score0.61

Suppressed

GENERIC BANNER

Standard upgrade available messaging

Suppressed — ignores expiration urgency and tier context entirely

This is the offer every passenger sees regardless of context. It misses the three signals that make Marcus's moment actionable: expiring miles, seat availability, and Platinum proximity.

NBA score0.09

Stage 7: Business Impact

Loyalty revenue protected, upgrade seat filled, and miles expiry converted to engagement

Contextual upgrade offers convert expiring miles into revenue, fill high-yield seats, and deepen loyalty engagement — without replacing the airline's existing systems.

Operational Value

Upgrade revenueBusiness seat filled at upgrade yield rate

Miles expiry converted25,000 miles redeemed — 8,200 expiry risk reduced

Seat utilizationEconomy seat freed for resale or last-minute upgrade

Loyalty engagementFirst upgrade redemption — activates previously dormant miles user

EQM accelerationMarcus earns 1.5x EQM — Platinum path shortened

Offer relevancePersonalized to expiration urgency — vs. generic banner for all

Loyalty Conversion

generic banner

same offer / all passengers

→

converted

contextual offer / right moment

Stage 8: Outcome

Same passenger. Same miles. A contextual offer instead of a banner.

Without Redis, Marcus sees the same upgrade banner as every economy passenger. With Redis, the offer is built around his expiring miles, his seat availability window, and his path to Platinum.

Generic Loyalty Experience

Marcus Webb

Economy 24C — AP-847 JFK→LAX

Upgrade offer

Business class available — upgrade from $299

Miles balance

Not shown — cash upgrade only displayed

Personalization

None — same offer shown to all passengers

ignored

offer

unused

miles expire

missed

conversion

Redis-Powered Loyalty

Marcus Webb

Gold — 47,200 miles — Passenger 360 assembled

Upgrade offer

Business class — 25,000 miles, 2 seats left

Use your miles before 8,200 expire in 23 days. Earn 1.5x EQM — only 14,800 from Platinum.

Why this wins

Expiring miles + open seats + Platinum path converge

Passenger 360 assembled in <12ms. 52 features. 0.83 confidence.

83%

confidence

<12ms

decision

converted

upgrade

Stage 9: Architecture Recap

Redis becomes the loyalty context engine between existing airline systems

PNR, loyalty management, seat inventory, and partner earn systems stay in place. Redis assembles the Passenger 360 fast enough to influence the check-in and booking moment with a contextual offer instead of a generic banner.

Data Sources

Passenger Name Record (PNR)

Booking details, seat assignment, SSR codes, itinerary

Loyalty Management System

Miles balance, tier status, earn/burn history, qualification progress

Seat Inventory System

Real-time seat map, availability, fare class buckets, upgrade inventory

Partner Earn System

Credit card partner transactions, hotel, car rental pending miles

Revenue Management

Dynamic upgrade pricing, load factor, co-pay calculations, yield targets

Kafka / Event Stream

Check-in events, booking events, flight status changes, gate activity

Ingest Layer

Redis Data Integration (RDI)

Syncs PNR, loyalty state, and seat inventory into Redis — always-fresh context

Redis FeatureForm

Serves passenger loyalty features and upgrade propensity scores online

Unified Context Layer

Redis RAM

Hot check-in state, live seat inventory, active upgrade window

Redis Flex

Miles history, tier qualification trajectory, partner earn records

Redis FeatureForm

Feature store — miles balance, upgrade propensity, expiration urgency, tier velocity features

Redis Context Retriever

Assembles the Passenger 360 — miles balance, tier status, upgrade eligibility, expiring miles, and seat availability — and exposes it as structured MCP tools for the decision engine

Decision Engine

NBA Ranker

Ranks upgrade, co-pay, bonus, and tier offers by expected value and acceptance probability

Upgrade Eligibility Rules

Miles threshold, tier requirements, inventory constraints, blackout rules

Revenue Arbitration

Yield trade-off between upgrade revenue, seat fill, and loyalty engagement value

Redis Search

Peer segment matching and historical offer acceptance similarity

Output Actions

Miles Upgrade

Full upgrade offer using loyalty miles

Miles + Co-pay Upgrade

Partial miles plus cash co-pay option

Bonus Miles Offer

Earn bonus miles on this segment — no upgrade

Tier Nudge

Progress notification — miles to next tier status

Decision Target

<12ms

Primary Goal

Loyalty engagement + upgrade revenue

Redis Role

Passenger 360 in real time