Frequent-flyer programmes with real-time redemption.
Earning, tiers, redemption and partner ecosystems for airline loyalty, engineered for the latency and peak-season load an award search has to survive.
- Characterise
- Parallel run
- Staged cutover
- Operate & extend
An airline loyalty programme is a second airline.
It has its own currency, its own liability on the balance sheet, its own partners and its own peak season. It is also, at many carriers, the system with the oldest code. Award search that takes too long. Promotions that need a release. A tier calculation that only two people understand, one of whom has retired. And a migration everyone agrees is necessary and nobody will sign, because the programme cannot stop for a weekend.
The migration is the part we take most seriously. Parallel run against historical transactions, staged cutover by segment, and a plan built to keep the programme operating while it happens — the platform overview describes the method.
Six modules, configured for airline economics.
Everything below is a setting on the running engine, versioned and simulated before it goes live — not a change request to a vendor.
Miles & tier engine
Accrual rules by fare class, route, cabin and partner; tier qualification on miles, segments or spend with soft-landing logic and qualification windows. All configuration, not code — and every rule version stays readable for the audit.
Real-time redemption
Award search and burn flows engineered to a latency budget agreed with you — the difference between browsing and booking. Points are reserved on search, committed on ticketing and released on failure, so a PSS timeout does not strand a balance.
Partner ecosystems
Airline, hotel, card and retail partners on a shared currency with clean settlement. Each partner has its own rates, caps and settlement calendar; the ledger nets the month out and the partner portal shows them the same numbers you see.
Tier prediction
ML models flag members trending toward churn or upgrade, so offers land before the status email would have. Scores refresh regularly, feed segments, and are measured against holdouts rather than against last year.
Disruption-aware
Goodwill and recovery flows for the days aviation happens — configurable, budgeted, logged. A cancelled flight triggers a rule, not a spreadsheet, and the cost of goodwill is visible per event.
IATA-world integration
PSS, DCS and e-commerce integration patterns for airline estates. Flown-segment events drive accrual, check-in events drive recognition, and the booking engine calls redemption over a versioned API.
What it does, in detail.
Award search engineered to a latency budget.
Award search is the moment a member decides whether the programme is worth it, so it is engineered to a latency budget rather than a best effort. Balances come from a hot ledger, award availability is pre-computed where the partner allows it, and the search-to-ticket flow is a two-phase movement that releases points when a booking fails downstream. The budget is agreed with you, measured at p95 and reported through peak season.
- Redemption is its own service — a partner feed backlog does not slow a member’s search
- Reserve on search, commit on ticketing, release on failure
- Service latency is on your dashboard, alongside PSS and payment timings
Accrual rules your revenue team can read.
Fare-class accrual, route multipliers, cabin bonuses, partner rates and promotional boosts are rules with a condition and an outcome, not branches in a codebase. Change a rule, simulate it over last year’s flown segments, see the miles issued and the liability it creates, then schedule it. Promotions that used to wait for a release become configuration. The rules shown here are examples.
- Simulate any rule set against historical flown segments before it goes live
- Tier qualification on miles, segments, spend or any combination
- Every accrual carries the rule version that produced it
Example rules. Yours are configured with you.
Goodwill that is budgeted, not improvised.
Disruption is when a loyalty programme earns or loses trust, and it is usually handled by a spreadsheet and an inbox. In Loyalty & Retention, a cancellation or a long delay is an event: a rule decides who gets what, the credit is issued against a goodwill budget, the member is told through the channel they prefer, and finance sees the cost per event. Contact-centre agents get a bounded discretionary allowance instead of an unlimited one.
- Rules by delay length, cabin, tier and cause, with a budget per event type
- Every goodwill credit is on the ledger with its trigger and approver
- Recovery offers measured against holdouts like any campaign
Tier prediction that acts before the status email.
A member a few thousand miles short of Gold in November is either a retention win or a churn statistic, and the difference is whether anyone noticed in time. The prediction models score every member regularly for upgrade and churn probability; segments update; offers go out with a budget and a holdout. Redemption engagement is reported against those holdouts, which is how you know a campaign moved it rather than the season.
- Regular scoring with model versions and evaluation metrics on the record
- Tier-nudge, win-back and status-match offers with budgets and caps
- Uplift reported against a holdout, not against last year’s season
One programme, four seats.
Run the programme at marketing speed.
Tiers, award charts, promotions and partner rates are edited, simulated and scheduled in the console. Every change is versioned and attributable, so the programme moves quickly and stays auditable.
- Simulate a promotion on last year’s flown segments
- Schedule rule versions to a date and time
- Engagement, redemption and liability on one screen
A miles liability you can trace.
Every mile carries its rule version, partner and cost basis. Liability by expiry bucket, breakage and partner settlement are reports over the ledger, reconciled to the transaction — not a spreadsheet beside the system.
- Liability by currency, partner and expiry bucket
- Settlement statements per partner and period
- Goodwill cost per disruption event
PSS in, ledger events out.
Flown-segment, check-in and booking events arrive over Kafka or webhooks; the engine consumes them idempotently and publishes ledger events back. OpenTelemetry traces run from the PSS call to the balance update.
- Versioned REST and event APIs for the booking engine and app
- Sandbox tenant with replayable sample traffic
- Runbooks, SLOs and dashboards shared with your ops centre
Answer the member on the first call.
Agents see the member’s full movement history with the rule behind each line, a bounded discretionary allowance, and the recovery flows already applied. Retro-claims, status matches and goodwill are guided actions, not free text.
- Full ledger history with rule versions per line
- Bounded goodwill allowance per agent role
- Retro-claim workflow with PSS lookup
Where it sits in an airline estate.
Legacy engine versus Loyalty & Retention.
Patterns for airline estates.
PSS
Bookings, ticketing and flown-segment events for accrual, redemption ticketing over a versioned API
DCS
Check-in and boarding events for recognition, upgrades and disruption triggers
Booking engine & app
Award search and burn over REST or GraphQL; balance and tier in the member profile
Alliance & partner carriers
Interline accrual and redemption with per-partner rate tables and settlement
Co-brand card issuers
Card-spend accrual feeds and settlement statements per period
Hotel, car and retail partners
Earn and burn at partner rates, partner portal reporting
Marketing platforms
Segment sync and delivery through email, push and SMS providers
Data warehouse & BI
Ledger events to your warehouse; Power BI and Grafana models
Migration designed to avoid a freeze.
Characterise
The legacy engine’s behaviour is pinned with golden-master tests over historical transactions — including its quirks. Fare-class accrual, tier maths, award charts and partner settlement are configured to match.
Parallel run
Loyalty & Retention runs shadow calculations on live traffic until outputs match the record, exception by exception.
Staged cutover
Earning first, then redemption, member segment by segment — every step reversible, every rollback rehearsed. Full cutover across segments is planned to avoid a freeze on programme operations.
Operate & extend
Tier prediction goes live, partners onboard, and promotions become configuration. The same team that built it runs it — uptime agreed with you according to your needs, observability throughout, and incident discipline your ops centre will recognise.
Frequently asked questions
One engine, five verticals.
See it on your own data.
A guided demo with an engineer who operates the platform — configured toward your programme, not a canned deck.