High-scale transaction processing, deterministic and real time
ReactorCX is the enterprise loyalty platform built to process billions of loyalty activities a year through one deterministic engine, in real time, with sub-second responses across every channel. It was architected from inception for real-time processing at enterprise scale, runs the largest single retail loyalty deployment in production, and proves headroom by replaying production traffic before any cutover.
How this is measured
Count of activities evaluated by the ReactorCX rules engine across all production tenants in the trailing 12 months. As of 2026-Q3.
How this is measured
Real-time API requests per day served for the largest single ReactorCX deployment (retail); not platform-wide. As of 2026-Q3.
How this is measured
Platform-wide API response-time service level. As of 2026-Q3.
How this is measured
No unplanned production outage of the ReactorCX platform since it became the primary system of record (first SafeSwitch cutover, 7-Eleven, February 2020). As of 2026-Q3.
Real-time loyalty at scale is where "batch with a veneer" breaks
The commute spikes traffic three to four times. A launch drop does it in seconds. A holiday weekend does it across every property at once. A platform that is batch-first with a real-time layer bolted on top holds until exactly that moment, and then members see wrong balances, staff see spinning screens and finance sees a reconciliation problem.
The other failure is quieter: non-deterministic evaluation, where the same basket produces different outcomes depending on ordering or load, and nobody can explain why.
One engine, one schema, one execution log
- Polymorphic activity engine: purchases, stays, play, bookings, reversals and non-transactional events through a single extensible schema.
- Deterministic rule execution: predictable, ordered, logged; hundreds of concurrent promotions resolved without conflict.
- Synchronous responses at the pump, the register, the front desk and the floor, inside a platform-wide API SLA.
- Real-time event publication to Kafka, Kinesis, EventHub and webhooks, so downstream systems see every event as it happens.
- Cloud-native on AWS: microservices, NoSQL, multi-AZ, continuous delivery without maintenance windows.
How an activity moves through ReactorCX
Activity
Any channel, any type, one schema with standard and extended fields.
Rules
Earn, burn, tier, promotion and arbitration rules evaluated deterministically.
Member state
Balances, sub-balances, tiers and aggregates updated with lineage.
Events
Level-1 change events and Level-2 business events published in real time.
Handlers and log
Downstream handlers, scheduler, and a per-activity execution log.
Read the processing model, event catalog and architecture in the open, then bring your diagram.
Book an architecture reviewWhat enterprise architects ask about, answered
| Question | Answer | Read more |
|---|---|---|
| Is it real-time or batch? | Real-time-first: synchronous evaluation at the transaction, with batch through FeedXChange for partner and bulk data. | Integration and Events |
| How is peak handled? | Horizontal scaling across multiple availability zones on AWS; NoSQL data tier; Apache Kafka event backbone. | Enterprise Reliability and Scale |
| How is headroom proven? | SafeSwitch replays production traffic at multiples of normal volume before cutover. | Safe Migration |
| How are releases handled? | Continuous delivery without maintenance windows; lower environments kept in sync with production. | Architecture and Extensibility |
| How is correctness proven? | Deterministic execution with a per-activity execution log; every point traces to its rule, activity and member. | Financial Integrity |
| What does downstream see? | Every loyalty event published as a real-time signal to Kafka, Kinesis, EventHub, Pulsar, GCP or webhooks. | Event Catalog |
The largest single retail loyalty deployment
7-Eleven's 7Rewards runs on ReactorCX across its US and Canadian stores, with real-time recognition at the pump and the register and capacity sized well beyond current requirements. Read how the program migrated and what it runs today.
How this is measured
7-Eleven stores in the US and Canada on which 7Rewards runs through ReactorCX, per 7-Eleven's public store count at the as-of date. As of 2026-Q3.
How this is measured
Distinct physical sites transacting through ReactorCX integrations, each counted once. As of 2026-Q3.
Scale questions
How does ReactorCX process loyalty activity at scale?
Every activity, whether a purchase, a stay, a play session, a booking, a reversal or a non-transactional event, passes through one polymorphic activity engine and a single extensible schema. Rules evaluate against member state, the ledger is updated, and events are published, with deterministic and auditable execution and sub-second response across every channel.
What happens at peak?
ReactorCX runs cloud-native on AWS with horizontal scaling across multiple availability zones and a NoSQL data tier, and absorbs commute, holiday, launch-day and event-weekend spikes without degradation. SafeSwitch replays production traffic at multiples of normal volume to prove headroom before cutover.
Is real-time processing deterministic?
Yes. Every promotion outcome is predictable, ordered and logged, so nothing conflicts and everything is traceable. Each activity keeps an execution log of what ran, why and what changed.
Validate the architecture before the sales call.
The documentation is public. When you are ready, the Architecture Review is a working session with Loyalty Methods engineering using your landscape diagram.