Scalable Loyalty
Enterprise loyalty you can trust in production
ReactorCX runs enterprise loyalty programs at production scale. It is the enterprise loyalty platform behind billions of loyalty transactions a year, tens of millions of API requests a day in its largest deployment, within a platform-wide API response SLA under 200 milliseconds, on a multi-AZ, multi-region cloud architecture that upgrades without maintenance windows.
Operations
Platform operations · sample view
API p95 latency
Availability
Deploy status
Rolling · no restartMicroservices · replicated NoSQL · Kafka backbone
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.
What "zero downtime" means here
Zero unplanned, member-visible production downtime for programs where ReactorCX is the system of record, measured from the date published with the metric, excluding scheduled maintenance windows agreed with the client. Platform upgrades use rolling deployment and configuration publishing without restarts, so upgrades do not consume maintenance windows.
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.
Definition: zero unplanned member-visible production downtime for system-of-record programs. Scheduled windows excluded.
How the platform stays up
Three architectural choices support continuous delivery: microservices, a replicated NoSQL data layer and an Apache Kafka event backbone.
Multi-AZ, multi-region
Deployment across multiple availability zones and regions for disaster recovery, latency optimization and regulatory data residency where required.
Microservice architecture
Scales up and down with traffic volume, stays resilient through individual instance failures and replaces failed instances transparently.
Replicated NoSQL data layer
Real-time transaction processing with real-time replication and fault tolerance, supporting the platform's horizontal scaling profile.
Apache Kafka event backbone
Distributed message brokers handling hundreds of millions of events a day, carrying both replication events and business events downstream.
Rolling upgrades, service-level rollback
Containerized services roll forward one at a time and roll back per service, without interrupting member-facing service.
Publish without restart
Configuration and rule changes publish into production without a platform restart, so program changes ship daily without downtime.
Region A
AZ 1
services
NoSQL replica
AZ 2
services
NoSQL replica
AZ 3
services
NoSQL replica
Event backbone
Kafka event backbone
Replicated across zones
Standby region
Region B
Disaster recovery, residency
Replicated services and data
Always warmWalk through the availability model with a platform architect.
Book an architecture reviewHow the platform stays fast
Every activity is evaluated through the same real-time path. Batch rows become individual real-time transactions, so batch behavior is consistent with synchronous behavior. The platform-wide API response SLA is not an aspirational target; it is the operational standard maintained across production volume.
Peak throughput is proven on production traffic, not benchmarks: every SafeSwitch™ replay doubles as a load test on the program's own history. See Safe Migration.
How this is measured
Distinct active promotions per month in the largest single retail deployment. As of 2026-Q3.
How this is measured
Distinct physical sites transacting through ReactorCX integrations, each counted once. As of 2026-Q3.
How this is measured
Platform-wide API response-time service level. As of 2026-Q3.
Proven where complexity is the baseline
7-Eleven
Sub-second response at the register, faster than the previous system, with headroom validated by replay before the switch.
Western Union
Multi-currency, 24/7/365 operation across dozens of countries.
MGM Resorts
Gaming, hotel, food and beverage and entertainment on one program, with no PII in the loyalty platform.
bp
Fuel sites on different underlying systems, unified under one program.
Frequently asked questions
What is the ReactorCX API response SLA?
A platform-wide API response SLA under 200 milliseconds (<200 ms Platform API response SLA). The percentile and measurement method are published with the metric.
How does ReactorCX handle peak volume?
Real-time evaluation on a horizontally scalable microservice layer, a replicated NoSQL data layer and an Apache Kafka event backbone, with replay-based performance evidence from each migration.
How are upgrades deployed?
Rolling, containerized deployments with service-level rollback. Configuration and rules publish without a platform restart, so upgrades do not consume maintenance windows.
Where does ReactorCX run?
AWS, cloud-native, across multiple availability zones and regions for disaster recovery, latency and data residency.
See the platform under your load profile.
An Architecture Review covers availability, throughput, environments and the integration surface, with the engineers who operate ReactorCX in production.