Loyalty & Identity Systems
Points in 200ms. Not in overnight batch.
Loyalty platforms for millions of customers with real-time processing, flexible rules engine and partner integration.
Loyalty platform as a core system¶
A loyalty program isn’t a marketing add-on — it’s a core system that processes millions of transactions, integrates dozens of partners, and directly impacts customer retention. When a customer pays and points aren’t credited, they call the helpline. When a partner doesn’t get data, they refuse the discount.
Real-time processing¶
The traditional approach of “points will be credited in overnight batch” is obsolete. Customers expect instant gratification:
- POS/e-shop sends transaction event to Kafka
- Loyalty Engine consumes event, evaluates rules (tier multipliers, bonus campaigns, partner rules)
- Account Service updates balance (Redis for hot data, PostgreSQL for persistence)
- Notification Service sends push/SMS with update
- Entire flow < 200ms
Flexible Rules Engine¶
Marketing needs to change rules daily — without waiting for deployment:
- Tier rules — Bronze/Silver/Gold/Platinum with automatic upgrade/downgrade
- Bonus campaigns — 2× points on weekends, 3× for category, partner multiplier
- Segmentation — Different rules for different customer segments
- A/B testing — Testing campaign variants on customer subsets
- Expiration rules — Points expire after 12 months, with 30-day advance notice
Rules are configured in admin UI, versioned (audit trail), and propagated to production in seconds.
Partner integration¶
A typical loyalty program has 10-50 partners. Each needs:
- REST API with OAuth2 authentication
- Webhook notifications about point transactions
- Rate limiting per partner (system protection)
- SLA monitoring per endpoint
- Sandbox environment for integration testing
Partner API is versioned (v1, v2), backward compatible. Breaking changes with 6-month deprecation notice.
Scaling¶
- Redis for hot data (current balance, tier) — sub-ms latency
- PostgreSQL for transaction history — partitioned by month
- Kafka for event streaming — horizontal scaling of consumers
- Kubernetes — autoscaling based on throughput
GDPR and data protection¶
- Right to erasure (anonymization of point history)
- Right to portability (export in machine-readable format)
- Consent management — opt-in/opt-out per channel
- Data encryption at rest and in transit
Časté otázky
Yes. Dual-run migration — old and new systems run in parallel, data is reconciled. Customer doesn't notice the transition. Typical migration takes 2-4 months.
Under 200ms from payment confirmation. Event-driven architecture — POS sends event, loyalty engine processes rules, customer sees updated status immediately.
Yes. Each partner has their own API credentials, rate limits, rules. Central loyalty engine aggregates points from all sources. Dashboard per partner for reporting.