Unknown breaking point
No production-grade load model, so the team could not say how many TPS before degrading.
Case StudyFintechDigital Banking
Neo-Bank Payment Rails
Scaling a Singapore neo-bank's payment rails to handle 10x transaction growth.
ProjectOverview
A MAS-licensed digital bank was outgrowing its payments core. Zimozi re-engineered its payment-rail integrations and performance-testing strategy so it could clear PayNow, FAST and GIRO volume at scale, without downtime, ahead of a major customer-acquisition push.
Industry: Digital Bank / Neo-bank · Region: Singapore (MAS-regulated) · Services: Payment Rails Integration, Performance Engineering · Engagement: 7 months. Client identity withheld under NDA; figures representative of the engagement outcome.
TheChallenge
The bank had launched fast and won customers quickly, but its payment integrations were tightly coupled to a single processor with ad-hoc retries. Leadership was about to switch on a large acquisition campaign and needed certainty that real-time payments would not buckle.
Brittle rail integrations
PayNow / FAST / GIRO flows were coupled and synchronous, so slow responses cascaded into timeouts.
Resilience & compliance gaps
Limited failover and observability made it hard to evidence MAS TRM resilience expectations.
Release risk
No automated performance gate meant regressions could reach production unseen.
The Solution
We modelled the real world, then engineered for it.
We reconstructed true traffic shapes, payday surges, promo spikes and end-of-month GIRO batches, into a repeatable performance-test suite running against a production-like environment.
We introduced an asynchronous payment-orchestration layer with idempotency keys, queue-backed retries and circuit breakers, so a slow rail degrades gracefully instead of cascading, hardened with active-active deployment across availability zones and ISO 20022 message handling.
Load tests were wired into CI/CD so every payments change is benchmarked automatically, regressions fail the build, not the customer.
KeyFeatures
An orchestrated, observable payments backbone that absorbs spikes instead of failing.
01
Realistic
Realistic
load modelling
Reconstructed true traffic shapes into a repeatable performance-test suite.
02
Decoupled
Decoupled
payment rails
Async orchestration with idempotency keys, queued retries and circuit breakers.
03
Active-active
Active-active
resilience
Multi-AZ deployment and ISO 20022 message handling for clean interoperability.
04
Live observability
A real-time dashboard for throughput, latency and success rate across every rail.
05
Performance
Performance
as a release gate
Every payments change is load-tested automatically in CI/CD.
TechStack
PayNow, FAST and GIRO integrations on event-driven microservices, Kafka queues and Kubernetes, hardened with active-active multi-AZ deployment and k6 load testing in CI/CD.
Kubernetes
Kafka
AWS
Business Impact
Confidence to scale, and the evidence to prove it.
The platform sustained 12,000+ transactions/sec at peak load with p99 latency under 350ms and 99.99% availability post-launch, giving the bank 10x throughput headroom and ~70% fewer payment errors.
Get started