Manual Regression Testing Cut 94% for an Exchange Platform

Our client runs an enterprise financial exchange platform on a hybrid microservices and monolith architecture, with deep integrations across message brokers, legacy monoliths and databases. Pynest engineers embedded in the client's QA Automation group and built end-to-end and integration test coverage across 16 business processes, a customized internal framework and AI-powered regression analysis tools. Manual regression testing before releases dropped from 160 hours to 10, on-call regression analysis went from 3 hours to 30 minutes, and 3 previously unautomatable business processes came unlocked.

94%

Less manual regression time

83%

Shorter regression duty shifts

3

Processes newly automated

Manual Regression Testing Cut 94% for an Exchange Platform

Client Overview

A large financial exchange infrastructure built using a hybrid microservices and monolithic architecture, integrating RESTful services, asynchronous messaging through ActiveMQ Artemis, and multiple databases, including PostgreSQL and Firebird. The platform supports critical business processes such as reporting systems, securities listing, cross-market operations, and historical data stores. It is designed to meet the reliability and data integrity requirements expected of exchange-grade infrastructure while supporting communication across its different architectural components.

CompanyEnterprise

RegionEurope

IndustryFintech

CompanyEnterprise

RegionEurope

IndustryFintech

Business Problem

Situation Before

A large chunk of the regression testing had to be done manually because of the complexity of business processes involving several microservices, asynchronous queues and legacy monolith components. Preparation for each release involved more than 160 hours of manual testing. A subset of integration tests could not be reliably automated because of synchronization issues across services, message brokers and databases. On-call regression shifts involved an average of 3 hours of manual analysis per session to diagnose test failures and localize problem areas. Lack of complete functional documentation added further friction to test development and incident analysis.

Core Pain

Manual Regression LoadPreparing each release took more than 160 hours of manual regression testing, which slowed releases.

Unautomatable ScenariosCertain scenarios could not be automated because of sync issues across services, brokers and databases.

Slow On-Call AnalysisOn-call shifts averaged 3 hours of manual analysis per session to diagnose failures and localize problems.

Risks

A large amount of manual regression testing increased time-to-market for releases and increased the chance of human error
Unstable or unautomatable integration processes meant that critical business processes didn't have dependable automated coverage
When incidents arose, it took longer to localize them during regression duty shifts, which slowed down release decisions
Older, monolithic systems and incomplete documentation led to gaps in knowledge, making it harder to spot defects in complex

Constraints

Complex hybrid architecture (microservices, monolith, message broker, multiple databases) with no service-level test environments
Asynchronous business processes required custom synchronization approaches before E2E automation was feasible
Incomplete documentation required independent system analysis and close collaboration with developers and analysts
High reliability standards of exchange-grade infrastructure demanded stable, reproducible test execution

Key Steps

09/24–03/25

Reporting systems automation

Pynest QA engineers embedded in the client QA Automation group under a team augmentation model, working Scrum with regular sprint syncs with analysts to clarify requirements and business processes. The first phase automated end-to-end and integration testing for the reporting systems, the first of the 16 target business processes.

04/25–09/25

Securities listing and internal library

The second phase added securities listing to the automated coverage and built out the in-house test framework and reusable Python libraries for teams to share. Pynest also designed the custom service synchronization mechanism that coordinates state across microservices, monolith, queues and databases before tests run, the unlock behind 3 previously unautomatable processes.

9/25–3/26

Cross-market and historical data

The third phase extended coverage to cross-market operations and historical data stores, bringing the automated suite across all 16 target business processes. Pynest built the LLM and LangChain regression duty tool, which cut shift duration from 3 hours to 30 minutes, with Jenkins running the suites in CI/CD.

1 / 3

09/24–03/25

Reporting systems automation

04/25–09/25

Securities listing and internal library

9/25–3/26

Cross-market and historical data

Results & Impact

94%Less manual regression timeFrom 160 hours to 10 hours

83%Shorter regression duty shiftsFrom 3 hours to 30 minutes

3 Processes newly automated Previously unautomatable

Business impactRelease preparation effort was reduced by 94%, which meant new features could reach users faster. On-call regression analysis used to take 3 hours. Now it only takes 30 minutes, easing the burden on QA engineers and helping the team make faster release decisions. Previously blocked business process automation was unlocked, extending quality coverage to areas where defect risk had gone unchecked before.

Operational improvements

Project Snapshot

Team size & roles18 FTE: 1 QA Lead, 13 Automation QA, 2 DevOps, 2 BA

Engagement modelTeam augmentation

Duration18 months, September 2024 to March 2026

Core tech stack

PythonPyTestActiveMQ ArtemisPostgreSQLFirebirdJenkinsGitKubernetesAllure TestOpsLLMLangChain

See Other Works

Secure LLM Gateway for a 63,000-Person Enterprise

Secure enterprise AI access with centralized governance, lower latency, content controls, and full token usage visibility

Bond Settlement Cut from 2 Days to 30 Minutes

End-to-end capital markets platform automating bond issuance, trading, settlement, and investor servicing

API Response Time Down 80% for a Bloomberg Rival

Financial analytics platform delivering real-time market data, news processing, and scalable backend services for fintech

Power Station Analytics: Algorithm Suite Delivered

Real-time equipment analytics platform for a power station with automated calculations, monitoring, and reliable fault detection

Show all