Custom Credit Emulator: QA Testing 2-3x Faster in 3 Months

Our client is a B2B2C fintech platform that connects with third-party credit services both in the US and Europe. In order to accelerate their QA process and decrease their dependency on back-end systems for integration testing, Pynest built a fully-fledged admin SPA from scratch, mimicking credit services. Application and transaction management, configuration of hierarchical settings, logging and dual ID search became possible from a single interface. Search time for transactions decreased by 80%, and data update time dropped from minutes to seconds thanks to WebSocket and Server-Sent Events technologies.

80%

Faster transaction search

90%

Lower data update latency

2-3x

More testing throughput

Custom Credit Emulator: QA Testing 2-3x Faster in 3 Months

Client Overview

The product is a web-based admin application designed to replicate external credit services for testers and developers building credit provider integration solutions. Through a single integrated UI, users can build and maintain test applications, transactions, accounts, refunds, offers, configurations, and webhooks without direct access to backend services or databases. The platform supports multi-level configuration, from Global to Location, Application, and Transaction levels. It also provides live-streamed events and cross-service search capabilities, enabling users to work across all services from a single interface.

CompanyB2B2C SaaS

RegionEU / US

IndustryFintech

CompanyB2B2C SaaS

RegionEU / US

IndustryFintech

Business Problem

Situation Before

Before the platform was developed, testing of credit-service integrations was conducted directly via the system's backend and databases. The search for one transaction took 5 to 10 minutes due to the lack of common IDs and the absence of a fast way to query data. Data was outdated because any updates appeared only after 1 to 2 minutes. Replication of a complicated scenario was not easy, and hierarchical settings had no separate user interface, which made every parameter change slow and risky.

Core Pain

No test toolingQA had no dedicated tools to build test scenarios or track the live state of the system.

Cross-system blind spotsTracing one transaction across systems with different identifiers was nearly impossible.

Onboarding backlogIntegration testing piled up, so credit-provider onboarding kept falling behind.

Risks

Slow QA cycles delayed new credit provider integrations, slowing the time-to-market for new lending products.
A lack of real-time data in QA resulted in inaccurate test results due to outdated data.
Manual backend testing was difficult for non-technical QA team members and increased the risk of configuration errors.

Constraints

Complex business logic: hierarchical configuration inheritance and numerous transaction processing scenarios
High performance requirements: large paginated tables, real-time updates and complex forms all in one interface
Iterative delivery: QA needed to start using the system before all features were complete

Project Goals

Give QA engineers and developers a unified UI for managing test credit service scenarios, eliminating direct backend and database dependency
Accelerate integration testing cycles to speed up credit provider onboarding
Ensure consistent reproduction of complex edge cases through a purpose-built emulation environment

Design a modular SPA using Feature-Sliced Design architecture (React/TS/Vite)
Establish clear separation between server state (TanStack Query) and UI state (Zustand)
Deliver real-time interface updates via WebSocket and SSE
Build a dual-ID search system (external ID and MongoDB ObjectId)
Implement hierarchical configuration management with parameter inheritance across four levels
Implement hierarchical configuration management with parameter inheritance across four levels

Transaction search became noticeably faster for QA workflows
Real-time data updates appear within seconds after system events
Hierarchical configurations can be managed in the UI without backend help
The interface stays responsive with large tables and live updates at the same time
New modules can be added without rewriting the architecture

Solution

The frontend is a modular admin SPA built with Feature-Sliced Design. The codebase is split into independent functional slices, keeping it maintainable by a single developer despite constant API changes. TanStack Query manages server state and backend synchronization, while Zustand handles UI state, including modals, filters, and local logic. WebSocket and SSE deliver real-time events, with TanStack Query cache invalidation updating the UI without page reloads. Server-side pagination, React.memo, and useMemo keep tables and forms responsive under heavy data loads.

There were 5 modules shipped. Module 1 gave QA full CRUD access to test apps and transactions, with server-level pagination and filtering through TanStack Table. The forms are validated with React Hook Form and Zod. Module 2 introduced a four-layered configuration hierarchy from Global to Transaction with inheritance and override capabilities. Module 3 provided real-time event streaming over WebSocket and SSE along with an Activity Log with deep-linking to relevant entities. Module 4 provided dual-ID search with external ID and MongoDB ObjectId. Module 5 covered CSV export for QA.

GitHub Actions runs CI/CD and Vite handles the build, so deployment stays the same for every merge. The source code is managed at GitHub, and a pull request process is maintained so that every code change must pass through code review before getting merged. ESLint and CI are checked on every merge, which keeps up the quality of the code as the codebase grows, even with only one developer on the frontend. MSW mocks the APIs for tests, so the frontend is testable without a live backend. The whole app runs on AWS.

Platform access is based on OAuth single sign-on. There is a single Axios layer where all the calls to the API are performed, reducing the risk of receiving malformed or unauthorized requests. The role-based access control is implemented by means of conditional rendering and route protection, making sure that each role has access only to its relevant part of the interface. TypeScript strict typing reduces the surface area for runtime security issues across the codebase.

Architecture decisions

Technical initiatives

Infrastructure & DevOps

Security & Compliance

Key Steps

Oct 2025

Kickoff and architecture

In October 2025 the team set up the project, designed the architecture, scaffolded the Feature-Sliced Design structure and aligned the API contract with the backend team. This groundwork let modules land later without rework.

Nov 2025

Core management modules

Through November 2025 the team built application and transaction management and the configuration hierarchy and added a typed Axios layer so backend contract changes stayed contained in one central place.

Dec 2025

Real-time and search

Across December 2025 the team delivered the WebSocket and SSE real-time layer, the Activity Log, dual-ID search and performance tuning, which together removed manual refreshes and slow crosssystem lookups.

Jan 2026

Reporting and delivery

From January 2026 the team added CSV export and reporting, ran final QA and set up CI/CD for delivery. The MVP shipped inside the planned three-month scope, and work then continued with new emulation scenarios.

1 / 4

Oct 2025

Kickoff and architecture

Nov 2025

Core management modules

Dec 2025

Real-time and search

Jan 2026

Reporting and delivery

Results & Impact

80%Faster transaction searchFrom 5-10 min to 1-2 min

90%Lower data update latencyFrom 1-2 min to seconds

2-3xMore testing throughputRun via UI, not the backend

40%Fewer wasteful re-rendersInterface load down 40-50%

Business impactThe platform took the backend out of the QA equation entirely. Engineers got an interface built for the job, which meant complex credit integrations no longer required database access or backend workarounds. Tighter testing cycles brought credit providers online faster and new lending products closer to launch.

Operational improvementsThe QA engineers conduct all their regular testing solely using the UI now as there is no backend or database to deal with. That has cut down errors and made it quicker to get new employees up to speed. Event streaming through WebSocket and SSE has done away with manual refreshing and the stale-data errors that came with it. Dual-ID search has solved a longstanding identification issue across systems. The Feature-Sliced architecture lets any developer add modules without breaking existing work.

Project Snapshot

Team size & roles6 FTE: 1 FE, 2 BE, 1 QA, 1 PO, 1 PM

Engagement modelDedicated team

Duration3 months for MVP, ongoing

Core tech stack

ReactTypeScriptTanStack QueryZustandWebSocketSSEReact Hook FormZodNode.js

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