Architectural certainty in modern core banking: The case for "Testing-as-a-Service"

In the cloud-native core banking space, generic performance benchmarks are no longer sufficient. While baseline certifications and system-level testing prove the underlying capability of a platform, they cannot account for the infinite variability of a modern bank’s architecture and rapidly changing product lines in response to market trends. Every institution operates with bespoke financial products, unique infrastructure constraints, and highly specific traffic shapes.

Relying on generic benchmarks to predict real-world performance is a significant operational risk. To achieve true architectural certainty, banks need a "Production-Twin", a high-fidelity replica of their exact environment, products, and load profiles.

Thought Machine’s Testing-as-a-Service (TaaS) offering provides exactly this. By partnering directly with the engineers who built Vault Core, banks can proactively stress-test their unique implementations before they hit production. Through three distinct strategic case studies, this paper demonstrates how TaaS transitions banks from reactive troubleshooting to proactive scaling, saving critical engineering hours and preventing costly production failures.

A standard TaaS engagement is a highly focused, three-week process designed to deliver boardroom-ready performance data. Under strict ISO 27001 and SOC-2 data security protocols, clients share their bespoke financial products and anticipated load profiles. In return, Thought Machine builds a temporary, high-fidelity Production-Twin to execute targeted test scenarios.

The outcome is not a generic health check, but a definitive map of a bank's architectural limits and clear, actionable infrastructure tuning recommendations.

Lunar, operating a high-velocity Banking-as-a-Service (BaaS) setup, faced highly concentrated traffic profiles during periods of peak demand. They needed to know the physical limits of their standard ledger architecture before a surge in traffic caused a production outage.

By extracting Prometheus metrics directly from Lunar’s production Grafana dashboards, the TaaS team built a statistically accurate emulation of their high-velocity traffic profile and seeded the database with 600,000 active accounts to replicate real-world data entropy.

The TaaS engagement successfully pinpointed the exact operational threshold, identifying the point at which system performance, specifically Round Trip Times (RTT), would begin to degrade under extreme load.

ROI: Armed with this precise, data-backed insight, Lunar was able to move beyond guesswork regarding its scaling limits. They made an objective business decision to proactively tune their architecture, effectively future-proofing their BaaS offering and entirely mitigating the risk of a high-profile outage during peak demand.

When deploying complex enterprise software, a common friction point arises: “It works on the vendor’s machine, why is it slow on ours?” The client was experiencing performance bottlenecks in their testing environments, seeing peak online postings capped at just 429 TPS before breaching our round-trip-time (RTT) SLAs and account opening speeds limited to 51 TPS.

Rather than investigating every possible variable affecting performance in the dark, the client engaged TaaS and utilising their exact Smart Contract, the TaaS team spun up a similarly sized Production-Twin, with the same levels of historical data ingested into the database.

The TaaS environment comfortably processed 2,630 TPS for online postings (a 6x increase) and 2,320 TPS for account openings.

ROI: This definitive data proved the Smart Contract was not the bottleneck, saving both Thought Machine and the bank's engineering teams untold hours of unnecessary code refactoring. Armed with this insight, the joint engineering teams pivoted to investigate the infrastructure, quickly identifying missing database flags and restricted Kubernetes horizontal scaling as the true culprits.

The client required absolute certainty that Vault Core could support their aggressive 2-year growth projections: scaling to 59 million accounts for their Virtual Bank.

The Thought Machine team ingested the bank's projected volumes into a Production-Twin to test complex End-of-Day (EoD) and End-of-Month (EoM) scheduled processes and also duplicated their handling of repayment processing in batches.

The initial tests revealed 3 critical vulnerabilities:

• Burst mode inconsistencies: High-volume repayment bursts caused financial logic inconsistencies, something the client was unable to determine themselves without the service.
• Algorithmic complexity: A highly complex revolving line of credit smart contract, which included 1-year overdue logic, timed out due to O(n) computational complexity. In production, this would have stalled the EOD/EOM window.
• Database strategy: A single database would have been incapable of managing the anticipated high volumes of traffic, whereas a multi-database approach could scale far more effectively.

ROI: TaaS caught these inefficiencies in a sandbox, not in production. By recommending specific PostgreSQL autovacuum settings and a multi-database architecture, the bank ultimately reduced EoD processing time by 30%.

Historically, banks seeking bespoke performance validation had two options: build an expensive internal performance testing framework from scratch, or hire a traditional "Big 4" consultancy (e.g. KPMG, Deloitte). Large-scale software testing and QA consultancy engagements routinely cost upwards of $150,000 to $300,000+, take months to mobilise, and rely on generalist testers who do not intimately know the underlying core banking engine.

Thought Machine’s TaaS disrupts this model on two fronts:

The TaaS offering is not a bespoke, isolated framework requiring separate maintenance overhead. It is powered by the exact same hardened, sophisticated testing frameworks used by our internal Release Certification teams to validate Vault Core.

This shared architecture creates an unbeatable speed advantage over traditional System Integrators (SIs). When Thought Machine develops a new feature, the testing framework to validate it is built concurrently. As our product capabilities grow, our TaaS capabilities automatically grow in parallel.

In contrast, a consultancy approach is inherently reactive. Consultants must await the release of a new version of Vault Core, dedicate several weeks to understanding the changes in the software, and then construct a testing framework for validation. TaaS bypasses this discovery phase entirely, turning performance testing from a bloated capital expenditure into a lean, repeatable operational safeguard available on day one of a release.

At Thought Machine, one of our core company principles is: We are transparent. In the enterprise banking software industry, vendors frequently publish highly optimised "lab statistics"; synthetic, perfect-world benchmark numbers designed for marketing brochures, but practically impossible to replicate in a real-world production environment.

We fundamentally reject this approach. Our entire quality framework, and by extension the TaaS offering, is anchored in the concept of RBT.

RBT mandates that we do not test in a vacuum. We test with realistic database entropy, full-scale account volumes, and the complex, bespoke Smart Contracts that actually run a bank's unique financial products. When a bank engages with TaaS, they are not buying a best-case scenario; they are buying empirical truth. The performance thresholds and scaling limits identified in a TaaS report are the exact numbers you should achieve in production.

Stop guessing your scaling limits based on generic benchmarks. True architectural certainty comes from testing the exact products, traffic shapes, and infrastructure that run your business.

Whether you are validating multi-year growth, debugging complex infrastructure bottlenecks, or preparing for high-velocity traffic, Thought Machine’s Testing-as-a-Service provides the empirical truth you need before a single line of code hits production.