Why Bank Software Keeps Getting More Expensive: The Quality Model Built with Deloitte
Deloitte Consulting’s banking clients kept raising the same concerns: systems growing more complex, costs rising faster than functionality, and no shared way to reason about software quality. A research collaboration with Deloitte’s technology strategy practice turned those concerns into SQuAP, an empirically validated quality model, later formalised as an open ontology.
The situation
Deloitte Consulting’s Technology Strategy & Architecture practice in Milan works with all the main Italian banking groups. Across engagements, their clients described the same pattern: information systems that had grown into layered accumulations of custom applications, software packages, and point-to-point interfaces. Maintenance budgets kept rising, documentation lagged behind the code, and reverse engineering had become slow and expensive.
The senior managers raising these concerns had no shared framework that explained what was happening. International standards for software quality existed, yet none of the banks used a comprehensive quality model, and the standards treated quality, architecture, and process as three separate subjects. Deloitte and researchers at the University of Bologna set out to build the missing model together.
Software quality in banking cannot be assessed one standard at a time. Quality, architecture, and process behave as a single interacting system, and business decisions propagate through it: cost-cutting at executive level resurfaces as architectural stratification, degraded maintainability, and rising total cost of ownership.
The engagement
The study ran as a four-phase Delphi process over more than a year. A panel of 13 senior managers, CIOs, chief data officers, and IT architects drawn from Deloitte’s expert network brainstormed, narrowed down, and validated the sector’s software quality concerns in facilitated rounds. A second panel of 124 carefully selected domain experts then evaluated every concern through a structured survey. Together, the panels represented banks holding 82% of the Italian banking sector’s aggregated net worth and 95% of its revenues.
The result was a set of 28 empirically validated quality factors, each mapped to the three relevant international standards: ISO/IEC 25010 for software quality, ISO/IEC 42010 for architecture description, and ISO/IEC 12207 for software processes. That mapping became SQuAP (Software Quality, Architecture, Process), a meta-model that connects the three standards through the factors practitioners actually struggle with.
What the research found
- Cost pressure drives the quality spiral. Under budget and time constraints, banks stratify core systems instead of evolving them: new layers are added on old software to ship functionality quickly. Each tactical layer raises integration costs and makes the next change harder, so the pressure that created the shortcut intensifies.
- Nobody had a complete map. None of the organizations studied claimed a complete overview of the source code running their information systems. Code ownership was fragmented across vendors, outsourcing partners, and internal teams, which inflates the total cost of ownership and makes strategic planning guesswork.
- Legacy skills retire faster than legacy systems. Core banking still runs on COBOL, which processes millions of batch transactions reliably. The developers who master it are retiring, new developers are not learning it, and the quality of custom development was reported to be declining as layers of new paradigms accumulate on old code.
- Documentation is the first casualty and a lasting cost. Under deadline pressure, documentation is the first activity to be skipped, and suppliers often withhold technical documentation as a lock-in strategy. The predictable consequences: expensive reverse engineering, difficult supplier switches, and system analysis built on incomplete information.
- Short-term executive incentives inflate long-term cost. The panel linked quarterly goals and fast C-level turnover directly to the preference for incremental fixes over structural solutions. These are business decisions, yet they surface as architectural anti-patterns and rising maintenance costs.
SQuAP gives executives a management tool to map business decisions into IT categories and trace their impact on the system at large: how a budget decision surfaces in the architecture, and how an architectural shortcut resurfaces in next year’s maintenance costs. Because the model is built on ISO standards, IT and non-IT leadership share an accountable vocabulary, and the total cost of ownership conversation stops being anecdotal.
The model was later formalised as SQuAP-Ont, an ontology developed with the Semantic Technology Laboratory of the Italian National Research Council and published openly at w3id.org/squap. Quality assessments built on it are machine-readable, so tools can reason over quality factors against a common reference.
Are rising IT costs pointing at a deeper quality problem?
Stratified systems, thinning documentation, and growing maintenance budgets follow a pattern that can be mapped and traced. That is what a quality model is for.