The Taiwan Banker

In April 2018, following four years of preparation, the UK-based TSB Bank completed a major core banking migration. Following its separation from Lloyds Banking Group, TSB decided to move the accounts of 5.2 million customers onto Sabadell’s Proteo4UK platform in a single “big bang” transition. The data migration itself was completed, but the new platform failed almost immediately after launch. Around 1.9 million customers were locked out of their accounts. Some found their mortgage records missing, while others could access information belonging to different customers. Small businesses were unable to process payroll, branch terminals stopped functioning, and call centers were overwhelmed by complaints. One week into the crisis, TSB CEO Paul Pester publicly admitted that the bank was “on its knees.” He later resigned. An independent 262-page investigation concluded that TSB’s board had never seriously debated whether a large-scale, all-at-once migration was the right approach, nor how its risks should be mitigated. The UK Financial Conduct Authority (FCA) and Prudential Regulation Authority (PRA) ultimately fined TSB £48.65 million. The disaster cost the bank £330 million and 80,000 customers. By 2020, TSB abandoned its plans for an in-house core platform and outsourced its entire IT operation to IBM. The lesson is not that TSB lacked technical expertise; rather, its underlying approach to transformation was fundamentally flawed. This case is highly relevant to Taiwanese banks because the temptation of a “rip-and-replace” modernization strategy exists in every boardroom discussing core banking transformation. The logic seems compelling: legacy systems accumulate technical debt, maintenance costs rise continuously, and COBOL engineers are becoming increasingly scarce. Why not make one major investment and solve the problem once and for all? The flaw in this thinking is that a bank’s core system is not a factory that can be demolished and rebuilt, but a highway carrying traffic every second of every day. You cannot shut down the highway to repave it; you must upgrade it while traffic continues to flow. IBM research published in 2025 found that 94% of core modernization projects exceeded their original timelines. This statistic highlights a fundamental reality: big bang migrations run counter to the natural evolution of complex systems. Global trends are increasingly clear. IDC forecasts that 40% of banks worldwide will adopt a “sidecar” strategy by 2026, deploying modern systems alongside existing cores rather than replacing them outright. By 2028, that figure is expected to reach 70–80%. Incremental modernization is becoming the industry standard, not because it is conservative, but because it delivers the best risk-adjusted returns. This shift also reinforces the importance of hybrid cloud architecture. Hybrid is not a compromise between public and private cloud; it is the architecture best suited to banking operations. Bank workloads are inherently heterogeneous. Real-time transaction processing requires deterministic latency and near-perfect availability, making private infrastructure the optimal environment. In contrast, AI-driven analytics, risk modeling, anti-money laundering detection, and other data-intensive applications are better suited to the scalability of public cloud platforms. A 2024 CIO Dive survey found that workload-based deployment across hybrid environments has become standard practice among retail banks. Nearly half of respondents favored phased migration strategies, while another 40% modernized through digital wrappers, APIs, and containerization rather than full core replacement. The most successful modernization plans follow a consistent sequence: assess, segment, and migrate. Banks first map workloads and data flows, then categorize them according to latency requirements, regulatory obligations, and data sovereignty constraints before selecting the most suitable computing environment for each layer. Mature CIOs no longer ask when everything can be moved to the cloud, but instead which workloads generate the greatest business value in which environment. Taiwan’s banking sector enjoys a unique advantage in this regard: a relatively clear and rigorous regulatory framework. Requirements relating to data localization, audit trails, and business continuity provide valuable architectural guidance. Rather than treating regulation as an obstacle, banks can view it as a design parameter. Requirements such as data residence, auditability, and recovery time naturally inform hybrid-cloud deployment decisions. Leading international banks have already adopted this approach, keeping transaction engines within private environments while deploying customer engagement, analytics, and AI services on public clouds. This represents a practical form of risk segmentation: maintaining strict control over the most regulated functions while enabling innovation in areas that require agility. Another common reason core transformation projects fail is not technology selection, but inability to demonstrate business value. IT teams can often explain the merits of microservices, containers, and API gateways, yet struggle to answer the simple board-level question of how it will affect business results in the next quarter. The incremental approach addresses this challenge by requiring clear KPIs at every stage. A payment gateway modernization project might target a reduction in transaction failures. A cloud migration of customer data platforms could improve marketing conversion. AI-driven fraud detection models can accelerate suspicious transaction identification. Each phase becomes a measurable business hypothesis rather than a promise of benefits that may emerge years later. A useful principle observed in many AI transformations is the “70–20–10 rule”: 70% of value comes from process optimization and incremental improvements, 20% from system integration and data connectivity, and only 10% from entirely new technologies. Core banking modernization follows the same pattern. The greatest returns typically come from eliminating existing bottlenecks rather than pursuing radical replacement. A 2025 Earnst & Young study also supports this view, finding that many institutions expect to complete modernization programs within three to five years while already realizing benefits during implementation. In other words, value should be generated at every milestone, not only at the finish line. One statement from the TSB investigation stands out: “The board never adequately considered whether alternatives to the big bang approach existed.” In hindsight, this seems difficult to understand, yet many organizations have become preoccupied with new technologies while neglecting the more fundamental question of decision-making frameworks. Subsequent transformation efforts often struggle to recover when the initial strategic direction is flawed. Leading banks worldwide have demonstrated that the greatest danger in core modernization is not moving too slowly, but attempting to move too quickly. The appeal of the big bang approach lies in the illusion of a permanent solution. In practice, however, regulations evolve, customer expectations change, and competitive pressures constantly shift. Architectures must be designed for continuous evolution, not for another high-stakes overhaul twenty years later. Ultimately, beyond just a technological choice, the hybrid cloud represents a change in mindset. Success in the digital transformation journey will belong not to those who move the fastest, but to those who execute each step with the greatest precision. The author is chair of the Taiwan Internet & E-Commerce Association