What It Actually Costs to Move Off Legacy Infrastructure: 6 Real Migrations, and 3 That Went £56 Million to £10 Billion Wrong

Key Takeaways

• Six companies, Veidekke, Vista Center, Vorwerk, Siemens Healthineers, Ukraine’s Medical Procurement of Ukraine agency and EPD International, migrated off legacy, on-premises or document-based systems onto modern cloud infrastructure, and every one of them can point to a specific, measurable improvement: Veidekke cut data centre costs 20% and provisioning time from months to minutes, Vorwerk hit 99.9% externally validated data quality across an eight-country SAP consolidation.
• The stakes behind getting a migration wrong are well documented. RBS, NatWest and Ulster Bank’s 2012 batch-processing outage, triggered when a software rollback was performed without having been tested, locked at least 6.5 million customers out of banking services for around three weeks and cost the group £56 million in FCA and PRA fines.
• The NHS’s National Programme for IT, launched in 2002 to build a single centralised patient record system across the health service, was formally dismantled in 2011 after roughly £10 billion in spending, undone in large part by a top-down system imposed on every NHS trust without enough local adaptation or testing before rollout.
• Sainsbury’s wrote off £290 million in 2004 after an automated supply chain and depot system, rolled out aggressively across multiple sites before it had proven reliable at scale, failed to move stock into stores and left the retailer hiring roughly 3,000 extra staff to shift goods by hand.
• None of the six companies profiled here were migrating anything close to a national health record system or a bank’s core payment infrastructure. What they share with the failures is the same underlying discipline, migrating in stages, validating data quality as you go, and treating rollback and governance as part of the plan rather than an afterthought, just applied at a scale where getting it wrong is expensive rather than catastrophic.

“Move off legacy infrastructure” sounds like a single decision, the kind of thing that goes on a roadmap slide as one line. In practice it’s dozens of smaller decisions, how much data to migrate at once, whether to run old and new systems in parallel, how to validate that nothing broke along the way, and it’s in those smaller decisions that migrations tend to succeed or fail. Six recent projects show what the successful version of those decisions actually looks like. A handful of well-documented, considerably larger failures show what happens when they don’t.

Six migrations, six specific, measurable outcomes

Veidekke, one of Scandinavia’s largest construction and civil engineering companies, was running a complex legacy IT landscape of on-premises data centres and hundreds of fragmented applications that was driving up costs and limiting flexibility. Cognizant partnered with Veidekke and Microsoft on a full cloud transformation onto Microsoft Azure, building a scalable cloud foundation with a governance model, modernising network and security architecture around zero-trust principles, and migrating applications, data and digital workplace environments. “Together with Cognizant and Microsoft, we designed and built a modern cloud platform in Microsoft Azure. We reduced platform costs by around 20 percent. Provisioning that used to take months now takes minutes. We have significantly lifted the security,” said André Hellum, COO and Head of IT at Veidekke ASA. The migration cut total data centre costs by roughly 20%, reduced provisioning lead time by more than 99%, and lowered Azure consumption by about 18% from peak levels.

Vorwerk Group, the family-owned German manufacturer behind Thermomix and Kobold appliances, had eight European countries running on outdated, heavily customised SAP ECC systems approaching end of life, leaving the company with poor inventory visibility and no clear way to track profit and loss by country. Cognizant led a two-phase SAP S/4HANA cloud migration consolidating Germany, Austria, France, Switzerland, Italy, Portugal, Poland and Spain onto a single Microsoft Azure-hosted platform, selectively migrating 20 years of historical data and more than 200 integration points. “The SAP S/4HANA-led transformation was a business-critical project for Vorwerk. Any delay or data quality issue could have seriously impacted operational continuity. We had full confidence in Cognizant’s ability to deliver on time and to the highest quality, and they didn’t let us down,” said Christian Horschinegg, Senior Vice President, Business Transformation at Vorwerk. The migration delivered 99.9% externally validated data quality and 100% accurate automated trade declarations across all eight countries.

Siemens Healthineers needed to scale a monitoring platform tracking hundreds of thousands of CT scanners worldwide, a migration to Azure Cloud that required coordinating across multiple vendors rather than one team working in isolation; Sigma Software’s team migrated the Usage Analytics component and its ETL pipelines feeding Qlik and Power BI, improving duplicate-removal functionality through parallel computation. “Sigma Software has strong Data Engineering competence, and they provided us with an expert team in a very short time thanks to their global presence and wide talent network,” said Alexey Toryanik, Head of Business Data Analytics at Siemens Healthineers. Ukraine’s Medical Procurement of Ukraine agency, working through the U.S. Government-backed SAFEMed project, had Sigma Software build the Stock and Inventory Management Module for its eStock system on AWS infrastructure built to HIPAA and ISO standards, supporting 10,000 concurrent users with triple replication across availability zones, replacing a fragmented system that couldn’t catch regional supply imbalances. EPD International, which operates the International EPD System helping businesses across 50 countries produce environmental product declarations, had Sigma Software rebuild a legacy, document-based process into structured digital workflows with version tracking and validation logic, eliminating manual invoicing through a Fortnox integration. And Vista Center, a Palo Alto nonprofit serving visually impaired and blind clients, had Xantrion migrate outdated on-premises infrastructure to the cloud and move documentation into SharePoint, with accessibility prioritised throughout given the organisation’s client base, resolving what had been unreliable network connectivity and a reactive rather than proactive IT relationship.

What happens when a much bigger version of the same decision goes wrong

The six projects above range from a mid-sized nonprofit’s IT cleanup to an eight-country enterprise SAP consolidation, and every one of them treated migration as a staged, validated process. The handful of infrastructure failures that have made the news over the past two decades share a common thread: somewhere in the process, that discipline broke down, usually under time or cost pressure, and the gap showed up at a scale ordinary businesses rarely reach.

Over the weekend of 16 and 17 June 2012, RBS’s technology division, which had recently offshored much of its IT operations, applied an upgrade to its CA-7 batch scheduling software, the system that runs the overnight processing every bank depends on to settle transactions. When problems emerged, an inexperienced operator performing the rollback wiped the payment-processing queue and the day’s scheduling data rather than safely reverting it, a rollback that had not itself been tested before being relied on. The disruption ran for roughly three weeks and affected at least 6.5 million customers across RBS, NatWest and Ulster Bank, some locked out of their accounts entirely, with figures for the full group sometimes cited considerably higher depending on how “affected” is counted. The Financial Conduct Authority and Prudential Regulation Authority fined RBS and NatWest a combined £56 million in 2014, with a separate £2.75 million fine from the Central Bank of Ireland for Ulster Bank Ireland; regulators specifically cited the absence of tested rollback procedures and inadequate change-management resilience as the root cause.

The NHS’s National Programme for IT is a different kind of cautionary tale, less a single bad weekend than a slow-motion failure of approach. Launched in 2002 with the goal of building a single, centralised electronic patient record system across the entire NHS, the programme was formally dismantled in 2011 after roughly £10 billion had been spent, well above original budget estimates, with government estimates putting realised benefit at only a fraction of that. The documented cause, laid out across National Audit Office and Public Accounts Committee reviews, centred on a top-down, one-size-fits-all system imposed on every NHS trust with insufficient local adaptation, limited clinician engagement in the design, and rollouts that reached hospitals before the systems had been proven to fit local clinical workflows. It’s a useful contrast to the six projects above, several of which, Vorwerk’s phased country-by-country rollout, Vista Center’s accessibility-first approach tailored to its actual client base, were built around adapting to the specific environment being migrated into rather than imposing a single template everywhere at once.

Sainsbury’s 2004 supply chain overhaul is the clearest example of a rollout paced faster than the technology had earned. As part of a broader IT transformation, the retailer introduced a new automated system for picking and replenishing stock across its distribution depots, rolled out across multiple sites at a pace that outran the system’s proven reliability. The automation failed to reliably move stock from depots into stores, leading to empty shelves and forcing Sainsbury’s to hire around 3,000 extra staff to move goods by hand while the automated system was fixed. The company disclosed a £290 million write-off in its October 2004 business plan, split roughly between redundant IT assets, the automated depot equipment itself, and stock losses from the disruption. Several of the six migrations profiled above moved in deliberate phases rather than switching everything over simultaneously, Vorwerk’s two-phase, country-by-country SAP rollout being the clearest example.

What this actually means for a founder weighing a migration

None of the six companies above were ever going to make front-page news for getting their migration right, and that’s rather the point: a well-run infrastructure migration is supposed to be boring. The three larger failures above didn’t fail because migration itself is inherently risky, cloud migrations, SAP consolidations and system modernisations succeed constantly, including in every case study here, they failed because a specific discipline was skipped under pressure: RBS’s rollback wasn’t tested before it was needed, NPfIT’s rollout reached hospitals before the system fit how those hospitals actually worked, and Sainsbury’s pushed automation across depots faster than the technology had proven it could handle. The founders and IT leaders behind the six migrations in this piece weren’t dealing with anything close to that scale of risk, but they were making the same category of decision, at every stage: validate before you commit, adapt to the specific environment rather than forcing a single template onto it, and treat the rollback plan as seriously as the rollout plan. It’s not a glamorous lesson. It’s also, on the evidence here, the entire difference between a case study and an incident report.