When asset-intensive enterprises buy digital capability the way they buy steel, trucks, and turbines, the discipline that built their competitiveness can work against their digital transformation. Lidl abandoned an ERP program after an estimated €500 million of investment. Birmingham City Council's Oracle implementation, budgeted at £19 million for its first phase, is now forecast at £144 million for the full program including reimplementation. MillerCoors and its system integrator ended a major SAP program in litigation. None of these organizations lacked procurement discipline — they applied a discipline built for a different class of purchase.
Your procurement function is capable. It has spent decades extracting value from suppliers of equipment, materials, and EPC contracts, and it does what it does well: writes a detailed specification, runs a competitive tender or RFP, evaluates bids across weighted criteria — price, delivery, quality, warranty, service, supplier standing — and awards to the compliant offer that scores best. Often not the cheapest bid; the discipline is genuinely multi-criteria, and for its intended purpose it works.
Eighteen months later the program is late, over budget, the "compliant" system does not fit the operation, and the vendor relationship has turned adversarial. The pattern is common enough to have its own literature: Peter Smith, a former CIPS president, filled a book, Bad Buying, with organizations wasting large sums this way, technology purchases prominent among them. Failure-rate estimates for large IT programs range, depending on how failure is defined, from roughly a fifth failing outright to well over half falling significantly short of budget, schedule, or benefits. Almost none of that risk is visible in the bids: a vendor that priced it realistically would rarely win a tender, so offers systematically understate it — the dynamic the megaproject literature calls strategic misrepresentation and auction theory calls the winner's curse. The evaluation model then optimizes over numbers that were shaped to win it.
What failure looks like: four public cases
The grocery group spent seven years replacing its in-house merchandise system with SAP, then stopped the program in 2018 and reverted to the legacy system, after an investment estimated in press reporting at around €500 million. The software itself was not the primary cause. Lidl required the new platform to mirror its existing processes — notably valuing inventory at purchase rather than retail price — triggering customizations that undermined the platform's integrity, compounded by heavy reliance on external integrators without matching internal ownership. Reporting suggests the valuation conflict was identified early and management chose to proceed regardless — as much a governance lesson as a specification one.
The sourcing lesson stands nonetheless: "we specify, the vendor complies," the bedrock of industrial tendering, converts a product purchase into a high-risk bespoke build.
Europe's largest local authority set out in 2018–19 to replace SAP with Oracle Fusion, with a phase-one budget of about £19 million. The council went live in April 2022 with a system unable to support compliant financial management, operated for more than two years with degraded financial control, and spent millions on manual workarounds; a full reimplementation is under way. Grant Thornton, the external auditor, put the excess over the original budget at £90 million or more in its February 2025 report; by early 2026 the total program forecast through 2027/28 stood at £144 million. The auditors cited inadequate governance, a culture in which bad news was unwelcome, insufficient investment in business change, departure from the program's own adopt-standard-functionality design principle, and a shortage of in-house expertise. The implementation contributed to the council's section 114 notice in 2023, though the primary driver was an unrelated equal-pay liability.
The sourcing lesson: a technology purchase treated as a transaction — sign, delegate, file — fails, because the "product" is a multi-year organizational change program the buyer must be staffed and governed to run.
The brewer contracted HCL to implement SAP on terms that assumed scope could be fixed up front. The program broke down over whether the blueprinting phase was complete; delay pushed deployment toward a peak-season blackout. MillerCoors sued for over $100 million alleging understaffing and failure to follow the contracted methodology; HCL countersued; the parties settled in 2018 without admissions.
The structural lesson is visible in the dispute itself, whichever account is right: fixed-scope contracting does not remove uncertainty — it converts it into blame allocation.
Hertz engaged Accenture to rebuild its website and mobile apps, terminated the engagement after missed launch dates, and sued to recover the roughly $32 million it had paid — alleging, among other things, that the delivered product lacked a tablet-optimized version and that remedying it was treated as chargeable additional scope. Accenture disputed the claims and the case settled — but the lesson holds either way: when requirements live in a contractual document rather than a working product, every gap becomes a commercial dispute instead of a design conversation.
Queensland Health's payroll replacement deserves mention alongside these: a total cost of ownership (TCO) estimated by KPMG at around AU$1.2 billion against an initial contract in the single-digit millions, severe enough to trigger a commission of inquiry.
Different industries, different vendors, different decades — and in each, a sourcing and contracting model built for well-specified physical deliverables was applied to open-ended digital capability, compounded by the governance and change-management failures the inquiries documented.
The root cause: value behaves differently in digital
Traditional procurement was never really about lowest price. But every criterion it weighs — price, delivery, quality, service, standing — is an attribute of the offer, knowable at the moment of evaluation. However sophisticated the weighting, the model rests on three assumptions about where value lives: that value is specified in advance, so requirements define what "good" is; that value is embedded in the deliverable, so a turbine that meets spec largely is the value; and that value is fixed at signature, so the best offer on award day remains the best value over the asset's life.
For physical goods these assumptions hold well enough, most of the time, for the model to succeed — though the megaproject record shows they are approximations even there. For enterprise technology, all three are materially weaker.
Value is substantially emergent
A meaningful share of a platform's eventual value tends to come from uses nobody wrote into the requirements, discovered once planners, operators, and analysts work with the system and its data. The reverse also holds: a system can satisfy every written requirement and create little value in operation — "compliant but not fit for purpose" is the characteristic failure mode of specification-driven technology buying. A specification is best treated as a hypothesis about value, not a definition of it.
Value is co-created rather than delivered
Most of a system's value is unrealized at handover. It materializes as people adopt it, processes change around it, and data flows through it — a joint production between buyer and vendor sustained over years. This is how Birmingham could take delivery of a major ERP and end up operationally worse off for a period: the deliverable arrived; the conditions for value did not. The delivery milestone changes meaning accordingly: in traditional procurement, acceptance largely concludes the relationship; in digital, go-live begins its most demanding phase.
Value is dynamic — it compounds or erodes
After signature, what you bought keeps moving: compounding through the vendor's roadmap, ecosystem, accumulating data, and each new integration; eroding through customization debt, renewal escalation, and deepening vendor lock-in. It includes option value — a platform is partly worth what it lets you do next, a component the AI cycle is enlarging. The offer that scored best on award day can be poor value by year five; Lidl's selected solution ultimately delivered less than the legacy system it replaced.
Where the traditional procurement playbook breaks
Set the traditional procurement moves against these digital realities one by one.
The exhaustive up-front specification
For a crusher, an exhaustive specification is engineering rigor; for enterprise software it is a fiction that ages quickly — some requirements are unknowable until users interact with the system, and they change across a multi-year delivery window. Hundreds of "shall" requirements invite paper compliance while real fit goes untested. And the dynamic runs both ways: a fixed-price tender against an incomplete specification forces rational vendors to price heavy contingency or bid low and recover margin through change orders — buyer-side contracting helps create the vendor behavior buyers then complain about. The alternative: outcome-based requirements, tested through scenario demonstrations and paid pilots before commitment at scale.
The award to the best-scoring compliant offer
License and subscription fees are commonly a minority — often cited at a third or less — of ten-year cost of ownership once implementation, integration, data migration, change management, upgrades, and exit are counted; Birmingham's cost trajectory shows where the larger sums sit. The value side is even less visible: adoption, fit, and roadmap appear in no bid table. An offer-anchored evaluation optimizes the only part of the economics it can see. The alternative: evaluate expected lifetime value — demonstrated fit, multi-year cost of ownership, vendor trajectory, exit cost.
The fixed-price, fixed-scope, big-bang contract
It converts inevitable learning into disputes and, occasionally, litigation — MillerCoors–HCL. The alternative: staged contracting, funding in phases with decision gates, retaining the option to stop or redirect. Capital-projects organizations already run stage-gated investment (FEL gates, FEED before EPC commitment); the adjustment is to apply the same staging to technology rather than treating the tender as one irreversible event.
Forcing the system to replicate existing processes
It breaks platform integrity and creates stranded customizations; it was central to Lidl's failure and, per the auditors, to Birmingham's. The alternative: platform standards by default, customization reserved for genuine differentiators — with governance strong enough to hold that line under internal pressure, because in both cases the principle was known and overridden.
Treating the purchase as a transaction that concludes
Cloud and SaaS turn technology from a capital asset into an operating relationship: subscriptions that evolve continuously and never become something you own, whose value leaks through unused licenses, renewal escalation, and scope drift. Gartner's SaaS vendor-management research warns that these contracts deliver value only under ongoing contract, performance, and risk management, and that unmanaged SaaS and AI sprawl drives overspend and loss of visibility. The alternative: a standing vendor-management discipline covering performance, risk, and renewals for the life of each platform.
Adversarial negotiation on margin
It weakens a counterparty whose goodwill, staffing, and roadmap you will depend on for a decade — and whom you cannot easily leave. Replacing an ERP is a multi-year program, as Birmingham's own SAP-to-Oracle transition demonstrates; once data, processes, and integrations live in a platform, the vendor holds structural leverage for years. The alternative: negotiate hard on the terms that matter — data ownership, exit assistance, price caps, service levels, the core argument of Tollen's Tech Contracts Handbook — then operate a governed partnership.
The relay and the piecemeal buy
A process in which procurement leads, IT advises, and the integrator delivers leaves the real success factors — fit, architecture, adoption, governance — owned by no one, a pattern visible in both the Birmingham and Lidl records. And buying systems one by one on individual merit produces silos and integration debt, because value comes from the integrated stack. The alternatives — architecture first, cross-functional ownership, distributed decision rights — are operating-model changes, not tender tweaks.
The required shift in technology procurement
Add the clock-speed problem — an eighteen-month tender cycle is prudent for major plant but risks outlasting the assumptions it was based on, especially as generative AI shifts the question from "which tool do we buy?" to "which capabilities should we own — build, compose, collaborate with a provider, or buy as outcomes?" (Nishar & Nohria, HBR, April 2026) — and the conclusion is hard to avoid. You are buying a capability, not a product. As Dovgalenko's Technology Procurement Handbook frames it, technology procurement is the management of a demand pipeline and a vendor lifecycle, not the execution of transactions. The required shift is from selecting the best compliant offer to maximizing expected lifetime value under uncertainty — and it reaches well beyond the tender, into how vendors are selected, and into who holds decision rights over the award and how the buying organization itself is governed.
- Lidl/SAP: Handelsblatt reporting on the termination of the eLWIS program (2018) — the source of the ~€500M investment estimate; Lidl has not published a figure.
- Birmingham/Oracle: Grant Thornton, external auditor's report on the Oracle implementation, Birmingham City Council (February 2025); council audit-committee reporting (January 2026) for the £144M program forecast.
- MillerCoors LLC v. HCL Technologies Ltd., U.S. District Court, N.D. Illinois (filed 2017; settled 2018 without admissions).
- Hertz Global Holdings v. Accenture LLP, U.S. District Court, S.D.N.Y. (filed 2019; settled without admissions).
- Queensland Health Payroll System Commission of Inquiry, Report (2013); total cost-of-ownership estimate ~AU$1.2B (KPMG).
- Peter Smith, Bad Buying (Penguin Business, 2020)
- Bent Flyvbjerg & Dan Gardner, How Big Things Get Done (Crown Currency, 2023) — base rates, fat-tailed risk, and strategic misrepresentation across project types, IT among the worst
- Sergii Dovgalenko, The Technology Procurement Handbook (Kogan Page, 2020)
- David W. Tollen, The Tech Contracts Handbook, 3rd ed. (ABA)
- Deep Nishar & Nitin Nohria, "The End of One-Size-Fits-All Enterprise Software," HBR, April 2026
- On failure-rate estimates: Standish Group CHAOS reports and the McKinsey–Oxford large-IT-program studies; for methodological critique, Eveleens & Verhoef, "The Rise and Fall of the Chaos Report Figures," IEEE Software (2010)
Technology executive — enterprise AI, automation & digital strategy for enterprises, NGOs, and UN agencies. 15+ years, primarily across Africa's emerging markets.