What went wrong in four public SAP projects, why it happened, and what a CIO or CFO can do to lower the risk.
Analysis | ~15 min read | By J. Torre
Most SAP projects do not end in court or in the press. The SAP project failures that do reach court or the press show what goes wrong, and that makes them useful. In this article we look at four public cases: National Grid US, Revlon, Haribo and Zimmer Biomet. We analyze the facts and allegations, we follow the causes that repeat from one case to the next, and we then set out what a CIO or CFO can do to lower the risk.
Our main point is that in these cases the main causes were not the software itself. They sit in scope, governance, data, testing, people and timing, and these are areas where management can act. That is also why the risk can be reduced with methods that already exist. This article builds on earlier ones in this series: on the business case for S/4HANA that the board approves, and on the transition paths to S/4HANA that define the scope of a project.
How we read these cases
The evidence differs from case to case, so we say which kind we are using. A fact is documented when it comes from a company filing (for example the 10-K, the annual report that U.S.-listed companies file with the SEC, the U.S. securities regulator), a court decision or a statement by a company executive. It is alleged when it comes from a lawsuit, which is one party’s version and has not been proven. It is opinion when it comes from a consultancy or the press. Two of the four projects are S/4HANA (Haribo and Zimmer Biomet). National Grid and Revlon used the earlier SAP ERP. We include them because their root causes are the ones we see in S/4HANA programs today.
We present the cases in chronological order, because each one adds something to the one before.
National Grid US and Wipro: choosing and checking the integrator
National Grid US started in 2008 to replace its back-office systems with SAP, covering finance, HR, supply chain and procurement. Deloitte was the first partner. In 2010 Ernst & Young became project manager and lead integrator, and Wipro became systems integrator, the firm that carried out the implementation until go-live. Payments to Wipro reached about $140 million. National Grid sued only Wipro. Wipro says it joined in 2010 and that several vendors worked on the project.
According to the lawsuit National Grid filed in 2017, the system went live with serious errors. More than 15,000 supplier invoices were unpaid two months after go-live, and the financial close went from 4 days to 43. The lawsuit says stabilization cost about $30 million a month and took more than two years, and claims over $300 million in total. The case settled in August 2018 for $75 million, with no admission of liability.
The causes are mostly allegations, and Wipro rejects them. National Grid said Wipro overstated its experience, had almost none with SAP at a regulated U.S. utility, and assigned inexperienced consultants. It also said the design was too complex and that the system was customized too much instead of using standard configuration. Wipro answered that the claims misstate the facts and that National Grid had given it positive evaluations and an award in 2014. A 2014 audit, as reported by The Register, points in a similar direction. According to that report, the audit said National Grid was not ready, had chosen an integrator with little U.S. utility experience, and had spoken with few other utilities running SAP.
If we take both sides together, the case is about choosing and checking the partner. Verifiable references from comparable projects, standard configuration by default with each customization approved on a business case, and tests of financial close and payroll before go-live would each have addressed part of the problem. They also depend on the customer being ready to do that checking, which leads to the next case.
Revlon: a go-live that went ahead despite known problems
Revlon wanted to replace 21 ERP systems in 50 entities with one SAP system. It reused the template of a subsidiary it had acquired. The largest site was the factory in Oxford, North Carolina, and the go-live was postponed four times because of problems there. The company went live at Oxford on 1 February 2018 anyway. Warehouse and production did not integrate well, and training was not sufficient.
Here the evidence is better, because the facts come from Revlon’s own filings. In March 2019 Revlon delayed its annual report because of a material weakness in internal control, and its share price fell 6.87%. The 2018 10-K (Revlon’s annual report to the SEC) reports $64 million in lost net sales and $53.6 million in additional charges. The 10-K also says that the company had not run a continuous risk assessment that identified the effect of SAP on internal control, and that it had not kept enough trained staff in U.S. operations. Shareholders added allegations: the template came from a different business, the team had no experience in retail or with such a large plant, and the go-live went ahead despite test problems.
The shareholder class action (Lachman v. Revlon) was dismissed in September 2020. The court found that Revlon had warned of the risks before launch and acknowledged the damage afterward. This is a ruling on disclosure to the market, not on project quality.
Compared with National Grid, the problem here is less about the integrator and more about decisions inside the company. The four postponements show that the problems were known. The filings show no continuous assessment of the effect on internal controls, and the court record does not mention go/no-go criteria that made the date depend on test results. A template proven in a comparable environment, training measured by role before cutover, and go/no-go criteria agreed in advance are the controls that address this.
Haribo: scope, data and timing
Haribo’s “One Haribo” project replaced decentralized ERP systems, including a merchandise system almost 30 years old, with a single S/4HANA instance. It was the largest IT project in the company’s history, in a group with 16 factories in 10 countries and sales companies in 26. The project ran at the same time as a headquarters move, a new factory, a new logistics center and a new managing director in Germany.
The merchandise system changed in October 2018, a few weeks before the peak season. In December 2018 the press reported supply problems in practically all products. The managing director said in 2019 that failed deliveries reached ten percent or more per month until February 2019. In an interview with Handelsblatt in November 2019 (as cited by IT-Zoom), he also put the cost at a three-digit million euro amount. This is the strongest cost figure of the four cases because it comes from the company, although it is not itemized or audited. Haribo’s sales decline is not attributed to the project. Haribo’s management points to earlier decisions, and sales were already falling before go-live.
Haribo’s management has also been open about the causes. It said the complexity was underestimated and that the scope was the widest possible, covering all processes from raw material to finished product. It said master data was a major problem: with more than 1,500 items per factory, whole items would not run on the line because of problems in the master data. And it said too much changed at once. Third-party commentators add insufficient testing and a go-live close to the peak season.
This case shows two things that the previous two do not. The first is that the way the project is divided matters. Splitting the scope by process and by site, treating master data quality as a milestone with an acceptance criterion, and rehearsing the cutover with real data and full item coverage would have reduced the size of each step. The second is timing. A go-live in the weeks before the peak season, at the same time as other corporate changes, leaves no room to absorb errors.
Zimmer Biomet and Deloitte: contract, change control and independent review
Zimmer Biomet is the most recent case and the only one with litigation still open. Zimmer and Biomet merged in 2015 and ran several ERP systems. Deloitte did a ten-month analysis and then an S/4HANA program for North America and Latin America, under a 2021 work order of $69 million.
Everything that follows is from Zimmer’s lawsuit of September 2025 and is alleged. The go-live date moved five times, there were 51 change orders, and the work order grew to $94 million. The system went live on 4 July 2024, and Zimmer says it was barely able to operate in the third quarter: it could not ship or receive product, issue invoices or produce basic sales reports. Zimmer claims at least $172 million: $94 million in fees, $15 million billed during fix attempts and $72 million in its own costs after go-live. Its 2024 10-K reports an impact on 2024 sales of less than 1%.
Zimmer alleges that Deloitte overstated what it could do and promised little customization, that the team lacked experience, that the warehouse module was defective, and that Deloitte advised going live without warning of critical risks. Deloitte says Zimmer did not give notice of breach within the contract period, described the project as a success for years, and caused part of the problems itself. A motion to dismiss is pending. UpperEdge, a buyer-side advisory firm, adds an opinion based on the lawsuit: a direct award without competition, few acceptance criteria and payment holdbacks, and a go-live that coincided with a restructuring, a change of CEO and later a CIO departure.
Whatever the court decides, the case shows how contract and governance work together. Acceptance criteria and payment holdbacks tied to milestones, a limit on change orders, and an independent readiness review before go-live, rather than relying only on the integrator’s view, give the customer information and leverage when it needs them. It also repeats the lesson from Haribo about timing: go-live should not overlap with other major corporate changes.
What the SAP project failures have in common
Read together, the four cases repeat a short list of causes.
The first is the partner and the team. In National Grid and Zimmer the customer says the integrator lacked the right experience, and in Revlon the allegation is about the team and the template. These are mostly allegations, but they all point to the same weak spot: the experience of the partner and the team was a risk the customer needed to check.
The second is the go-live decision. Revlon postponed four times and went live anyway, Zimmer moved its date five times, and National Grid went live with errors in financial processes. In these cases the date seems to have carried more weight than readiness. Haribo chose a date close to the peak season.
The third is scope and change. Zimmer had 51 change orders, National Grid is alleged to have customized the system heavily, and Haribo covered every process at once. When scope grows or is too wide, testing, data and training all come under more pressure.
The fourth is data and testing. Haribo’s master data errors and National Grid’s incorrect financial calculations are cases where the problem could have been found before go-live by tests with real data and business participation.
The fifth is the customer’s own readiness: training and trained staff (Revlon), internal controls (Revlon), and the ability to oversee the integrator (National Grid, Zimmer).
One more point is about reading the results. Not every loss after go-live belongs to the project. Haribo’s falling sales had started earlier, and the Revlon ruling is about disclosure, not about project quality. A fair analysis separates what the project caused from what happened around it.
Zimmer alleges a defective warehouse module, but apart from that, none of these causes is a software defect. Each is a management decision or a missing control, which is why they can be managed.
How to reduce the risk
A method does not guarantee success, but it removes most of the paths to SAP project failures seen above. We group what we recommend into the places where the four cases went wrong.
Governance and business involvement
In each case above, governance played a part. The steering committee should include the business, not only IT. The CFO and the heads of operations, supply chain and sales should own the decisions on scope, process design and go-live, because they are the ones who live with the result. Underneath the committee, each business process needs an owner with time assigned. Users should take part by role from design to testing, and business users, not only IT, should run the user acceptance tests. SAP Activate proposes a business product owner for the backlog in the build phase, which does the same thing in an agile setting.
Sponsorship needs to be stable. Where the sponsor or the leadership changes during the project, as around Zimmer and at Haribo, the mandate should be confirmed again.
Decision rights and staffing matter as much as structure. Horváth’s 2025 study of 200 companies found lack of decisions, scope expansion and weak project management among the main causes of delay. On average, projects ran 30% longer than planned and fewer than 10% finished on schedule. In the same study, 75% of participants said the staffing of project management and the team had been neglected.
Continuous communication
Communication should follow a fixed schedule, with progress, risks and decisions reported to the steering committee and to the affected business units. A single status format with objective thresholds (red, amber, green with a defined meaning) brings bad news early, when there is still time to act. And what the board and investors are told should match what the project team reports.
Scope, design and change control
Fit-to-standard workshops, as used in SAP Activate, are the tool to decide what to adopt as standard and what to change. Each customization needs a business case and an owner. Our series on Clean Core explains how to keep the standard core and place extensions where they do not block upgrades. Change orders should be limited and each one approved by the steering committee, with its effect on cost and date stated. Phasing is the other protection: a first release with fewer processes or sites reduces both the risk and the cost of a wrong decision (we compare the options in our article on transition paths to S/4HANA), which is the opposite of the “all processes at once” approach at Haribo.
Data, testing and cutover
Data migration should be its own workstream with an owner (see also our article on the data trust gap in SAP finance). Master data quality gates should be a condition before testing and cutover, and several data loads should be reconciled with the business, not only with IT. Full integration tests and user acceptance tests should cover the processes with the highest risk: financial close, payroll, invoicing, shipping and receiving. The cutover should be rehearsed with realistic data.
SAP’s own cutover readiness framework describes the approach that fits the Revlon and Haribo cases. It sets predefined criteria, regular checkpoints and monitoring from an early stage, so that the go/no-go decision does not depend on the pressure of the last moment.
Go/no-go decision, timing and independent review
Go/no-go criteria should be defined in advance, with decision makers. If the criteria are not met, the date moves. Peak business seasons and other corporate events, such as restructurings, moves or leadership changes, should be avoided. Hypercare after go-live needs enough staff, with the integrator still involved.
An independent health check at key points, for example before the build phase and before go-live, gives the steering committee a view that does not come from the integrator. The reviewer should not be paid by the integrator. An assessment of the effect on internal controls before go-live addresses the Revlon outcome directly.
People and change management
Training and adoption should be a workstream with budget, schedule and metrics. The measure to track is how many users, by role and by site, are trained before cutover, and whether enough trained staff will be present on the first day.
Contracts and partners
Before signing, ask for references from projects of similar size and sector, and speak with those customers. Check also who will actually staff the project, since experienced SAP people are scarce (see SAP talent in 2026). Write acceptance criteria into the contract, tie part of the payment to them, and keep the right to review the composition of the integrator’s team. Compete the selection where possible, because a direct award removes the comparison of price and quality.
The role of the methodology
SAP Activate covers much of this. It has six phases (discover, prepare, explore, realize, deploy, run), fit-to-standard workshops, an agile build with a business product owner, full integration and user acceptance tests before deploy, and a hypercare period. It has workstreams for project management, data management, solution adoption, testing and extensibility. A methodology on its own does not prevent failures. What matters is that the company applies it and that governance enforces it, especially at the decisions where the cases above went wrong.
A short checklist for the CIO and the CFO
Before approving a go-live, ask:
- Who owns each business process in the new system, and have they signed off?
- Did the master data pass its quality gate, and who confirmed it?
- Were the financial close, payroll and end-to-end processes tested with real data?
- Was the cutover rehearsed, and what was the result?
- Are the go/no-go criteria met, or are we moving the date by decision?
- How many users, by role, are trained?
- Did someone independent review readiness and internal controls?
- Is the go-live away from peak season and from other major changes?
- What do we tell the board and investors, and does it match what the project team reports?
Conclusion
The four SAP project failures differ in size, sector and outcome, but the causes are close to each other: a partner whose experience was not verified, go-live decisions driven by the calendar, scope that was too wide or kept changing, data and tests that did not stop a faulty go-live, and a business that was not fully involved. These can be managed. A company that applies business ownership, clear governance, regular communication, objective go/no-go criteria and independent review reduces the risk a lot, though it cannot remove it.
Sources and reliability
- National Grid: The Register (6 Aug 2018), TechTarget (Dec 2017), Wipro filing with the Bombay Stock Exchange.
- Revlon: Lachman v. Revlon, E.D.N.Y. 19-cv-2859 (as published by Simpson Thacher); Revlon Form 10-K FY2018 (SEC); Reuters.
- Haribo: IT-Zoom (15 Dec 2018, 27 Nov 2019, summarizing a Handelsblatt interview), Tagesspiegel/dpa (14 Dec 2018), ComputerWeekly.
- Zimmer Biomet: lawsuit as reported by the Journal Gazette; Deloitte’s motion as reported by MassDevice; Zimmer Biomet Form 10-K (SEC); UpperEdge (opinion); Loeb & Loeb press release (law firm for Zimmer).
- Methodology: SAP Learning, “Describing the Methodology Structure” (SAP Activate); SAP Community blog, “SAP Project Cutover Readiness Assessment” (2021); Horváth, S/4HANA transformation study, 2025.
