National Grid v. Wipro: A $585M Cleanup

Of the public IT-project disputes, National Grid v. Wipro is among the most instructive, because the failure mode is so vivid: a new SAP system went live, and almost immediately it paid people the wrong amounts and tangled the supply chain. The cleanup reportedly cost more than the project did. It is a textbook illustration of risk arriving all at once, at the moment of cutover.

What follows is drawn from the 2017 complaint and contemporaneous reporting. It describes a reported legal matter — not a finding of fault, and not legal advice. National Grid made the allegations, the case was resolved by settlement, and there was no public liability judgment.

What the project was supposed to be

National Grid’s U.S. arm set out to replace its core ERP with a new SAP platform spanning payroll, supply chain, and finance — the operational backbone of a large electric and gas utility. The system went live in November 2012. Wipro served as the systems integrator on the program.

What reportedly went wrong

The trouble surfaced at go-live and compounded fast. According to later reporting and the complaint:

• Payroll miscalculated. The new system reportedly paid employees incorrectly — some too much, some too little, some not at all — for time, pay rates, and reimbursements.
• The supply chain seized up. Within weeks of go-live, National Grid’s backlog of unpaid supplier invoices reportedly ran into the tens of thousands, and inventory record-keeping was described as in disarray.
• The damage was operational and immediate. These weren’t latent quality issues; they hit real money and real people the moment the system carried live work.

National Grid’s 2017 suit, filed in the U.S. District Court for the Eastern District of New York, alleged the implementation delivered little usable value for what it paid.

The numbers, stated carefully

Three figures matter, and they should be kept distinct:

Figure	What it is	Confidence
~$140M	Fees National Grid reportedly paid Wipro, and sought to recover	Reported in filings/press
~$585M	Widely reported cost of the multi-year cleanup that followed	Reported cleanup cost, not a judgment
~$75M	Amount Wipro reportedly paid to settle, 2018	Reported settlement, no liability finding

The ~$585 million headline is the cost of remediation — the multi-year effort to fix payroll, supply chain, and finance after the troubled go-live — not the original build. Reporting described hundreds of contractors working over more than two years at a monthly cost in the tens of millions. Different outlets frame the total slightly differently; IEEE Spectrum, for example, put the additional spend near $600 million on top of the original development cost. We cite roughly $585 million as a reported cleanup cost, not an adjudicated damages number — the case settled, reportedly for about $75 million from Wipro in 2018, with no public finding of fault.

Why the failure surfaced only at cutover

This is the part worth sitting with. A payroll and supply-chain system can look complete in testing and still pay people the wrong amount in production, because the real test is real data at real volume — every pay rule, every union agreement, every supplier term, every edge case the test environment didn’t capture. In a big-bang cutover, all of that meets the new system on the same day, with the old system already switched off. There is no partial fallback. The first time the system is truly exercised is the first time the business depends on it — and any gap becomes an incident, not a finding.

That is the core argument we make in big-bang vs. incremental: the all-at-once switch doesn’t just risk failure, it guarantees that any failure will be maximally disruptive and minimally recoverable.

How we’d shape a cutover like this

A payroll-and-supply-chain migration is exactly the kind of system that must not be cut over blind. We work through it slice by slice behind a strangler facade, so the existing system keeps paying people and processing invoices while the new one is proven alongside it. Before any slice carries live traffic, we reconcile its output record-for-record against the legacy — same pay calculations, same invoice postings, same inventory movements, on real data — and we shadow live traffic through the new path without it affecting anyone until it matches. Traffic shifts only on green, and rollback stays a flag away.

In that model, a payroll miscalculation shows up as a failed reconciliation on a single slice in a test, weeks before it could ever reach a paycheck — not as a $585 million cleanup. Where the concern is the exposure an at-risk system already carries before any project starts, a legacy system liability assessment names it.

Not a ruling on who was at fault

We’re not adjudicating fault here — the case settled without a public liability finding, and Wipro is a major integrator that delivers large programs successfully. Nor were we involved in this matter in any way; we weren’t. It stands as a public, well-documented example of how a big-bang go-live concentrates and hides risk until the worst possible moment. To scope a reversible, parity-proven cutover for a system you can’t afford to break, book a discovery call at /meet or reach the team at sales@modernlift.ai.

This guide summarizes a reported legal matter for illustration. It is not legal advice, and any judgment about the parties’ conduct belongs to the courts and counsel.