Zimmer Biomet has filed a $172 million lawsuit against its system integrator over a SAP S/4HANA program that went live and then could not ship product. That figure changes what "stalled implementation" means on a board slide. A project two quarters late and 30 percent over budget is no longer a project-management anxiety. It is a disclosure risk, a covenant risk, and in that case, an allegation of fraud. The right question stops being whether to keep going. It becomes which of three doors to walk through, and what each one costs.
The Four Failure Signatures
Before any decision, name the failure. Stalled NetSuite implementations split cleanly into four signatures, and the triage path depends on which one is live.
Scope Drift
Scope drift shows up as a change-order ledger that grows faster than progress. A fixed-fee statement of work balloons through amendments, each one small enough to approve and large enough to compound. In the Zimmer Biomet complaint, the original work order ran $69M and reached $94M across 51 change orders, a scope that was described as "minimal customization" at signing. If the integrator is sending monthly change orders and the steering committee is still approving them, that is not a schedule problem. It is a sizing problem that was never reopened.
Data That Will Not Reconcile
Data is the quietest failure and the most expensive one. Chart-of-accounts mismatches, missing historical journals, and item-master duplicates do not surface in unit testing. They surface in parallel run. By then the cost of a chart rebuild is counted in quarters, not weeks. Experienced teams expect this; the fact that bad data is always discovered during the load is worth internalizing before the discovery.
SuiteScript That Nobody Owns
A NetSuite implementation that has accumulated hundreds of SuiteScript customizations, workflow extensions, and SuiteBundler installs will stall at UAT. The partner cannot regression-test what they did not write; the client cannot audit what they do not understand. The warning sign is a defect backlog that stops going down for two consecutive sprints.
A Partner Who Cannot Finish
Not every rescue is the client's fault. Published post-mortems on Nike, MillerCoors, Revlon, and Israel Chemical trace the failure back to a mismatch between the integrator's core competency and what the engagement actually required.
The Triage Decision
Before choosing rescue, re-scope, or pivot, run a two-week diagnostic that produces four artifacts: a live defect inventory by severity, a data-quality report from the proposed cutover extract, a list of customizations with owners and test coverage, and a steering-committee vote on which three business processes the system must support on day one. The point of the exercise is to force a binary. If the system cannot post a clean trial balance for a single legal entity in a sandbox, the project is not in a rescue. It is in a re-scope or a pivot.

Benefits realization is the external benchmark the audit committee cares about. If tracking is under half of the business case, the partner will propose extending the timeline. The finance leader's counter-question should be whether an extension buys measurable movement on the defect burn-down, or whether it just resets the invoice.
Path One: Targeted Rescue
Rescue is the right path when the four diagnostic artifacts show a clean trial balance in sandbox, a bounded customization inventory, and a partner that can be replaced without losing the data migration work already done. The headline cost is the one most teams underestimate.
A rescue typically needs a replacement partner, a hard freeze on new scope, a rebuild of the integration layer, and a rewritten cutover plan with a documented abort trigger. On the governance side, a rollback condition is where governance starts, because a go-live date without one is often why the first attempt failed. The audit committee briefing for this path leads with the trigger, not the date.
Path Two: Re-scope and Phase
Re-scope is the right path when the platform fit is broadly correct, the data is salvageable, and the failure is one of ambition. The move is to collapse day-one scope to the general ledger, AP, and a single subsidiary close, then defer revenue recognition, allocations, inter-company, and management reporting to a phase two that has its own budget and steering committee. Coverage of a 2024 Panorama study found roughly a third of enterprise software projects were completed late and over budget, and most had attempted to deliver everything in a single wave. A phased plan replaces a heroic go-live with a credible promise that the next milestone will land where stated.
Path Three: Pivot to an Agentic Platform
Pivot is the right path when the diagnostic shows the platform itself cannot support the day-one process without customization that re-creates the original debt. Veronica's Auto Insurance paid over $111,000 under a 2021 NetSuite services agreement before abandoning the system in 2024 over missing functionality, failed data migration, and non-functional vendor payment processing. Paying more does not fix the shape of a mismatch.
An agentic pivot narrows the surface area rather than widening it. Role-based agents run AP, AR, and controller workflows on top of a deterministic ledger they never write to directly, with reconciliation produced as a daily artifact rather than a month-end project. The ERP migration with reconciliation proof approach and the controller-side math on what NetSuite costs to keep versus replace are the two references to work through before sequencing a mid-flight switch.
What to Watch for the Next 90 Days
The signal to watch is the high-severity defect burn-down against change-order volume. If severity-one defects are closing at a rate that would clear the queue before the next steering-committee meeting, the chosen path is working. The number that should worry a controller is a severity-one queue that holds steady for two consecutive weeks while change-order fees accelerate. That combination is how the first attempt stalled, and it is how a rescue, re-scope, or pivot goes the same way if nothing about the governance has changed.
