If you've been following the multi-million pound rollout of Palantir inside Britain's national health service, you've probably heard the headline numbers. Ministers and health bosses love quoting them: 110,000 additional hospital operations and a 15 percent drop in discharge delays. They painted a picture of a revolutionary software suite sweeping through hospital wards and fixing systemic backlogs overnight.
There's just one problem. The statistics watchdog just stepped in, and those impressive figures don't actually prove what the health service said they did.
Following an investigation by the UK Office for Statistics Regulation (OSR), NHS England had to make a major retreat. The watchdog forced the health service to publicly qualify its claims about the software's impact, update its public-facing material, and admit a basic truth about data: correlation isn't causation.
Here is what really happened behind the scenes, why the data crumbled under scrutiny, and what this means for the future of patient data in the UK.
The Problem With Flawed Statistics
When NHS England handed US tech company Palantir a £330 million contract in late 2023 to build the Federated Data Platform (FDP), the backlash was immediate. Critics pointed to Palantir's background in defense contracting, its American corporate leadership, and privacy concerns regarding patient records. To win over skeptical doctors and the public, officials needed hard proof that the platform delivered real value.
That proof came in the form of sweeping national metrics. Health officials claimed trusts using the new platform saw a surge in elective procedures and far fewer delayed discharges for long-stay patients.
Yet when analysts look closer at how health systems operate, those claims don't hold water.
A hospital might perform more operations in a given year because it hired more surgeons, opened extra weekend wards, or simply recovered from post-pandemic backlogs. Crediting a single piece of software for those extra surgeries without controlling for any other factor breaks basic statistical rules.
In methodology notes buried on its website, NHS England quietly added a disclaimer acknowledging it could not "draw conclusions about cause and effect as other variables have not been controlled for." Basically, they admitted that while these improvements happened after the software went live, there's no proof the software actually caused them.
The Watchdog Steps In
That quiet admission wasn't enough for the statistics regulator.
The OSR launched a probe after complaints surfaced about how these figures were being presented to the public and used by ministers in parliament. The watchdog's ruling was clear: NHS England must make its caveats prominent rather than burying them in footnotes.
Going forward, the health service must:
- Explicitly state on its main webpages that its headline figures don't prove causation.
- Ensure ministers and media teams stop using unverified figures in official briefings.
- Clearly label case studies to show that NHS England hasn't independently verified or authored them.
- Publish broken-down trust-level data so independent researchers can verify performance claims.
To make matters worse, several hospital trusts admitted to errors in the underlying data used to calculate discharge delay reductions after independent media investigations uncovered irregularities across dozens of trust filings.
Is the NHS Trapped in Bad Data Contracts?
The situation highlights a massive issue in how public sector organizations purchase and evaluate technology. When government bodies spend hundreds of millions of pounds on external software, they face immense political pressure to demonstrate immediate success. That pressure often leads to cherry-picked data and premature victory laps.
We've seen this play out before, but the scale here is huge. The NHS isn't just buying off-the-shelf administration software; it's handing over the central nervous system of patient operational data to a third-party vendor.
Contrast this with regions like Greater Manchester, where local health leaders opted out of the national FDP rollout entirely. Instead of adopting the Palantir platform, they built their own data tools using local NHS talent, British universities, and domestic tech partners. Early reports suggest those homegrown systems perform exceptionally well without incurring identical costs or privacy disputes.
When local, open approaches yield solid results while national rollouts rely on disputed stats to justify their budget, leaders need to ask tough questions.
What Happens Next
The OSR investigation is officially closed for now, but the political fallout is just getting started.
Ministers face a huge milestone with a contract break clause arriving in February 2027. Campaign groups like Foxglove are already pushing the government to pull the plug or order a complete independent audit of the platform's actual impact.
If you run operations, manage data, or oversee procurement in healthcare or the public sector, this entire situation offers three vital takeaways:
- Demand raw, disaggregated data. Never accept vendor-supplied performance figures at face value. Demand raw trust-level data that can be independently audited and cross-referenced with control groups.
- Separate correlation from cause. Baseline operational changes, staffing shifts, and seasonal swings account for massive variations in throughput. If your analytical team doesn't control for external variables, your reporting won't hold up under public scrutiny.
- Build transparent governance early. If you bury methodological caveats in obscure footers, watchdog groups and the public will eventually find them. Transparency at the start prevents reputational damage down the road.
The health service has pledged to fix its public messaging and involve its professional statistical teams more closely in future media releases. Whether that restores public trust in its digital strategy remains to be seen.
