Governance

Cockcroft’s Folly

Ten years of looking wrong

Take any decision that matters and follow it along its chain:

1. Governancewho decides
2. Riskwhat could go wrong, and what it would cost
3. Compliancewhat must be true
4. Resiliencewhat all three amount to when the thing actually happens

So work a real one through. This example is unlike the others here in one respect that matters: it is a link that held. It held late, it held expensively, and the man it belonged to was laughed at for a decade for making it hold.

The lumps on the chimneys

In the late 1940s Britain was building two air-cooled reactors at Windscale, in Cumberland, to make plutonium. They were a national priority, they were behind schedule, and their chimneys were already designed.

John Cockcroft, director of the atomic energy research establishment at Harwell, had read a report that uranium oxide had been found on the ground near a graphite reactor at Oak Ridge. He concluded that the Windscale stacks needed filters.

The stacks were already designed. Putting filters in at that stage meant putting them at the top — hundreds of feet up, where they had to be hauled and housed, at great cost, as pronounced lumps on the skyline of a plant that had no budget for ornament.

They were built. And they were named, as such things are named, by the people who had to pay for them and look at them: Cockcroft’s Folly.

October 1957

The piles required periodic annealing — a controlled heating to release energy stored in the graphite. In October 1957, during one of these operations, Pile No. 1 did not behave. By the tenth it was on fire, and it burned for days.

Radioactive material went up the chimney. What came out of the top of it went through Cockcroft’s lumps, which captured most of it.

What got past was still enough to matter. Iodine-131 settled on pasture, pasture was eaten by dairy cattle, and milk from an area of some five hundred square kilometres was banned and poured away for weeks. That was the residue. The assessment since has been that without the filters a large part of Cumbria would have been contaminated.

What the story is actually about

It is tempting to read this as a story about one clever man, and that reading is useless to anybody. Cleverness is not transferable and cannot be designed into an organisation.

What can be designed is the answer to a narrower question: on that programme, at that moment, could the person who understood the hazard impose a cost on the people who owned the schedule?

On this occasion, yes. Not because the argument was won on its merits in any visible way — it plainly was not, or the thing would not have been called a folly for ten years — but because of where Cockcroft stood and what his standing allowed him to insist upon. The filters were not built because everyone agreed. They were built because one person who could not be ignored declined to stop saying it.

That is a governance fact, not a technical one. And it is the fact that does not survive an organisation chart being redrawn.

The asymmetry that never goes away

Notice the shape of what was being asked. The cost of the filters was known, immediate and itemised. The benefit was unknown, deferred, and probabilistic — and, if the programme ran well, would never be visible at all.

That asymmetry is not a bias to be trained out of managers. It is the permanent condition of every preventive control ever proposed. Which means a governance structure cannot rely on individuals being brave enough to lose that argument repeatedly. It has to carry the asymmetry itself — by putting the decision somewhere that the schedule does not own.

And there is a corollary worth stating plainly, because it is the part that makes people uncomfortable: most good governance is never vindicated. Cockcroft got a fire. The overwhelming majority of sound, expensive, unpopular precautions are never followed by the event that would have justified them, and their authors retire with the nickname and without the vindication. A structure that only protects the ones who turn out to be right is not protecting anybody, because the protection has to arrive years before the answer does.

Every organisation has its own lumps on the chimney — the control everyone agrees is disproportionate, the check nobody can see the point of, the cost that is obviously too high for the thing it is guarding against. Some of them genuinely are. The question is not whether you have any. It is whether you would be able to tell, and whether the person who can tell is in a position to make you pay for it.

The keep

• Governance is not the signature. It is the distribution of who can stop things, and what stopping them is allowed to cost.
• Being right early is indistinguishable from being wrong, for exactly as long as the event takes to arrive. Structures that cannot tolerate that stop producing people who are right early.
• Cost is known, immediate and itemised; benefit is unknown, deferred and invisible. The structure has to carry that asymmetry, because no individual can carry it indefinitely.
• Most good governance is never vindicated. Judge the arrangement, not the outcome — the outcome is mostly weather.
• The person who knows and the person who pays are rarely the same person. Governance is the standing arrangement between them, written down before anyone knows who will need it.

First move: find the last time someone in your organisation stopped or materially changed something on technical grounds, against a cost or a deadline. Name the person and the date. If you cannot think of one in the past year, that is the finding — and it is not that nobody objected. It is that nobody’s objection was permitted to be expensive.

Sources

Cockcroft’s insistence on the filters, the Oak Ridge report that prompted it, their placement at the top of the already-designed stacks and the nickname that followed: John Cockcroft and Windscale fire. The annealing operation, the fire and the filters’ effect: Britannica, Windscale fire. The milk ban over some five hundred square kilometres, and iodine-131 as the contaminant of concern: radioactivity.eu.com, Windscale accident. Checked 20 September 2026. The counterfactual — how much worse it would have been without the filters — is an assessment made after the fact, and is reported here as such rather than as measurement.

‹ The Second Opinion