In Abundance, Ezra Klein and Derek Thompson talk about the cost per mile of highway and railway in the US as a metric of how hard it is for the state to build national infrastructure. It’s disproportionately more expensive to build road and railway there than in the rest of the world. Organisational complexity and outsourcing destroy productivity. The book gives an alternative example – the Madrid Metro was built for a fraction of the cost of US equivalents at least in part because a small team of state‑employed engineers ran the project themselves, instead of the whole thing being outsourced.
It made me wonder if there’s a comparable metric for the way we ship services in government. Something that would confirm or disprove my feelings that NHSE is slowed by outsourcing and blocked by process in a way that other departments aren’t. I had a dumb idea that you could give a cost per kb or something. So, given AI, I set a research project for Claude. It went through GitHub, audited accounts, contract registers and usage stats for NHSE, GDS and a Norwegian equivalent. The tl;dr is that there’s not enough openly published data about spend and many hidden repos. Sad face – but not a surprise.
Claude found a study of the Norwegian department that radically in‑sourced everything, showing that in‑sourcing was no cheaper but faster and better quality. Getting Claude to do the research made me wonder if there was new potential for open, accountable government.
Supply
The Performance Platform was an early part of GDS. The theory was that if government published how its services performed, everyone would benefit. Ministers would see what things cost. Teams would compare themselves and improve. And the public, an “army of armchair auditors” would hold government to account.
GDS attempted to mandate supply of open data.
Demand
I worked on Performance Platform for a few months in 2014. There were massive gaps in the data (departments rarely, if ever, published), and metrics that everyone agreed were admirable in principle but nobody found practical value in. There were very few examples of anyone doing anything with the data. Some supply, but as far as anyone could tell us, no demand. The departments themselves didn’t use it, because it wasn’t detailed enough to be useful to them. Other than some internet‑famous bloggers, no one was using it. No journalists, armchair or otherwise. Performance Platform was retired in 2021.
Better demand
If the demand‑side problem was time and ability – that reading government data takes weeks of skilled work nobody could spare – that problem might now be resolved.
If anyone can now read and analyse the data, the constraint moves back to the supply side. Supply is hard. Contracts Finder publishes less than half of NHS contract awards. The mechanisms of open government don’t have the teeth to be useful, and AI can’t analyse data that was never published. It requires government to put more effort into being open and accountable. But if it did, maybe the impact would be felt this time.
Making things open might make things even better.
I’m on holiday for 3 weeks. Byeee x