Three-quarters of government evaluations are commissioned to improve implementation. Two-thirds are commissioned for accountability.

They're not the same thing, and building a system that serves both turns out to be harder than it looks.

This episode is about the structural tension between backward-looking accountability — did we deliver what we promised? — and the forward-looking question of whether we're actually learning anything useful.

Caroline argues the current measurement architecture is structurally bad at three things: it locks in categories at the moment of maximum ignorance, it treats evidence as binary when almost everything useful lives in the middle, and it has no way of capturing what you found out that you didn't know you were looking for.

There's also a detour through lifts, chin hairs, tattoo-level decisions, and Marie-Antoinette's diningware.

Referenced in the pod:

The APS State of Evaluation report (2026) — the list of why government commissions evaluations
Andrew Greenway's article on Sir Olly Robbins and The Accountability Act
Adam Kucharski's article Do we need more 'aluminium-standard' evidence?
Inside the Box: How Constraints Make Us Better by David Epstein
Tom Loosemore Universal Credit TWT episode

This podcast was recorded on Kaurna land, and we recognise Kaurna elders past and present. Always was, always will be.

Now for some appropriately bureaucratic disclaimers....

While we have tried to be as thorough in our research as busy full time jobs and lives allow, we definitely don’t guarantee that we’ve got all the details right.

Please feel free to email us corrections, episode suggestions, or anything else, at thewestminstertraditionpod@gmail.com.

Thanks to PanPot audio for our intro and outro music.

'Til next time!