Essays, frameworks, research notes, investigations
Short pieces on how we work with evidence, questions and knowledge. Each one is tied to something running in production.
Why questions come before research
Roughly four in five organisational questions seek confirmation. A study that answers them thoroughly produces a large report and a poor decision.
framework · 2026-08Movement and confidence are different numbers
A trend score says how much something is changing. A confidence score says how much the evidence lets you claim it. Merging them produces a number that means nothing.
research note · 2026-07Divergence is a finding, not an error
When two authoritative sources disagree on the same indicator, the honest output is both values.
technical investigation · 2026-07The 68% that did not exist
A fabricated figure in an early Context Brief motivated the numeric-anchoring gate. This is what it does.
framework · 2026-06When a network is the wrong picture
A trade relationship involves a municipality, a product, a country, a period and the signals that affect it. Five edges lose what one hyperedge keeps.
research note · 2026-05Refusing to resolve is a valid result
An entity resolver that always returns a match is lying some of the time. Ours records why it declined.