Research Notebook
Notes from building FirmRanker
Methodological observations from designing a system to measure how AI discovers and recommends law firms — the problems we encounter and how we handle them.
Research Note 001
Why we’re separating “recommended” from “suggested”
A firm placed in a list and a firm singled out as the best fit are commercially different events. Measurement has to keep them apart.
Methodological observationRead the note→Research Note 002Why one AI response cannot establish visibility
A screenshot is one observation. Visibility is a pattern that only shows up across repeated, comparable observations.
Methodological observationRead the note→Research Note 003Why entity resolution matters when measuring law firms
Before you can count how often a firm appears, you have to be sure which firm each mention refers to.
Methodological observationRead the note→Research Note 004Why AI citations need careful interpretation
A source that is cited is not necessarily a source that caused the answer. Treating citations as causes invites false conclusions.
Methodological observationRead the note→