Journal · 19 Apr 2026
Sampling session notes without turning them into surveillance
Shared documents after a workshop are a gift and a hazard. They contain the only durable trace many hybrid rooms will ever have. They also contain jokes, names of people who were struggling, and half-formed doubts that were never meant to travel to a dashboard.
Sampling, as we teach it, is a political act with a small technical surface. You decide in advance which sections of the notes you will code: the opening contract, a mid-session conflict, the closing list. You copy those passages into a working file with names reduced to roles where possible. You do not ingest the whole corpus into a model “just in case.” Just in case is how archives become leverage.
United Kingdom organisations already live with UK GDPR language about purpose limitation. Session Quality Analytics should inherit that instinct even when the notes are “only internal.” Internal forwards are how a private parking-lot item becomes a performance story. Our After-Action Scoring lab therefore forbids keyword searches across a year of workshops unless a named governance group has written why.
If your legal team asks for a retention period, start with ninety days for coded extracts and keep the original notes where they already lived, under the rules that already applied. Inventing a second archive because analytics feels scientific is how good methods become creepy.
The quality claim you can defend is always smaller than the dataset you could hoard. That is a feature. A one-pager that cites three sampled windows will embarrass you less in six months than a warehouse of everyone’s asides.