A best-practices doc lists tactics. Quantum stores outcomes: which mechanic ran, on which vertical, what the lift was, which psychological driver it pulled on, and why it failed if it did. We query by context (vertical, funnel stage, driver) — not by tactic name.

Direct access, no. The database includes anonymised data from other engagements that's contractually protected. But every research report and test recommendation we deliver is a query against Quantum — clients get the answers, just not the raw rows.

Continuously. Every test we ship — across every active client — gets tagged and added within 48 hours of the result reading. The database grows by 60-80 records per month.

Two-person review. The strategist who designed the test tags it; a second strategist validates the tag against the actual mechanic. Disagreements get a third reviewer. We've audited the corpus twice — tag accuracy is around 94%.