Schools → ComparisonFinalists chosen
A small, repeatable loop — baseline the current page, re-test sketches against it, roll out with TARS — applied one production page at a time before we commit to a direction. Built from Vitaly Friedman's "How to Measure UX" and the open questions on design‑crit‑dataviz.
"How to Measure UX" argues that success isn't happiness — it's task comprehension — and that most design measurement fails because teams jump to a solution before mapping the problem, then try to take credit for business metrics too many steps removed to prove. The talk walks through KPI trees, value architecture, correlation-vs-causation traps, the TARS framework (Target · Adoption · Retention · Satisfaction), and top-task usability testing.
Recorded 2026‑10‑27, two parts. Full writeup below; raw transcripts linked for reference.
A GitHub-wide search for customer chart/dashboard complaints came back empty, so we asked PMs directly instead. Three answers, one theme: trust and relevance, not visuals.
Independently, two PMs described a data-trust / perceived-value problem rather than a visual one — and a third supplied a named framework (preattentive attributes, five color functions) to design against instead of a taste call.
Each page gets its own measurement plan before any sketch is chosen for it — same five-step method, applied fresh to what that page actually gets wrong.