Most readers on this page already know how to parse a methods section. The pieces below are for ranking evidence: from a clean randomised trial down to two dashboard series that merely moved in the same week.
Ranks causal claim strength on a ladder, from a genuine randomised trial down to two numbers that merely rose together on a dashboard.
Sorting tool: ranks what actually supports a causal claim against two numbers that only moved in the same week.
Read more ›Thirteen published experiments that misled while telling the truth about their numbers: sound statistics, broken inference.
Inference gap: clearing p < 0.05 leaves the story built on top of that number still open to challenge.
Read more ›Shows why a before-and-after metric comparison fails as measurement, even when the change looks like a win.
Team habit: the pre/post slide most groups already trust is the structure this report dismantles.
Read more ›Explains why a pilot that worked so often collapses once the same design rolls out at scale.
Scale problem: pilot-to-rollout failure is a methods issue with a literature behind it, not random bad luck.
Read more ›Tversky & Kahneman's 1991 job-choice study, with the design's real strengths and the places the method stops short of the claim people usually attach to it.
Reading practice: strengths and limits of one seminal paper in the same pass, past the abstract's headline result.
Read more ›