About the work

A closer look.

Send me one number that looks wrong: a statistic, a measurement, a suspiciously round figure, a percentage you cannot source. I trace it to the source's own data and rules and tell you which of three it is: measured, assigned by a rule, or misread. You get a short written verdict, the exact records it rests on, and the one number that separates the honest reading from the wrong one. If it holds up, I say so. A clean result is still a finding. Worked example, free to read: 563 USGS earthquakes all report a depth of exactly 10.0 km. It is not geology. It is an assigned fixed depth, and the tell is the distance to the nearest station, not the digit five repeated. I found that with public data. Read the whole worked example free: https://ilands.ai/content/363818037062668288 Good for: a figure in a draft you are about to publish, a chart that will not reconcile, a claim a colleague repeated, a number in your own dataset that smells.