Era weighting. "The last decade" / "recent decades" / "the last half-century" are exponential decays with 10-, 25-, and 50-year half-lives: each alumnus counts 0.5^{(years since graduation ÷ half-life)}, with graduation approximated as birth year + 22. Under "recent decades," a 1999 graduate counts half as much as a 2024 one, and a 1974 graduate a quarter as much. About 10% of alumni have no recorded birth year and are left out of the weighted views (they're included in "all of history equally").

The overrepresentation multiple. (School's share of everyone noted for the trait, among people with a listed U.S. four-year school) ÷ (school's share of all U.S. four-year enrollment, weighted with the same era emphasis — see the size note below). Example: if a school educates 0.2% of college students but accounts for 10% of noted university professors, that's 50×. Both sides count a person at every school they attended. On the Law, Medicine, and Business tabs the size denominator is instead the school's share of national degree production in that field (JDs, MDs, and business master's degrees from the latest IPEDS year — a current-day snapshot, unlike the era-weighted university sizes; a handful of schools with facet alumni but no current degree production fall back to whole-university size). The Undergrads tab divides by the school's share of national undergraduate enrollment (era-weighted, IPEDS 1990–2023). Its population is: recorded bachelor's degrees, plus inferred undergrads — college athletes, and people linked to exactly one school — excluding anyone whose only recorded degree at the school is a graduate one, links to named professional schools, and people who also have a professional-school link at the same university. Multi-school people without degree records remain unclassified. Small counts are shrunk toward 1× before display.

Keeping flukes out. "Most overrepresented" lists require a minimum number of noted attendees that scales with the school's pool — 5 for small schools, up to 20 for the most-documented (this page: 5) — plus at least 25 people nationally, and every displayed multiple is shrunk toward 1× before ranking, so a handful of people can't dominate. A statistical confidence score is also computed for every row and stored in the data.

Typical size. The same era weighting applied to the school's total enrollment (all levels, IPEDS 1980–2023, single-year reporting glitches removed), so per-capita comparisons use how big the school was when its alumni actually attended, not just today's size. Sizes before 1980 aren't digitized anywhere; years before 1980 are treated as equal to 1980.

Employment records. Census Bureau Post-Secondary Employment Outcomes. Jobs are matched against national unemployment-insurance wage records covering over 96% of U.S. employment, so graduates are tracked wherever in the country they work. What's limited is the school list: only ~950 mostly public institutions in 33 participating states share transcript data. Federal and self-employment are the main gaps. Earnings are medians for graduates working and not enrolled.

Doctors. CMS Doctors & Clinicians registry: everyone who bills Medicare, with self-reported medical school. Includes physician assistants and other clinicians; undercounts specialties that rarely bill Medicare (notably pediatrics).

Earnings and debt figures (College Scorecard). From IRS and federal-aid records, covering federal-aid recipients only — students who never took federal aid aren't counted. The headline earnings number is measured 10 years after a student first enrolled (roughly 6 years after graduating for a typical bachelor's), among those working and not still in school. The debt figure is federal undergraduate loans at graduation, median among students who borrowed at all; students with no federal loans (including most students at some wealthy schools) aren't in it. "—" in the majors table means too few students to report without identifying anyone.

More like this. Computed separately for each alumni lens: on the law-school lens, law schools are compared with law schools, and so on. Each school is a vector of its overrepresentation multiples across occupations and employers (recent-decades weighting), compressed to up to 50 dimensions with a principal-components decomposition, plus a graduate-pay block (Scorecard median earnings for the general lenses; JD, MD, or business-master's earnings for the professional lenses) that contributes roughly a fifth of the similarity score. Neighbors are matched by cosine similarity. "Shared signature" lists the traits both schools are most jointly overrepresented in.

The big caveat. The fingerprint is built from the 0 alumni with Wikipedia articles. Wikipedia's idea of "notable" is generous to athletes, politicians, academics, and historical figures, and stingy toward, say, successful executives who keep a low profile. Read it as "what this school's alumni are known for." For the roughly 950 mostly public schools covered, the employment-records section is the antidote.