Find your kind of excellence.

Positions follow the numeric rating, including estimates, within the informed comparison. They stay the same when you filter. Numeric positions are not claims of statistical precision. Schools with only a broad fallback retain a numeric rating but have no meaningful individual position; they appear afterwards with a dash.

1.4× is an evidence-adjusted estimate of 40% greater career presence than the sampled cohort benchmark—not a 40% greater chance of being hired. Raw recorded ratios remain available inside each rating.

All direct ratings use the same gradual adjustment toward 1×. Small expected counts and disagreement between cohort definitions receive more adjustment; crossing a market-count threshold never changes the scoring formula. The stability setting is 25 expected appearances and is not a fitted probability model.

Missing observations may use compatible-school estimates, considering institution type, subject mix, size, selectivity, completion, geography and comparable traditional rankings. Donors contribute their adjusted ratings, not raw spikes. Sparse peer groups receive extra adjustment. No school receives a name-based boost.

If compatible peers are insufficient, or the industry's peer model fails its held-out checks, we show a neutral 1× fallback without an individual position. This is an unknown result, not evidence of average outcomes. The model does not verify programme availability.

The employment calibration sample is US-based. Estimates for other countries are explicitly marked as cross-country extrapolations, not locally observed employment rates. Neither observed nor estimated ratings are hiring probabilities.

Peer estimates did not beat both baselines; schools without direct evidence receive an unranked neutral fallback. These checks reconstruct historical signals, not future job offers or top-of-table accuracy.

Sampled roles: junior associate, legal assistant, paralegal. Markets: boston, california, london, new york, philadelphia + dc, texas. Employer baskets may overlap and cannot isolate individual-company outcomes.