How this works
Methodology
What we collect, where it comes from, how we order it, and the things we will not do. Federal data last ingested September 14, 2026.
The short answer
We do not publish our own ranking. We reproduce the positions that eight established publishers assign, attach each publisher's stated weighting so you can see what produced it, and join those positions to the federal admissions, cost and outcome data the rankings are built from. The one figure we compute ourselves is a cross-source consensus percentile, defined below. Every number on the site carries its source and reporting year next to it.
Where every figure comes from
Four kinds of data appear on this site, and each has a single canonical origin. We do not average across origins or fill gaps with estimates; a figure we do not have is shown as missing rather than inferred.
- Admissions data
- Acceptance rates, SAT and ACT middle-50% ranges, student-faculty ratios and aid participation come from each institution's Common Data Set, the standardised annual disclosure colleges publish themselves.
- Enrollment, cost and completion
- Undergraduate enrollment, published tuition, room and board, six-year graduation rates and first-year retention come from IPEDS via the National Center for Education Statistics.
- Net price, earnings and debt
- Average net price, median earnings ten years after entry and median federal debt at graduation come from the U.S. Department of Education's College Scorecard. Earnings and debt figures cover federal aid recipients only, which is a real limitation we note wherever we use them.
- Ranking positions
- Rank, list name, denominator and year come from each of the 8 publishers' own published lists. Positions are their editorial property and are reproduced with attribution for comparison and commentary.
How cross-source consensus is computed
A rank means nothing without its denominator. Spelman College is #37 among 207 national liberal arts colleges; Purdue is #46 among 436 national universities. Reported as bare numbers those look similar, but the first is a stronger result.
So we convert every placement to a percentile — rank divided by the size of the list, times 100 — and take the unweighted mean across every list a school appears on. Lower is better. A school with a consensus percentile of 4 sits in the top 4% of every list it appears on, on average.
We also report methodology sensitivity: how many percentile points a school's standing moves depending on which published list you read. A small spread means its standing barely depends on whose formula you use. A wide spread means the ranking you happen to read substantially changes the impression you get, which is usually the more interesting finding.
Two limitations we will state rather than hide. Schools appearing on fewer lists have noisier consensus figures, and we do not penalise them for it. And the mean is unweighted, which implicitly treats all eight publishers as equally credible — a simplification, but a more defensible one than assigning our own credibility weights to other people's work.
How list order is computed
Every sort, filter and “best for…” list on this site is derived from published ranks and federal metrics — not from editorial favour or institution-specific flags. All of that logic lives in one module, src/lib/rankings.ts, so a change to ordering is always reviewable in a single place.
A unit test guards the module: it must not reference fields that exist only for editorial metadata, so list order cannot silently pick up non-data signals. Consensus percentiles, net-price sorts and hub orderings all flow through the same code path.
Where our dataset is too thin to make an honest list — some themed hubs name the schools that belong on a fuller version rather than padding the page. If we change a figure after publication, we update the site when the underlying source changes and refresh the snapshot dates shown on each page.
Automated access
This site is for people comparing colleges in a browser. We do not offer a public bulk export or scraping-oriented feed. Structured data on each HTML page describes that page for search engines; it is not a licence to republish the underlying compilation or third-party ranking positions.
Verification status of the current dataset
This site holds two kinds of data, and they are trustworthy to very different degrees. Conflating them would be the easiest way to mislead a reader, so they are tracked and labelled separately on every profile page and in structured data on those pages.
Federal figures: verified
Enrolment, admissions, tuition, net price, Pell share, first-generation share, graduation, retention, earnings, debt and student-faculty ratio for all 1913 institutions are ingested by script directly from the College Scorecard institution file (June 2026 release release) and IPEDS 2023. Nobody retypes them. The join key is the institution's IPEDS UNITID, the release is pinned in the ingest script so a new federal release is a reviewed commit rather than a silent change, and the generated file is committed so every figure is visible in a diff. These figures can be cited.
Where a figure is missing it is labelled Not reported with the reason, because the reason is usually the interesting part. A college that has withdrawn from federal student aid has no federally calculated net price, earnings figure or median debt — not because its outcomes are poor, but because the records those figures are derived from do not contain its students. A test-blind college reports no score range because it does not collect one. Rendering either as a dash would throw away the only useful thing we know.
Ranking positions: verified
For the 2026 edition we bulk-load and audit publisher placements in CI (npm test): U.S. News (national, liberal-arts and regional lists on this roster), Forbes America's Top Colleges (500-school list), Washington Monthly's tuition-and-tax ranking, WSJ/College Pulse Best Colleges in the U.S. (500-school print table), Niche Best Colleges (Overall), Money Best Value Colleges (star ratings), Princeton Review Best 392 Colleges (book membership plus category badges — no overall 1–392 rank), and QS World University Rankings (2026 edition nid, U.S. institutions on the global list). Each row records list name, rank (or listed status), list size (outOf), year, and rating text when the publisher supplies one.
Niche covers schools on the published Best Colleges list (~2,600 institutions); QS and Princeton Review only cover schools those publishers include. When every tracked publisher omits a school, the rankings table is intentionally blank — that is not a data gap. Hand-written profile copy is updated to match the bulk tables; cite the table on this site rather than older marketing lines.
What we will not do
- Publish a proprietary overall ranking. A ninth formula would add noise rather than clarity.
- Let an institution change a position, an ordering or a list placement.
- Remove a documented weakness from a profile at an institution's request.
- Use schema.org aggregateRating for third-party ranking positions. Those are other publishers' editorial placements, not ratings we issued.
- Pad a list to look comprehensive. Where our coverage is incomplete, the page names the institutions that belong on a fuller version.
- Generate pages programmatically to chase query volume. Every page here is written to be worth reading on its own.