How the numbers are built

The example school

Definitions are easy to nod along to and hard to check. So every score below is worked through on the same real profile: the University of Vermont, a public university in Burlington with 11,743 undergraduates. Nothing here is invented. Each figure is the one the tool actually publishes, and you can open its live profile and check every one.

We picked a school with a full set of data, not a flattering one. Where Vermont has no value for a metric (it pays no endowment tax, it awards no short-term certificates) we work the example on a different real school and say which.

Where the data comes from

Seventeen public data sources, thirteen of them official federal collections, all free to access. The first two carry most of the figures. The rest add the policy, research, labor-market, safety and demographic signals. OpenAlex, Wikipedia and Wikidata are community-maintained, and we use them only where the institution matches with confidence:

Reusing these figures. The federal collections above are U.S. Government works and carry no copyright. OpenAlex and Wikidata are CC0. One source is different: WICHE's Knocking at the College Door (11th ed.) is © WICHE and licensed CC BY-NC-SA 4.0. The three figures on this site derived from it (the home-state enrollment cliff, in-state graduating-class size, and the projected shift in the applicant pool's composition) are adaptations of that work: they are offered under the same CC BY-NC-SA 4.0 license, and if you reuse them you must credit WICHE, link the license, note that the material was adapted, and license what you build under the same terms. Wikipedia contributes pageview counts read from the Wikimedia REST API, which are facts rather than text, so no share-alike obligation attaches; no Wikipedia prose is republished anywhere on this site. IRS Form 990 figures are retrieved through the ProPublica Nonprofit Explorer API. Nothing here is published with the endorsement of, or in partnership with, any of these providers.

The data is pre-built into static files from local copies of these releases. There are no live API calls at page load and no exposed keys; the site's own analytics and marketing tags are listed in the cookie policy. The tool is just data and JavaScript. Null, suppressed (PrivacySuppressed) and blank values are treated as missing, never as zero. Several metrics are projections or scenarios, not current-period facts: the field-demand outlook (BLS, 2024-34), the enrollment cliff (WICHE projections), Grad PLUS exposure and the Parent PLUS cap gap (FSA volume read against the 2025 budget-law scenario), Workforce Pell exposure (certificate share read against the 2025 budget law's new short-term-Pell eligibility, effective July 1, 2026), and endowment-tax exposure (a 2025-budget-law scenario, not current law). The closure-risk score is a structural index, not a forecast. Each is flagged as indicative where it appears.

Coverage

Peer groups

What it is. The set of schools a college is measured against. Almost every number in the tool is a rank inside this set, so the group decides what the number means.

How it is built. A single national percentile would be misleading, because a regional teaching college and a research flagship do not compete for the same students or the same money. So each school is placed in one of 70 peer groups, built from two facts it reports to the federal government:

• Carnegie Classification, what kind of school it is: doctoral/research, master's, baccalaureate, associate, special-focus and so on.
• Sector, who runs it: public, private nonprofit, or private for-profit.

Combine those two and you get the group. Percentiles and medians are computed only inside a school's own group, so the comparison is always like-for-like.

When a group would be too small. A rank against four other schools is noise, so a group needs at least 10 schools. If a Carnegie-and-sector combination has fewer, its members are moved to a wider group, in two steps:

1. 1Regroup by region and sector, for example "Public institutions in the Southeast region".
2. 2If that is still under 10, regroup by sector alone, for example "All private nonprofit institutions".

When Carnegie has not classified a school. Roughly a third of the institutions in the tool carry no Carnegie classification: newer schools, single-campus branches, and much of the certificate sector. They used to be pooled into one residual "unclassified" bucket, which is not a peer group, it is a leftover. They are now grouped by institutional level and sector instead: four-year, two-year or less-than-two-year, crossed with public, private nonprofit or private for-profit. A two-year public is compared against other two-year publics.

There are 70 peer groups in all. 55 are a straight Carnegie-and-sector match, covering 3,563 institutions. 9 group the institutions Carnegie has not classified, by level (four-year, two-year, less-than-two-year) and sector, covering 2,108. 3 fall back to region and sector (40 institutions) and 3 to sector alone (76). The smallest group holds 10 institutions, the largest 1,323. Every profile names its own group and its size, so you can always see which rule applied.

The honest limit. A peer group is a starting point, not a claim that 93 schools are interchangeable. Size, state funding and mission still vary inside a group. The tool also offers a separate "nearest neighbors" set on each profile, matched on size, selectivity, cost, outcomes and student mix, for when the fixed group is too coarse.

Percentiles & standing

What it is. A percentile is a class rank. It answers one question: what share of this school's peers it sits above on this measure. A 97th percentile means it is ahead of about 97% of the group.

Percentile is not the same as good. Higher is not always better: a high median debt or a high net price is bad news. So we translate the raw percentile into a standing using each metric's direction:

For metrics where lower is better (debt, net price, default rate) the standing is flipped, so "Strong" always means "doing well", never just "a big number". Some metrics are neutral (enrollment size, Pell share). Those get a percentile but no good/bad standing, because more is not inherently better or worse.

When we publish no percentile at all. Two cases, both deliberate:

• Fewer than eight schools report the measure. A rank against one or two peers is arithmetic, not information. The value and the peer median still show; the rank does not.
• Every school in the group reports the same value. If a thousand schools all report 0%, no ranking among them means anything, whatever number you print. The shared value is shown as the peer median instead.

Where the rank rests on a thin group (eight to nineteen schools), the peer count under the card is shown in amber rather than grey.

Metrics with no direction get a mark, not a bar. About half the metrics here are neither good nor bad: enrollment size, share of students over 25, research volume, published tuition. Those cards show a small tick on an empty track, saying where the school sits, rather than a filled bar, which would read as a score. Only metrics with a stated direction are filled in and tagged Strong, Average or Below peers.

Which benchmark you are reading against

What it is. Every median and every rank on a profile is computed inside that school's peer group by default. The buttons above the metric cards move the reference: to the school's home state, to the whole national universe, or to a peer set you build yourself. A number does not change when you switch. What it is being measured against does, and that is usually the argument.

What moves when you switch, and what does not.

• Moves: the median printed beside the rank, the dashed reference line on every trend chart, and the spread bar under every rank. All three are read from one lookup, so a card cannot show one reference's median over another reference's spread.
• Does not move: the percentile, unless you are using a peer set you built. A rank is an exact count of institutions, and computing one inside a state needs every value in that state, not its summary. Rather than interpolate an estimate and print it in a slot that has always held a count, the rank line keeps saying in peer group and says so on every card.

Two floors, because a median and a spread are not the same claim. A reference needs 5 reporting institutions before we will print its median, and 8 before we will draw its spread, which is the same floor a percentile has to clear. A small state can therefore show a median with no bar under it. That is the honest outcome: a box plot drawn from six numbers invents a shape.

A hollow marker instead of a solid one means the school falls outside the track entirely, above the reference group's 90th or below its 10th. It is pinned to the end of the bar and reads as “at least this far out”, not as an exact position.

The three prices

A college does not have a price. It has at least three, they differ by a lot, and treating them as one number is the most common way to be wrong about cost. The tool keeps them separate and scores them differently.

The informative number is usually the gap between the low-income and the full-pay net price. A wide gap means aid is being aimed at need. A narrow one means the school discounts broadly, which helps its yield more than it helps its neediest students. Each price card carries an information mark explaining which of the three it is and why it is scored the way it is.

How old each number is

The trap. The Department of Education publishes a "most recent" institution file, and it is not one vintage. It is assembled element by element, taking each measure's most recent non-null value. A measure the Department stopped collecting in 2016 sits in the current file looking exactly as fresh as one collected last year. Labelling the whole file with its release year, which is the obvious thing to do, is wrong by up to a decade for some measures.

What we do instead. Every measure is dated individually, and the date is derived rather than assumed: we compare the current file against every annual release back to 2010 and take the year whose values match. Each card shows its own date, and where the file year still is not the measurement window, the card says so.

A stale federal figure is still the only federal figure that exists, so we publish it rather than leave the card blank, and we say how old it is. Earnings are the case to watch: they are measured in a single tax year for a cohort that entered college roughly a decade before, so a 2020-21 earnings figure describes students who started well before 2020.

Federal collections arrive at different speeds. The other federal source behind the tool, IPEDS, publishes each component on its own schedule, so the layers below are not all from the same year. Each card carries the year of the collection it came from, and a layer is only moved forward when the newer file still carries the fields it is read for.

• 2023-24. Derived 12-month enrollment: full-time-equivalent headcount
• 2023-24. Derived human resources: average salary by academic rank
• 2023-24. Institutional characteristics: published room and board charges Held back: the newer file no longer carries the published room and board charges.
• Fall 2024. Institution directory: names, locations, metropolitan area, control and level
• Fall 2023. Admissions: applications, admits, enrolled, test-score submission
• Fall 2023. Derived fall enrollment: student-faculty ratio, out-of-state and distance-ed shares
• Fall 2023. Staff by occupational category: instructional-staff headcount and demographics
• FY2022-23. Finance, GASB (public institutions)
• FY2022-23. Finance, FASB (private nonprofit institutions)
• FY2022-23. Finance, for-profit institutions
• Fall 2022. Residence and migration of first-time undergraduates Held back: reporting is optional in odd-numbered years, so the newer file covers fewer than half of all institutions and cannot be compared across peers.

2 newer files are published but not used, for the reason shown. A partial or column-short release would make a comparison look complete when it is not.

Ratios that cross collections. Endowment per undergraduate divides an IPEDS fiscal-year figure by a Scorecard enrollment count, and the two are not from the same year. Cards like that name both dates rather than picking the flattering one, and should be read as an order of magnitude.

Filings that arrive on their own schedule. Single audits and IRS Form 990 returns follow each institution's own fiscal calendar, so we show each school at its most recent filing. That means a peer comparison on those measures spans a few filing years rather than one, and the cards say so.

Financial Health Index (CFI)

What it is. The headline financial score, and the one number here that is not an Ibex invention. It is the NACUBO Composite Financial Index (CFI), the balance-sheet measure accreditors and college boards already use, and whose underlying ratios bond-rating agencies track. Unlike a percentile, it is an absolute score: it says how strong the balance sheet is, not how it ranks.

How it is built. Four ratios come out of the IPEDS finance filing. Each is converted to a "strength factor" on a common scale, then blended by fixed weights onto one scale running from about −4 (weak) to 10 (strong):

Reading the scale:

The honest limits. Below 3 falls short of the conventional threshold for financial health; below 1 signals acute stress. Schools carrying little or no building debt have the viability ratio dropped and the other weights re-normalized. The CFI is computed from IPEDS FY2022-23 and lags the current year by two to three years. It is calculated for public (GASB) and private nonprofit (FASB) institutions. Private for-profit colleges report under a different FASB standard with no comparable net-asset or reserve concept, so they get an operating margin, tuition dependency and revenue composition but no Composite Financial Index. Branch campuses that report finances at a parent or system level can show distorted ratios. Where a branch, online division or other child unit files no IPEDS finance survey of its own, its profile shows the parent or system's figures, labeled as such and left out of rankings and peer percentiles so the same finances are never counted twice. It is for informational benchmarking, not a credit rating, and not financial advice.

Financial Resilience Score

What it is. The 0–100 gauge on each profile. It is a class rank, not a health grade: it says how a school's per-student finances compare with its own peer group. 90 does not mean "90% healthy". It means "near the top of this group". A school can score high in a weak group, and a strong school can score middling in an elite one.

How it relates to the Financial Health Index. They are two different measurements and they can disagree, so read them together rather than as one verdict. The Resilience Score is a peer-relative ranking of per-student resource levels. The Financial Health Index is an absolute CFI-style score of the institution's own balance sheet and operating result, computed without reference to peers. A school with modest resources per student can still run a sound balance sheet, and a well-resourced one can still post an operating loss. Across the 2,821 institutions that carry both, the two are close to uncorrelated (r = 0.06), which is what you would expect from measures of different things, not a sign that one of them is wrong.

How it is built. Up to three per-student measures from the IPEDS finance filing. Each is turned into a percentile inside the school's peer group, and the score is the average of the percentiles it has. Higher is stronger. No weighting, no adjustment, no hidden inputs:

• Endowment per undergraduate, the cushion behind each student.
• Net tuition revenue per FTE, what it actually collects per student after discounts.
• Instructional spend per FTE, what it puts back into teaching per student.

The honest limits. Most scores are built from two inputs, not three, and that is the norm rather than the exception. Of the 5,787 institutions in the tool, 5,691 carry a score and 96 are suppressed for reporting none of the inputs. Of those 5,691, 3,356 (59%) are scored on two inputs and 2,335 on all three. The missing input is always the same one: endowment per undergraduate. Every for-profit lacks it, which is faithful rather than a gap, since they hold no endowment; the rest are publics and nonprofits whose endowment is not in the filing. A two-input score and a three-input score look identical on the gauge, so check the components shown beneath it.

This also bites inside a group: in 32 of the 70 groups, some schools are scored on three inputs and others on two. Where that happens the scores are ranked on the same scale but are not built from the same measures, so a close gap between two schools in a mixed group is weaker evidence than the same gap in a group where everyone reports the same things.

And because it is a rank, the same dollars produce very different scores in different peer groups. For an absolute read on financial strength, use the Financial Health Index instead.

Structural closure-risk index

What it is. A 0–100 index where higher means more structural pressure. It screens for the financial and demographic strain that has historically come before closures and mergers. It is not a prediction that any school will close, not a credit rating, and not a forecast.

How it is built. Four signals. Each is converted to its own 0–100 pressure reading on a fixed, published scale, and the index is the plain average of whichever are available. At least two must be present, otherwise the score is suppressed:

The honest limits. Equal weights are a choice, not a finding: we have no evidence that these four pressures matter equally, and we do not pretend otherwise. The scale endpoints are judgment calls, published above so you can disagree with them. Both money signals come from FY2022-23 filings, and the cliff is a projection to 2041. It is a signal to investigate, never a verdict.

Graduation vs. completion

What it is. Two different measures of student success, because the federal government measures it two different ways, on different students, over different time windows. They are not directly comparable, and a school can look very different on each. We show both, side by side, on every profile.

Both families are reported by the U.S. Department of Education and lag the current year by roughly two to three years.

Policy-exposure scenarios

What it is. Four signals modelling how a school could be affected by the 2025 federal budget law. Read them as "here is the size of the exposure if this lands as written". They are scenario estimates, not current law, and not a bill anyone has received.

Field-of-study analytics (CIP)

What it is. The "Top fields of study" section on each profile, which asks whether a school's individual majors pay off. It rolls up College Scorecard Field-of-Study data to the CIP 2-digit family and, when you expand a field, drills into its individual 4-digit-CIP majors. That is the finest grain ED publishes, since there is no 6-digit earnings series.

Completions come from eight dated yearly releases (2014-15 through 2022-23). Median earnings one and four years after completion are completion-weighted averages of program medians. ED moved debt and the earnings-threshold counts out of the latest dated files, so those come from the Most-Recent-Cohorts snapshot. Each 4-digit major links to a national field-of-study page ranking every U.S. college offering it. Major-level figures appear only where enough graduates exist to clear ED's privacy threshold, so many cells read "–".

• Earnings-premium test and the 2-of-3 mechanic. Compares a program's median earnings four years out against the ACS median for a high-school graduate (undergraduate credentials) or a bachelor's holder (graduate credentials) in the school's state. The 2 of 3 yrs badge flags a program that came in below that benchmark in two of its latest three reported cohort-years, which is the statutory trigger under the 2025 federal earnings-premium test (effective July 1, 2026), under which a program can lose Title IV eligibility. Indicative only: ED's official determination uses its own cohort definition and may differ.
• Threshold pass-rate (% > threshold). ED's own share of a program's graduates earning above the federal earnings threshold: the above-threshold count divided by the measured-earnings cohort. This is the literal "do-no-harm" pass rate the 2025 test refers to. ED publishes it sparsely and in different measurement windows across releases (the 4-year window appears only in the most-recent snapshot; 5- and 1-year windows appear in other files), so we pool all nine Field-of-Study releases and report each program from the best available window: 4-year preferred, then 5-year, then 1-year. A small chip marks any figure not on the 4-year window, and hovering names the cohort size and source release. Every cell is real ED-published data with no imputation, and pooling raises coverage from roughly one program in five to about five in six.
• Debt-to-earnings. Median student debt at completion divided by median earnings four years out, a gainful-employment-style value ratio. Below 0.5× shows green, above 1.0× red.

Field-demand outlook

What it is. A forward-looking read on whether the jobs a school's degrees feed are growing or shrinking. It is the projected 10-year (2024-34) change in U.S. employment for those occupations, from BLS Employment Projections, mapped through the NCES CIP-SOC crosswalk and weighted by the school's own program mix. The benchmark is the U.S. all-occupations projection of +3.1%.

Three fields are excluded because their crosswalk maps to "any job" and carries no coherent occupational signal: 05 Area/Ethnic/Gender Studies, 24 Liberal Arts & Humanities, and 30 Multi/Interdisciplinary. Excluding them is a judgment call, and it means a school heavily weighted toward those fields has a thinner signal here than the number suggests. It is indicative, and never a guarantee.

Risk flags

What it is. Short warnings that surface a pattern one number alone would hide: multi-year enrollment decline, heavy debt-to-earnings burden, or an unusually large endowment draw for the school's size, for example. Flags are fixed rules, not judgment and not AI. Each one is either true of the data or it is not. They are color-coded amber (watch) or red (acute), and shown only when the underlying data supports them.

They are signals to investigate, not verdicts, and a flag is never a claim about an institution's future.

Role lenses

The profile view offers four optional lenses: Enrollment VP, CFO, President, and Board / Trustee. A lens never changes a single number. It reorders the metric cards, stars (★) the measures that role watches most, and opens on the most relevant tab. It is presentation over identical data, so two colleagues can read the same profile through the lens that matches their job.

Rankings

The leaderboards rank on one disclosed metric at a time. No weighting, no composite, no editorial judgment. Schools that do not report the metric are left out. To stop tiny programs distorting the lists, balance-sheet rankings (financial health, endowment per student) require at least 1,000 undergraduates, and enrollment-growth lists require a meaningful starting base. Each list states its own filter in a note. These are data rankings, not endorsements.

Value, ROI & the OBBBA screen

Two sections of the tool derive a number rather than republish one, so both are set out here in full.

Value & ROI is one ratio. Years to recoup = (net price × 4) ÷ (median earnings 10 years after entry − the median earnings of a high-school graduate in the same state). Net price and earnings come from the College Scorecard; the high-school baseline comes from the Census Bureau's American Community Survey, taken state by state so a school is measured against the labor market its graduates actually enter. A school is listed only if it enrolls at least 400 undergraduates and its earnings premium is positive; where the premium is zero or negative the ratio has no meaning and the school is left out rather than shown as infinite. It assumes four years to a degree and a flat premium over time, neither of which is true for every student, which is exactly why it is published as one disclosed ratio and not as a score.

The OBBBA Compliance Watchdog is a screen, not a determination. The 2025 budget law adds a program-level earnings-premium standard from July 1, 2026: a program can lose federal loan eligibility if its graduates' median earnings fall below what a typical high-school graduate earns in the same state (for undergraduate credentials) or a typical bachelor's graduate earns (for graduate credentials). We apply that comparison to each 4-digit-CIP program a school reports in the College Scorecard Field of Study files, and rank schools by the share of completions in programs that fall below their benchmark, weighted by completions rather than by program count so a large failing program counts for more than a small one. Schools with fewer than four programs carrying published earnings, or fewer than 400 undergraduates, are not screened at all. This is our arithmetic on public data, run years before the Department of Education publishes its own; it is directional, it is not the official calculation, and it is not a legal determination about any institution.

Limits & what this is not

• It is not a credit rating or investment advice. The CFI tracks ratios similar to those agencies use, but it is not a rating and carries no opinion on any security.
• It is not a "best colleges" ranking. There is no overall quality score and no weighting of one metric against another.
• The data lags. Scorecard figures trail the current year by about 2 years; IPEDS finance by 2 to 3. Every metric is labeled with its source year, so check the label before quoting it.
• Most scores are ranks, not grades. Financial Resilience and the percentile standings say where a school sits among its peers. They do not say the peer group itself is healthy.
• System reporting can distort. Branch campuses that file finances at a parent/system level may show ratios that do not reflect the individual campus. Where a child unit files no finance of its own, its profile shows the parent/system figures, labeled as such and left out of rankings and peer percentiles.
• Small cohorts are suppressed. Federal privacy rules withhold earnings and debt for small programs. Those cells show as missing, never as zero.
• Blank does not mean clean. On the Single Audit metrics especially, an empty value means the school filed no audit, not that the audit found nothing.

Quick answers