Sources & methodology

University rankings

A conglomerate of public rankings. Domestic: US News and THE US, 2026. International: QS 2027, CWUR 2026 and available THE World 2026 records. The supplied THE World extract is partial; US entries are supplemented by the world-rank column in THE's US table.

Latest loaded editions only. Domestic schools may use a single-publisher fallback when THE or US News is missing; international requires at least two publishers. Equal-weight geometric mean of publisher positions; rank intervals use their midpoint, and open-ended bands retain their lower bound. Composite positions are among covered schools. The number of sources varies for the partial international table. Rankings are not official publisher placements.

Sources: US News, Times Higher Education, QS, CWUR. Imported from founder-supplied extracts. Publisher composites remain local-preview only pending commercial reuse review.

Univised career & value

Historical LinkedIn search appearances cover selected employer baskets, roles and locations, not unique alumni or hiring probabilities. Schools share the same recorded query panel. Missing queries are not zeros, and employer baskets cannot identify individual-company outcomes.

Career presence = recorded appearances divided by expected appearances. For each location, the reference rate is total recorded appearances divided by the compatible schools' total estimated field cohort. A school's expected count is that rate multiplied by its estimated field cohort. Overall comparisons divide summed observed counts by summed expected counts across supported locations. Location searches use that location's own reference; company and role filters rebuild the comparison for the available query scope.

1× means expected presence; 1.4× means 40% more recorded appearances than expected. The benchmark is cohort-weighted within the sampled schools, not the average applicant or all universities. Recent annual, field-weighted credentials are a proxy for historical alumni populations, not a matched graduating cohort. Repeated searches can count the same person. This is not a hiring probability, a causal school effect or a forecast.

Raw ratios above remain the evidence layer and the scoped-search measurement. Industry ranking headlines apply continuous regularisation to every direct record: rating = 1 + w × (raw ratio − 1), where w = effective expected count / (effective expected count + 25). Effective expected count is expected count divided by 1 + the absolute log disagreement between field and whole-institution ratios, with 0.1 added to each ratio for zeros. The 25-appearance setting is a conservative stability assumption, not a fitted probability parameter. Alternative settings of 10 and 50 are reported inside rating details. Raw zeros remain visible; an adjusted zero-count rating is not proof of positive placement.

No name-based adjustments or score jumps at positive-market thresholds apply. Geographic concentration remains visible. Industry publication still requires twenty compatible employment anchors, ten anchors with five appearances and three informative markets. Reputation alone cannot create an industry list.

Missing-school estimates use only direct employment anchors, after applying the same regularisation to them. Peers must have compatible institutional groups, degree levels, subject mix, size and academic fundamentals even when traditional ranks are available. Domestic and global ranking coordinates remain separate. Require at least five donors and four effective donors; shrink estimates further toward 1× for small groups, distance, missing publisher ranks and cross-country transfer. Estimates never become donors. A leave-one-school-out check must improve on both median and neutral baselines by 2%; otherwise missing schools receive a neutral fallback without a position. These checks do not validate unsampled institution types, countries or top-of-table accuracy.

Numeric positions follow adjusted scores, with rounded ties kept together. Estimated position groups are available in details, not statistical confidence intervals. A broad 1× fallback means unknown, not observed average performance. Filtering does not renumber the catalogue. An industry rating does not establish programme availability or hiring odds.

Market-transfer validation is separate from descriptive ordering. We hold out one market and compare its observed/expected signal with the remaining markets, excluding the evaluated school from the reference rates. The diagnostic uses mean absolute log1p error against a neutral 1× baseline, with at least twenty eligible schools, fifty comparisons and a 2% improvement required. This does not calibrate hiring probability.

Value combines observed institution-wide median earnings ten years after entry and a common supported basket of evidence-adjusted career ratings, with equal weight in a geometric average, then adjusts for relative annual tuition and living costs. Inferred peer outcomes never enter Value. Cost and earnings references share country, currency and vintage. Lower cost improves value; expensive schools need stronger outcomes. Incompatible horizons and incomplete costs are excluded. Value also applies a market-omission sensitivity check through the same evidence weights. It is not a degree-specific or lifetime ROI forecast; aid, taxes and financing are excluded. Use the ROI planner for personal scenarios.

Campus experience

Campus survey ratings in this local preview originate from the historical Niche archive, rounded to whole numbers out of ten. They are not new Univised surveys. Source records and vintages are retained internally. These ratings are excluded from production pending replacement or rights clearance.

Maps

OpenStreetMap contributors (ODbL), OpenMapTiles and OpenFreeMap. Venue snapshots are cached; basemap tiles load as you explore. Where available, Police.uk crime counts use approximate locations under OGL v3 and describe a dated snapshot, not personal safety risk.

Peer benchmarks

General industry ratings use the coverage model described above. Specific career-path searches remain separate: their peers must answer the same archived employer/role/location query and satisfy country, geography, size, academic and subject-mix checks. A broad traditional-ranking estimate is not substituted for missing company- or location-specific evidence. Estimated schools are never used recursively as observed donors.

Faith & local highlights

Institutional religious affiliations use College Scorecard and the NCES IPEDS IC2024 release. Affiliation is not the distribution of student beliefs; missing values are not treated as nonreligious. Student shares appear only where a university publishes an aggregate, with population and vintage retained. NCES affiliation definitions.

Houston and Provo introductions use factual place names and original descriptions informed by Visit Houston, Houston trip ideas, Explore Utah Valley and Visit Utah. Other city cards select diverse mapped venues and at most one archived sports team; they are not a verified top-six popularity ranking. Nearby map points are shared across campuses when their coordinates fall within the same radius.