Show the source. Explain the gap.

Core rules

What every result must do

1. 01
   Name the source and year

   A site-wide updated date is not enough because costs, admissions, programs, earnings, and projections come from different cohorts.

2. 02
   Keep different scopes separate

   School-wide graduation rates do not describe one degree. System-level earnings do not become campus-level outcomes because they appear on a campus page.

3. 03
   Leave missing values missing

   Use “Not reported,” “Suppressed,” or a more specific reason. Never convert missing data to zero or silently lower a school’s rank.

4. 04
   Distinguish facts from interpretation

   A source value is a fact. “Stands out for” is an editorial summary that needs visible evidence.

5. 05
   Avoid causal claims

   Historical earnings do not prove that a college caused the result. Related degrees do not prove that graduates enter a particular occupation.

What the numbers describe

Use the strongest measure for each question

Employer data boundary: no free authoritative national source shows which individual companies hire graduates from every college, so CollegeLineup does not claim one. An employer entered in a tool is a note beside the path and nothing more.

Attribution and license details for each source live on the data sources page.

Data model

Preserve the identifiers that explain scope

UNITID
  to institution facts and the campus page

UNITID + CIP6 + award level
  to exact program offering and distance flags

OPEID6 + CIP4 + credential level
  to historical earnings and federal debt outcomes

CIP6 to SOC6
  to possible occupations, wages, outlook, and work context

When an OPEID6-level outcome repeats across branches, the page must label it as system-level. When a four-digit field outcome is shown near a six-digit program, the broader scope must be visible.

Rankings

Answer one question per list

Every list needs a visible question, eligible universe, factor definitions, weights, missing-data rules, source years, and methodology version. A college cannot pay for placement or better editorial treatment.

Two orders are in use today. The balanced blend scales completion, affordability, and reported earnings from 0 to 1 across the eligible group, reverses net price so a lower cost scores higher, and averages the three with equal weight. The value and two-year lists sort on one reported number, net price, after an eligibility gate. Both print their eligibility rule, their factor list, and the count of colleges excluded for missing data at the top of the page.

Aim High, Realistic, and Safe

Estimated chances, and what they are built from

Public data provides overall acceptance rates and enrolled-student test percentiles. It does not provide a centralized view of essays, recommendations, course rigor, extracurriculars, major capacity, or institutional priorities. The estimate is therefore a rough guide for balancing a list, and it says so wherever it appears.

Each estimate is assembled in the open:

• The published overall acceptance rate sets the starting point. With no student detail entered, the estimate is that rate and nothing more.
• An SAT score moves it by where the score falls inside the reported 25th to 75th percentile range, on a normal curve fitted to those two published percentiles.
• A GPA moves it against a selectivity anchor, because no federal source publishes an admitted-GPA range by campus. That anchor is a stated assumption, printed on the card, not a measurement.
• A college with no reported acceptance rate gets no estimate. Missing data never becomes a number.
• The estimate never reaches 0 or 100, and the category always agrees with the number beside it.

• Let students place and move schools manually, whatever the estimate says.
• Show factor-level reasoning such as academics, cost, program, and location alongside the estimate.
• Print every input that moved a number, so no percentage arrives unexplained.
• Keep “Safe does not mean guaranteed” beside every result set.

Career and employer goals

Use an employer as context, not a promise

V1 should ask for a role first. An optional employer can add industry or location context. Results can connect the role to possible degrees, the degrees to colleges that offer them, and the occupation to public wage and outlook data.

Do not call a college a feeder without strong, current, citable employer-level evidence. Alumni presence, a published employment report, and a recruiting event are different signals and must be labeled separately.

Online programs

Confirm the exact program with the school

Distance education reporting can indicate whether all, some, or none of the credentials within a field may be completed remotely. It does not identify every marketed program, class schedule, residency requirement, state authorization, current availability, or online tuition.

Every online result should tell students to confirm the exact program, tuition, format, in-person requirements, and availability in their state before applying.

Privacy

Keep personal planning on the device in V1

• No account is required.
• GPA, test scores, family budget, saved colleges, employer interests, and checklist state remain in local browser storage.
• Personal inputs do not appear in URLs, analytics events, logs, or error reports.
• Storage access uses safe wrappers because it can fail in private or embedded contexts.
• Students can clear, export, and import their lineup as JSON.
• Analytics, if added later, must use coarse page events and exclude form values.

The privacy page explains these practices in plain language.

Freshness and operations

Refresh data outside the Vercel runtime

Public bulk files are downloaded and transformed in a separate offline workflow, and the validated output is committed to the repository. The deployment serves static pages and static data files, with no runtime provider calls and no required environment variables.

This build was produced on August 23, 2026 from transform version 2. Every raw file carries a recorded SHA-256, and the build fails rather than publishing partial data when a source changes shape.