Skip to main content

Our Methodology

By Ryan Kallander, Founder

Data systems expert who has built analytics and revenue infrastructure at B2B SaaS technology companies. Parent researching schools for my own family. Read more

Our Approach

We believe school quality should be measured by what a school adds to a child's education, not by the demographics of the families who attend. A school in a low-income neighborhood that helps students grow faster than expected deserves recognition. A school in a wealthy area where students coast on advantages deserves honest context.

That's why student growth is our primary signal. Schools that consistently improve student outcomes year over year score highest in our system, regardless of where they start.

Our poverty adjustment ensures that two schools with identical test scores but very different student populations get meaningfully different academic ratings. We use statistical regression to establish what's "typical" for a school's community, then measure how each school performs relative to that expectation.

A Worked Example

Abstract weights get easier to read once you see them applied to a real school. The numbers below pull from today's highest-composite public school in our index, Grinnell Grade School in Grinnell, Kansas. About 38% of its students qualify for free or reduced lunch, so the poverty-adjusted residual carries real weight in its composite. It ranks #1 of 1,292 rated schools in Kansas.

ComponentScore (0–10)WeightContribution
Student Progress (Growth)10.0050%5.00
Adjusted Academics10.0030%3.00
Resources & Environment10.0020%2.00
Composite—100%10.00

Applying elementary weights. A missing component redistributes its weight proportionally, which is why contributions can sum slightly differently from the raw percentages.

What We Measure

Every public school score is built from up to four components, weighted by school level:

Student Progress (Growth)

Elem 50% · Mid 40% · HS 25%
What we measure
Year-over-year change in state-test proficiency within each school.
Why it matters
Identifies schools that move students forward, independent of where they started.
Limitation
Needs 2+ years of tests; 2020–21 is down-weighted and missing-growth weight shifts to academics.

Adjusted Academics

Elem 30% · Mid 35% · HS 25%
What we measure
State test proficiency, blended 60% raw with 40% poverty-adjusted residual.
Why it matters
Wealthy schools that merely match peers earn near-zero; low-income schools that beat the line earn a positive residual.
Limitation
States with fewer than 30 rated schools skip the adjustment — the sample is too small for a reliable fit.

College Readiness (HS only)

HS 30%
What we measure
4-year graduation rate, AP participation, IB programme availability.
Why it matters
It's the sharpest outcome signal once students exit the state-testing window.
Limitation
~47% of high schools have grad rate only; ~22% have none, and the weight redistributes.

Resources & Environment

Elem 20% · Mid 25% · HS 20%
What we measure
Student-teacher ratio and per-pupil spending.
Why it matters
Captures the conditions students learn in — inputs, not outcomes.
Limitation
Spending doesn't always correlate with outcomes, and ratios don't capture class-level staffing. Public attendance values are withheld under correction. Existing Resources & Environment scores and overall ratings may still reflect an unverified attendance input.

How Poverty Adjustment Works

For each state, we fit a statistical regression line using every public school's poverty level (free/reduced lunch percentage) and test proficiency. This tells us: "For a school with X% low-income students, what proficiency level is typical?"

Poverty-adjusted proficiency for Grinnell Grade School Scatter plot of 1,226 Kansas schools, one dot each: the percent of students on free or reduced lunch on the x-axis and the average of each school's math and English proficiency results, in the state's latest test year, on the y-axis. A dashed straight line fitted to those schools shows the proficiency expected at each poverty level. It falls from 69% proficient at 0% free or reduced lunch to 19% at 100%. Grinnell Grade School, at 38% free or reduced lunch, is 86% proficient where the line expects 50%, 36 points above it.0%50%100%0%50%100%% Free/Reduced Lunch% ProficientExpected for FRL %+36 pts vs expected
Each grey dot is one of 1,226 Kansas schools, placed by its share of students on free or reduced lunch and the average of its math and English proficiency results in the state's latest test year. The dashed line is a plain straight line fitted to those schools: the proficiency expected at each poverty level. Our ratings draw this line from a slightly different set of schools and test results, so a school's own page can show a slightly different gap. Grinnell Grade School sits 36 points above the line, which is the gap our Adjusted Academics component rewards.

To protect against outliers, we winsorize FRL percentages at the 5th and 95th percentiles before fitting. States with fewer than 30 schools skip the adjustment entirely — the sample is too small for a reliable regression.

Each school's profile shows its expected proficiency vs. actual, so parents can see the math behind the adjustment.

Scoring by School Level

Different school levels have different educational priorities. Elementary schools focus on building foundations, while high schools prepare students for college and careers. Our weights reflect this:

ComponentElementaryMiddleHigh
Student Progress50%40%25%
Adjusted Academics30%35%25%
College Readiness——30%
Environment20%25%20%

K-12 and other multi-level schools use high school weights.

Programs & Offerings

Beyond ratings, each school page displays program data to help families understand what a school offers. This data comes from the Civil Rights Data Collection (CRDC), conducted by the U.S. Department of Education's Office for Civil Rights. Programs use official OCR public-use files where available; older rows may come through the Urban Institute Education Data Portal. AP/IB rows may also use NCES EdFacts.

What we show:

  • AP/IB Programs — number of AP courses offered, IB programme availability, AP enrollment and participation rates
  • Athletics — count of interscholastic sports offered, broken down by boys' and girls' programs, plus team counts (varsity and JV)
  • Advanced STEM Courses — which of five advanced subjects (Algebra II, Advanced Math, Calculus, Chemistry, Physics) the school offers
  • Gifted & Talented — whether the school offers a gifted and talented program

Limitations:

  • CRDC data is collected periodically. Our CRDC programs collection covers SY 2023–24. Each school row shows its recorded source and period; a program refresh does not redate teacher or chronic-absence data.
  • CRDC covers public schools only. Private school pages show AP/IB data (where available) but not athletics or support data.
  • Sports data shows the number of sports offered, not which specific sports. No school-level sport-by-sport data is publicly available.
  • We filter out obvious data quality issues: schools reporting more than 25 sports per gender are excluded as likely district-level aggregates.
  • AP participation and the IB indicator can contribute to the existing college-readiness score when available. Athletics, dual enrollment, advanced STEM subject availability, and gifted/talented indicators are supplemental information to help families evaluate fit.

Programs are displayed differently by school level. Elementary and middle schools typically don't offer AP courses or interscholastic sports, so those rows only appear when the school actually reports that data. Gifted and talented programs span all school levels.

What Our Scores Cannot Measure

No rating system captures everything that makes a school great. Our scores do not measure:

  • Teacher quality — individual teacher effectiveness isn't captured in public data
  • School culture and climate — the "feel" of a school, student well-being, and parent satisfaction
  • Arts and clubs — performing arts, visual arts, debate, robotics, and other non-athletic extracurriculars
  • Special education quality — how well a school serves students with IEPs or 504 plans
  • Safety and discipline — public data does not capture incident-level safety
  • English learner support — quality of ESL/ELL programming

A rating is a starting point for research, not the final word. We encourage families to visit schools, talk to other parents, and consider factors that no data set can capture.

Schools Without Direct Test Data

About 33,000 schools — mostly private schools and some specialized public programs — don't participate in state standardized testing. Rather than leave these schools unrated, we estimate their academic score using a geographic proxy: the average academic score of directly-measured schools in the same area. The weight reductions below describe how this proxy affects a public school's composite; private schools show the same kind of estimate, but it's informational only and isn't part of their composite (see the FAQ below for why).

City-Level Proxy (~21,000 schools)

When a school's city has at least 8 directly-measured schools, we use their average academic score as the proxy. This is our preferred tier: a city-level average captures local school quality better than a statewide figure.

Weight reduction: 20%. A city proxy academic score carries 80% of the weight that a direct score would. For example, if academics would normally be 30% of the composite, a city proxy counts for 24%.

State-Level Proxy (~12,000 schools)

When a city doesn't have enough measured schools, we fall back to the statewide average. This is less informative — a Texas state average represents 9,000+ schools spanning very different communities.

Weight reduction: 35%. A state proxy academic score carries 65% of its normal weight. The heavier penalty reflects the lower predictive accuracy of statewide averages.

These discount levels were validated through holdout back-testing: we removed scores from 10% of directly-measured schools, imputed them via proxy, and compared predicted vs. actual scores. City proxies predicted within 1.5 points on average (on a 10-point scale), and state proxies within 1.7 points — both within acceptable thresholds. Every school page clearly indicates when a proxy score is used.

Equity & Diversity

Each school page shows equity-related data including a campus diversity index, demographic breakdowns, and free/reduced lunch percentages. This information helps families understand the school community, but equity metrics are not factored into the composite score.

We made this deliberate choice because equity metrics are context, not performance. A school's demographic composition doesn't tell you how well it educates students — our poverty-adjusted academic score already accounts for socioeconomic context where it matters most: in measuring whether a school outperforms expectations.

How We Handle Missing Data

When a component cannot be calculated — for example, if a school has only one year of test data (no growth) — we redistribute the missing weight proportionally among the remaining components. A school rated on two of three possible components still gets a composite score.

Full ProfileAll components scored. Highest confidence.
Partial Profile2-3 components scored. Missing weights redistributed. Noted on the school page.
Limited DataOnly 1 component scored. Score based on very limited data — treat with caution.

Public elementary and middle schools need at least 2 scored components to receive a composite rating. Other public school levels need at least 3, and so do private schools, because those models carry more components and a rating built from only 2 of them can lean too heavily on whichever pieces happen to be available. For public schools, a component still counts toward this minimum when it comes from the city or state proxy described above, not only when it is directly measured. Private schools are scored on a different five-component model that has no academic component at all, so a proxied academic estimate never counts toward their minimum, even when one is shown on their page. Schools below these thresholds are shown as "Not Yet Rated," and so is any school with fewer than 30 students enrolled, before any component is scored: a handful of students is too few to rate fairly.

Data Sources

SourceWhat we use
NCES CCDSchool identifiers, enrollment, demographics, location, student-teacher ratios
State DOEs (46 states)Standardized test scores, graduation rates, accountability ratings
EdFacts / CRDCFederal test scores, AP/IB enrollment, civil rights data
Census ACS 5-YearNeighborhood income, education, poverty at ZIP level
Urban Institute EDPSupplementary federal data (programs, finance)
NCES PSSPrivate school enrollment, staffing, programs
Walk ScoreWalk / Transit / Bike Score for school address (Neighborhood Context section only — not in composite)
FBI UCRCounty-level Part-1 offense index (county granularity only — never per-school). Neighborhood Context section only — not in composite.

All data is publicly available and free. Schools cannot pay for better ratings. See our Data Sources page for state-by-state coverage detail.

Change Log

Version 2.1 — April 2026

  • Tiered proxy academic scoring for ~33K schools without test data (city proxy at 80% weight, state proxy at 65%)
  • Holdout validation confirming proxy discount levels (city MAE 1.5, state MAE 1.7 on 10-point scale)
  • Private school minimum components raised from 2 to 3 — prevents thin composite scores
  • Equity metrics shown as informational only, no longer displayed as a numeric score

Version 2.0 — April 2026

  • Added level-specific weights (elementary vs middle vs high)
  • Introduced poverty adjustment via state-level regression
  • College Readiness component added for high schools
  • Private school scoring model with available-field compensation
  • Growth component given primacy in elementary/middle weights

Version 1.0 — January 2026

  • Initial composite scoring model launched
  • Single set of weights applied across all school levels

Research & References

Our methodology is informed by academic research on school quality measurement:

  • Value-added models — Our growth-first approach draws on research showing that value-added measures are more informative than status measures for identifying effective schools.
  • Poverty adjustment — The use of regression residuals to compare schools serving similar populations is standard practice in education research (e.g., Downey & Condron, 2016).
  • GreatSchools methodology — We studied GreatSchools' Summary Rating methodology to understand industry standards, particularly their approach to growth weighting and equity considerations.
  • Simpson's diversity index — Our campus diversity measure uses Simpson's index (1 - Σp2), a well-established ecological diversity metric.

Questions people ask first

Why does my school have the score it has?

A school's score is a weighted average of up to four parts, each compared with public schools in the same state. To read it, start with its level, since the weights differ: growth is 50% of an elementary score, 40% of a middle school's and 25% of a high school's. Second, growth against proficiency: growth asks whether scores are rising, not how high they are. Third, the poverty adjustment, which compares academics with the rate expected for its share of low-income students. A part at the state median scores about 6.5, so a 4 means its main parts sit well below it.

What year is this data from?

Each input has its own school year. Test scores come from state education departments, SY 2024–25 to SY 2025–26 depending on the state, or federal EdFacts files (SY 2017–18 to SY 2021–22) in the 5 states whose own files are not loaded. Directory and enrollment: NCES CCD, SY 2024–25. Private schools: NCES PSS, SY 2023–24. Programs: CRDC, SY 2023–24. Neighborhood income: Census ACS 2019–2023 5-year estimates. The Data Sources page lists every state's year, and the Change Log records changes to the method.

Why can a school with high test scores score below its neighbor?

Because proficiency is only one input. In the worked example, Grinnell Grade School had 86% of students proficient where its state's line expected 50% for a school with 38% of students on free or reduced lunch. That 36-point margin is what the poverty adjustment rewards. A neighbor with the same proficiency but fewer low-income students has a higher expected rate, so it beats the line by less. If its results are also flat from year to year, growth pulls it lower, and growth is the largest weight for elementary and middle schools.

What does "growth" measure?

Growth is the year-over-year change in the share of a school's students scoring proficient on state tests, measured within that school. It asks whether results are moving forward, not where students started, which is why it is our primary signal. It needs at least two years of tests, and 2020–21 results count for less. When a school has no growth score, its weight shifts to academics. Growth is 50% of an elementary score, 40% of a middle school's and 25% of a high school's, as the weights table shows.

Why are private schools listed but not ranked?

Most private schools do not take state tests, so they have no growth or poverty-adjusted academic score to set beside public schools. They stay in our directory, but they are not ranked alongside public schools. Where enough data exists, a private school gets a separate rating from a five-part model: classroom attention, instructional commitment, diversity, community context and school profile. It is compared only with other private schools in its state and needs at least 3 scored parts. An estimated academic score on a private school page is informational and never enters that rating.

What happens when a school is missing data?

There is no penalty for missing data. When a part cannot be calculated, such as growth for a school with one year of tests, its weight is shared among the parts that can be, so a gap in public records is not scored as a zero. Public elementary and middle schools need 2 scored parts for a rating; other levels and private schools need 3. Below that, the page shows Not Yet Rated instead of a number. A public school without test results may carry an estimated academic score from its city or state average, at reduced weight, and its page says so.

Who pays for School Scout?

No school, district or advertiser pays us, and no school can pay for a rating or a higher place in a list. Every input is public data. We do not sell leads: we do not sell your personal information to realtors, schools or anyone else. Today the site has no ads and no revenue, and Layover Supply LLC, the small company that built it, covers its costs.

How do I report an error, and what happens next?

On any school page, use "See something wrong? Report an error" near the bottom, pick what is wrong and describe it. You can also write to us from the contact page. Each report is stored and sent to us, and we read every one. When our figure is wrong, we add a recorded correction that is reapplied every time the data reloads, so the fix survives the next refresh. When the government source itself is wrong, the fix has to come from the agency that published it. Leave an email if you want a reply.

Questions?

We publish the math because you shouldn't have to take a rating on faith. If you have a question about how a specific school scored, or you think we got something wrong, use the "Report an Issue" button on any school page. It comes straight to us.