A family I worked with years ago narrowed a cross-country move down to two towns, then settled it in an afternoon. One town's schools showed 9s and 10s on a popular ratings site; the other's showed 5s and 6s. They picked the 9s, signed on a house, and felt they had done their homework. What they had actually measured, without realizing it, was the median household income of the two zip codes. The ratings had told them which neighborhood was wealthier. They had not told them which schools were better at teaching.
That gap between what a rating appears to say and what it actually measures is the most important thing to understand before you let a single number steer a decision this large. The ratings are not fraudulent. They are built on real data and disclosed methods. But a number can be accurate and still answer a different question than the one you are asking.
What the Major Ratings Actually Measure
Start with the most widely cited consumer rating, the GreatSchools Rating. It is a 1-to-10 summary score, and per the site's published methodology it is a weighted blend of up to three themed components. The Test Score Rating reflects how well students meet the state's grade-level proficiency standards. The Student Progress Rating measures academic growth, how much students improve on state tests from one year to the next. For high schools, a College Readiness Rating folds in graduation rates, advanced coursework, and college-entrance results. The methodology gives the stronger of the progress and test-score components the largest base weight and distributes the rest across the others, adjusting for how much data a school has.
One detail matters for anyone reading an older explainer. GreatSchools reports test results broken out by student income and by race or ethnicity, and for several years an equity component fed those subgroup gaps into the overall score. As of the site's 2025 methodology update, those student-group ratings are still calculated and shown, but the documentation now states plainly that they do not factor into the overall GreatSchools Rating. If you read a guide that describes an equity input baked into the headline number, it is describing the older model.
Niche takes a different shape. It issues A-through-F letter grades and combines two kinds of input: hard public data and survey opinion. The academics grade, which carries the most weight, draws on federal data from the U.S. Department of Education, state and national test scores, graduation rates, and college outcomes. Layered on top are millions of reviews and survey responses from students, parents, and teachers, which feed the culture, teacher, and overall grades. The blend is the point and the catch. A school's letter is part measured performance and part reported sentiment, and the two are not separated for you on the badge.
Then there are the ratings your own state issues. Most states run an accountability system that assigns each school a letter grade, a star count, or a tier label. The ingredients vary, but the common ones are proficiency on the state test, academic growth, graduation rate, and, in many states, progress for English learners and for specific student groups. These state ratings are the official record, the one tied to funding and intervention, and they are often more rigorous than the consumer sites. They are also the least uniform: an A in one state and an A in the next are computed differently and are not comparable.
The Problem Hiding Inside a Proficiency Score
The single biggest ingredient in most ratings is proficiency, the share of students scoring at grade level on the state test. And proficiency, measured as a level, is tightly bound to family circumstance.
The clearest evidence comes from the Educational Opportunity Project at Stanford, led by sociologist Sean Reardon, which assembled a national database of roughly 350 million test scores. Their summary of it is blunt: average test scores correlate strongly with socioeconomic factors like income and poverty. As the project puts it, average scores tell you very little about how much children are actually learning in a given school. A high average can mean students arrived already ahead, which is largely a function of the resources at home and in the surrounding community, not of the instruction inside the building.
If proficiency tracks income, then a rating built mostly on proficiency is, to a large degree, a wealth signal wearing the costume of a quality signal. This is why ratings tend to map so neatly onto a region's housing prices. The score is partly re-describing the neighborhood you are already looking at. A wealthy district can post excellent proficiency numbers while adding very little to what its students would have achieved anyway, and a high-poverty school can be doing extraordinary work and still show middling proficiency, because its students started further back.
Growth Asks a Better Question
The fix that researchers and the better rating systems have converged on is growth. Instead of asking where students stand, growth asks how far they moved: how much did a given cohort improve on the state test from one year to the next. The Stanford team's central finding is that learning rates, their term for growth, are a far better measure of school quality than average scores, because they correlate much less with family income and come closer to isolating what the school itself contributed.
Growth is not a perfect instrument. It is noisier year to year than a proficiency level, it can be unstable for very small schools, and a school can post strong growth while its students are still, in absolute terms, behind grade level. A child who learns two years' worth of material in one year is doing something remarkable; that child may also still be below the bar. Both facts are true and both matter, which is exactly why growth should sit beside proficiency rather than replace it. The point is not that growth is the only honest number. It is that a rating leaning almost entirely on proficiency is leaning on the number most contaminated by income.
The Income-Adjusted Question: Is the School Beating Its Odds
There is a third question worth asking, and it is the one we built into allk12's data. Take proficiency and ask not whether it is high or low, but whether it is higher or lower than you would expect given the school's student population.
That is what our BeatsExpectations score does. For each state, we run a regression of school composite proficiency on the share of students who qualify for free or reduced-price lunch, the standard proxy for low income. The line that fits that data is the expected proficiency for any given poverty level. A school that scores above its predicted line is beating expectations; one below is falling short. The score is the residual, the distance between what a school posted and what its student demographics would predict. Schools in the top slice of that residual within their state are flagged as outperforming. We exclude specialized populations, such as state schools for the deaf or blind and therapeutic placements, because state ELA and math proficiency is not a comparable yardstick for them.
What this surfaces is the high-poverty school quietly outrunning its odds, the one a raw proficiency rating buries near the bottom and a wealthy coasting school the same rating crowns. It is not a measure of growth, and it does not pretend to be; it is an income-adjusted read on level. But it answers a question no raw rating asks: given who walks through this school's doors, is it doing better or worse than the data predicts? You can see the score on each school's page and on our best-schools rankings, which lead with it. For context when you read any single proficiency figure, the national average composite proficiency sits around 47 percent. A school at 60 percent in a wealthy area may be unremarkable for its demographics; a school at 45 percent in a high-poverty area may be a standout.
How to Read a School Without Getting Fooled
None of this means ignore ratings. It means read them as one input with a known bias, and triangulate. When you look up a school, on this site or anywhere, read three things together rather than fixating on one composite badge.
- Proficiency · where the students are. The share at grade level on the state test. Useful, but remember it is the number most tied to family income. Compare it to the state average and to the roughly 47 percent national composite, not in isolation.
- Growth · how fast they are moving. Year-over-year improvement, the closest thing to a measure of what the school adds. A school with average proficiency and strong growth is often teaching more effectively than a high-proficiency school that is coasting.
- An income-adjusted measure · whether the school beats its odds. A beating-expectations or similar residual score tells you if the school is over- or under-performing its demographics. This is the one that catches what raw ratings miss in both directions.
Read together, those three rarely all point the same way, and the places where they diverge are exactly where the interesting information lives. A high proficiency score with flat growth and a below-the-line income-adjusted result is a comfortable neighborhood school, not necessarily a strong teaching one. Mediocre proficiency with strong growth and a school beating its expectations may be the better bet for a child who needs momentum.
If you want to see how this plays out across a single state, the school and score pages for somewhere like California let you sort the same set of schools by proficiency and by the income-adjusted score and watch the rankings rearrange. The schools that hold their position near the top under both lenses are the ones worth a closer look. The browseable index of every state's schools lives at our schools directory.
The family I mentioned moved anyway, and their kids did fine; most kids do. But they made a six-figure decision on a number that was, in large part, a restatement of a zip code. The honest version of school ratings is less tidy than a single 9 or a clean A. It is three numbers that disagree with each other, and the disagreement is the point. Do not buy a house on one composite figure.
Sources
GreatSchools: Ratings methodology report
GreatSchools: About our ratings
Harvard Graduate School of Education: First National Database of Academic Performance Launches (Stanford Educational Opportunity Project / SEDA)
allk12 BeatsExpectations: per-state OLS regression of composite proficiency on free/reduced-lunch share; national average composite proficiency approximately 47 percent (allk12 data, NCES-based).
