What School Rankings Get Wrong — And What We Can (and Can't) Fix

You're on Zillow at 11pm. You found a house you can almost afford in a neighborhood you've never been to. You click the GreatSchools rating: 8/10. Okay, good school. You move on.

But what does 8/10 actually mean? What's it based on? And who decided that number should be sitting right there next to the listing price, quietly shaping one of the biggest decisions you'll ever make?

We started asking those questions. The answers bothered us enough to build something different. But building something different comes with an obligation: to be honest about what we can actually fix, and what we can't.

Who pays, and what that buys

GreatSchools is a nonprofit, and most of its money is philanthropy. Its 2023 tax return reports $9.07 million in revenue, $5.72 million of it from contributions and grants, and about $1.97 million in a category the return labels "Licensing." The return doesn't say what that licensing covers, and we're not going to guess. Separately, and visibly to anyone: GreatSchools ratings sit next to listing prices on Zillow and Realtor.com. A rating that travels inside a real estate listing gets read as a property feature, whether anyone meant it that way or not.

Niche is a private company, so nobody outside it knows its revenue mix. Its partnerships page invites schools to claim their profile, update their own data, and "upgrade to Premium" for enrollment marketing. The same page invites real estate sites to "add Niche school grades to your real estate content." So a school you're comparing may also be a Niche customer. We went looking for a published statement about whether paying can move a grade and didn't find one, in either direction — so we're not going to tell you there's a firewall, and we're not going to tell you there isn't. We're telling you what the arrangement looks like from outside, which is all we can see.

Neither of those is corruption. They're alignment questions, and alignment questions apply to us too. Being paid by parents wouldn't make us clean either. A parent-funded tool can still profit from anxiety, chase prestige, and paywall its own caveats. So instead of asking you to trust our business model, here is what we commit to:

• No paid placement. No school, district, or company can buy a rank, a badge, or a spot in a list.
• No school can change its score. Schools can send us corrections to factual data, and we want them. They can't move a number.
• Commercial relationships never touch the ranking. Whatever we sell and to whom, the scoring code doesn't know about it.
• The methodology and the source data are public. Every weight, every formula, every source file, at /methodology.
• Corrections are documented. When we get something wrong, we fix it and publish what changed and when.

Hold us to those five, because you can check every one of them. You can't check a business model.

And here's what those two do better than we do. GreatSchools covers all fifty states; we cover one. Niche has reviews from parents and students, and private school coverage, and we have neither. Both publish real methodology pages with real weights on them, so this isn't a story about secrecy. What we do differently is degree. We publish the correlations that embarrass our own formula, the sentences our software writes, and the dimensions we can't justify yet. Read on and you'll hit several of them.

What the data hides — and what we surface

Here's the thing that made us build this.

California reports CAASPP results in four levels, not two: Standard Exceeded, Met, Nearly Met, Not Met. The two big national sites start from a single number instead — the share of students who met or exceeded. That's a real choice, and it costs you something.

Take the 80 California elementary schools where 74% to 76% of students met or exceeded the standard this year. Across those 80 schools, the share who actually exceeded it runs from 25% to 59%. Same headline number, very different schools underneath it.

Do the division and the gap is stark. At the school with 59% exceeded, about four out of five kids who cleared the bar are in the top level. At the school with 25%, only about one in three are — the other two thirds sit in "Standard Met," the band directly above the line. We can't see where inside a band any child landed, and we won't pretend to. We can see which band, and that alone splits these schools in two.

Both big sites publish what they do here, so we can quote them rather than speculate. GreatSchools' Test Score Rating says: "We begin with the percentage of students performing at the 'proficient and above' level on the state's standardized exams." Niche's Public Elementary & Middle School Academics grade puts 60.0% of the weight on "State Assessment Proficiency," with 25.0% on the district's academics grade, 10.0% on student-teacher ratio, and 5.0% on its own parent and student surveys. Neither published method separates the kids who cleared the bar by a wide margin from the kids who barely cleared it. We surface that split, because we think the ceiling matters as much as the floor.

Two schools can post the exact same met-or-exceeded number and still look completely different once you break out who's in the top level — that's worth asking a school about directly, because the number alone doesn't tell you why. For the full breakdown, including what each level actually means and how to read it at your school, read our deep dive on the exceeded vs. met distinction.

We also surface growth. The number we score is a cohort comparison: this school's third graders in 2023 against this school's fifth graders in 2025, in test scale points. Same school, two years apart, and roughly the same group of kids — though not exactly, because families move, and they move most at schools serving unstable housing. Read it as the school's trajectory, not any one child's. A school whose cohort gains ground is doing something the raw proficiency number hides. A school whose cohort loses ground despite high overall scores is worth asking about; it may be coasting on kids who arrived already ahead.

And we show chronic absenteeism, the share of students missing 10% or more of school days, and suspension rates. Both are on every profile. Neither tells you as much as it looks like it does, and we get into why below. We haven't found either one inside the headline score on the rating sites we've looked at.

We have opinions. Here they are.

The Scope Score isn't a neutral aggregation of everything California reports. It reflects specific judgments about what matters.

We believe exceeded scores deserve extra weight. The proficiency rate stops counting the moment a student crosses one line, so it tells you nothing about how far past it anyone got. The exceeded rate reopens the top of the distribution. It carries the single largest weight in the elementary Scope Score. It still doesn't tell you what the school added. A school's exceeded rate tracks its poverty rate at r = −0.78. (That's a correlation: 0 means no relationship at all, and −1 would mean a perfect one, running in the opposite direction. So −0.78 is strong.) The richer the school's families, the higher the exceeded rate, almost every time. Growth is the one dimension that doesn't work that way, which is the next thing we have an opinion about.

We believe growth is worth more attention than raw scores get. We checked how closely each part of our formula tracks a school's free-lunch rate. The two proficiency measures track it hard, at −0.78. Cohort growth barely tracks it at all, at −0.13. So a school's average score mostly tells you which families enrolled there, and its growth doesn't.

Be careful about what that buys, because we were careless about it here for months. A weaker link to income is not proof that growth measures what the school contributed. It rules out the most obvious alternative explanation and nothing more. The research that actually validates growth measures against lottery admissions — the cleanest test anyone has — was run on adjusted growth models in Denver and New York City, not on our cohort comparison. Nobody has validated ours. What we can defend: growth is the one number on our page that isn't mostly a proxy for the neighborhood, which is why it's worth looking at.

Our growth is also a weaker instrument than the state's, and you should know that California publishes a better one. CDE's growth model follows matched individual students from one grade to the next in grades 4 through 8, rather than comparing two cohorts. GreatSchools publishes a Student/Academic Progress Rating built on state-reported growth data, and gives growth its heaviest weight when it has it — 0.45, against 0.27 for its other components. We import California's growth file too, and we show it on every school profile — we just don't put it in the Scope Score. The reason is a trade we'd rather show you than hide: the state's measure tracks family income about three times more closely than ours does, −0.36 against our −0.13. Give that up and the score quietly becomes more about the neighborhood again. On top of that, California has published only one year of it, so nobody yet knows how much a school's figure bounces — and at the elementary level it covers grades 4 and 5 only, while our score spans 3 to 5. So each measure is better at something, and we'd rather show you both and let them disagree than pick a winner on your behalf. Where they disagree about your school, that's a good thing to ask the school about.

We believe chronic absenteeism belongs on the page, and we're careful about what we say it means. A school where 20% of students miss more than 10% of days is a school where a lot of kids aren't there to learn, and that's worth knowing. It is not evidence that the school did something wrong. A rate that high reflects a high-need student body at least as often as it reflects a struggling school — absence tracks poverty, health, and transportation, and the best available work finds that raw school-level rates tell us almost nothing about a school's own effect on attendance (Liu, Fordham 2023). We count it, inverted so lower is better, at a light weight, and we cut that weight for exactly this reason. Cutting it reduced the problem without solving it. Until we can compare each school against demographically similar peers, the raw rate still penalizes schools serving high-need families, and we say so on /methodology.

We believe suspension rates say something about how a school responds to kids who struggle — but a raw rate doesn't say it cleanly. The strongest study here compares students assigned to schools with stricter discipline in one metro area, and finds real effects on arrests, incarceration, and education (Bacher-Hicks, Billings & Deming, 2024). What it establishes is that discipline practice varies between schools and matters. It doesn't establish that any given school's raw rate is free of who enrolls there — our own subgroup breakdowns exist because it isn't. So we include the rate, inverted, and treat it as a question to take on a tour rather than a verdict.

We also include ELPAC proficiency: the share of a school's English learners scoring at the top overall level, "Well Developed." That level satisfies the state's language criterion for being considered for reclassification — one of four required criteria, with the other three decided locally. (Students with significant cognitive disabilities take the Alternate ELPAC, where Level 3 serves the same role.) No study we've found validates a school's raw ELPAC rate as a quality measure on its own. We include it at a modest weight because schools demonstrably shape how English learners progress, and the rate partly reflects who enrolls: a school with many newly arrived students will show a lower number almost no matter what it does.

These are opinions. Reasonable people could weight things differently. You can read the exact weights and all our reasoning at /methodology. We think you should be able to adjust the weights yourself, too. We haven't built that yet.

Three lenses, not one verdict

The Scope Score is built around three lenses:

Academic Performance — exceeded rates, met-or-above rates, grade 3→5 growth, and, for high schools, graduation and college readiness. Growth belongs here in our view: a school whose cohorts gain ground is doing something different from one whose cohorts hold level.

School Climate — chronic absenteeism, suspension rates, and ELPAC proficiency. Three rates: how often kids are in the building, how often the school removes them, and how far its English learners have gotten in English. This is the closest public data comes to how a school runs day to day, and it's still only a set of rates. A low suspension rate can mean restorative practice or it can mean a student body that rarely triggers discipline, and the number won't tell you which. Treat this lens as the questions worth asking on a tour, not as a finding about how a school treats kids.

Community Profile — demographics, equity gaps across student groups, and per-pupil spending. This lens is context, not a scored input. We show it alongside the score so you can read the numbers against who walks in the door.

On spending: we pulled California's own SACS current-expense file for 2024-25 and ran the numbers ourselves, so treat what follows as our arithmetic on the state's data, not a figure the state publishes. Across the 622 districts with at least 500 students in average daily attendance, current expense per student runs from about $12,200 to about $41,000, and the middle district lands near $19,900. Statewide, total current expense divided by total attendance comes to about $21,200 per student.

That spread is enormous, and it doesn't line up with scores the way you'd expect. Manhattan Beach, South Pasadena, Palos Verdes Peninsula, and El Segundo all spend between roughly $16,000 and $18,700 per student — below the state figure — and their elementary schools average Scope Scores from the low 70s to the low 80s. A district spending $25,000 per student on a school scoring 40 is a different story from a district spending $16,000 on the same school. Read the spending as background you hold the score against, and ask where the money goes.

The Scope Score combines the first two lenses into a single number, because single numbers are useful for comparison. The number should be a doorway, not a verdict. On profiles we lead with a band — Strong, Solid, Developing, Needs Support — because the zone is the part you can act on, and the precise score is the detail underneath it. Every profile shows the component breakdown, so the summary can't hide the story.

What we can't see

Here's what the data genuinely doesn't capture, and where our score will mislead you if you rely on it alone.

Teaching quality. One of the most important variables in a child's school experience is whether their specific teacher is skilled, warm, and good at reaching kids like them. Test scores correlate loosely with average teaching quality over time, but they say nothing about the teacher your kid will actually have. There's no substitute for talking to parents whose kids are already at the school.

Arts, music, athletics, and extracurriculars. These don't show up in CAASPP data. A school with a strong arts program or a legendary running team offers something real that our score can't see. If that matters to your family, and for some kids it matters enormously, the Scope Score won't help you here.

IEP support and special education quality. California does publish CAASPP results for students with disabilities where group sizes allow it, and we show them on school profiles. But a test score is not an IEP. Whether services actually get delivered, whether accommodations follow your child into every classroom, whether inclusion is real or nominal, whether anyone returns your calls — none of that is in any dataset. For these families the difference between two schools is often enormous and almost entirely outside what we can see. Ask other IEP parents at the school. We don't have a better answer than that.

Parent and community culture. Whether the PTA is engaged, whether new families get welcomed, whether the community feels warm or competitive — you learn that by talking to current parents and showing up to a school event. It doesn't fit in a database.

Your specific kid. A school that's exceptional for a kid who needs structure might be wrong for a kid who needs room to wander. A school with high exceeded scores might have an intense, high-pressure culture. A school with middling scores might have the warm, unhurried classroom that's exactly right for your child. Only you know your kid, and only a visit will show you the fit.

How to use the data well

The Scope Score is most useful as a starting point that narrows the field, not a destination.

Look at patterns, not snapshots. A school that's been in the 75th percentile for several years is telling you something more reliable than a school that spiked this year after a run of average ones. Our historical scores are all recomputed under the current formula, so year-to-year comparisons on our site are apples to apples.

Compare schools that are comparable. A Scope Score of 65 in a high-income district where most schools score 70-80 means something different from a 65 in a district where most schools score 45-55. We show a statewide percentile and the district average for exactly this reason. Use those, not just the raw number.

Use the data to generate visit questions, not to skip visits. If absenteeism is high, ask current parents why. If growth is strong but exceeded scores are low, ask what the school does for kids who are already ahead. If suspensions are high, ask how the school handles conflict. Let the data make you a smarter visitor, not a more anxious one.

Trust your tour, for the things a tour can show you. When you walk into a school you pick up real information: how the front office greets you, how kids move through hallways, whether classrooms look alive or exhausted. That's real, and it's about fit, culture, and how the building treats people. It is not a reliable read on whether kids learn more there. The Gates Foundation's Measures of Effective Teaching project had trained raters score more than 23,000 videotaped lessons, and found that a single lesson is a weak read on a teacher even for an expert with a rubric (Ho & Kane, 2013). You are not going to beat that on a forty-minute walkthrough, and you don't have to. Use the tour for what only a tour can show you, and the data for what only data can.

Why we start with one state, not fifty

National rating sites face a hard problem: compare schools across fifty states with fifty different tests. Both solve it the same reasonable way, and both say so — GreatSchools rates a school "relative to other schools in the same state," and Niche converts each state's proficiency rate into a within-state percentile before comparing anything. What survives that translation is a proficiency rate. California's four performance levels don't, even though California publishes them.

We take the opposite approach. We show you the most useful data your state actually produces, in its native form, instead of compressing it into something comparable to a school in Ohio and not very useful for your decision.

California's CAASPP data is unusually rich. Four performance levels, not pass/fail. Reported by grade, by subgroup, by school. That's what lets us show the exceeded vs. met split, the growth trajectory, and the rest of the detail that disappears in a score built to work everywhere at once.

Starting with one state isn't a limitation we're embarrassed about. It's what lets us go deeper. When we expand, we'll do the same thing: show what that state's data actually says.

Hold us accountable

We checked whether the exceeded split and growth were measuring the same thing. They aren't quite. Across 5,372 California elementary schools, a school's cohort growth and its exceeded rate correlate at r = +0.21 — a weak positive link, which means the two share a little under 5% of their variation and go their own way with the rest. That is not independence and we won't call it that. It's enough to say the two are mostly not redundant, which is why we track growth as its own line rather than folding it into the score for proficiency.

Now the unflattering half. Growth is by far our least stable measure. Compare each school's growth against a different cohort's growth at the same school and the two agree at r = 0.57. Exceeded rates score 0.97 on the same test. That gap is the reason we cut growth's weight in August 2026 rather than raising it again. At the old weight, more than a quarter of elementary schools landed in a different band depending on which cohort we happened to measure — a parent reading the band was reading the coin flip as often as the school.

We won't say "half of growth is noise," because a correlation can't split signal from noise that cleanly, and some of that year-to-year movement is real change at real schools. What we'll say is narrower and worse-sounding: a single year of our growth number is unreliable enough that you should not act on it alone, and we've weighted it accordingly.

That cut cost us something, and here it is. Moving weight off growth and back onto the proficiency measures made the Scope Score track family income more closely, not less — |r| against free-lunch share went from 0.718 to 0.747. We took that trade knowingly, to buy stability: the score's own year-over-year reliability went from 0.905 to 0.934. We think a parent is better served by a number that doesn't move under them. Reasonable people would weight it the other way, and the modeling is re-runnable.

Absenteeism came out worse than we hoped. We wanted a signal independent of test scores. It tracks met-or-above at −0.65, and the higher a school's free-lunch share, the higher its absence rate — r = +0.58. That's why we cut its weight, and why the raw rate still isn't fixed.

What we still can't tell you: whether our growth measure predicts anything you'd care about later, like eighth grade math or high school graduation. Nobody has validated that for our measure, including us.

If our score doesn't match what you know about a school from experience — if the data says one thing and parents in the neighborhood say another — we want to know. That divergence is information. It might mean our methodology is missing something.

If you're a parent trying to understand California schools, start here. If you want to see exactly how we score schools, read our methodology. If you think we're wrong about something, tell us. We mean it.

Corrections, August 14, 2026: this post previously said California doesn't publish student-level growth data. It also described competitor methodologies as not fully public, which is wrong — both publish weights. And it argued our incentives were clean because parents fund us; we've replaced that with five commitments you can check.

Corrections, August 15, 2026, after a line-by-line source check of this post. We import California's growth model now and show it on every profile, so the text above says that rather than "it's on our list." We had described our own growth measure two different ways in the same article; the scored number is the cohort comparison, and only that description survives. We claimed growth and exceeded rates were "close to independent" at r = +0.21 — that's a weak positive link, not independence, and the "94% unexplained" line that followed it was doing work correlation can't do. We said roughly half of a year of growth is noise; a correlation can't make that split, so we cut the claim and kept the actionable part. We said a high absence rate means the school "has a problem" — the evidence says a high rate signals a high-need student body at least as often, and the section now says that. Two spending paragraphs cited an NCES file that doesn't contain the figures we quoted; those are now our own arithmetic on California's SACS file, labeled as ours. The classroom-observation study we cited covers early-childhood centers for ages 0 to 6, not K-12 tours, and we've replaced it with one that observed school-age classrooms. We asserted Niche has a firewall between payment and grades; we can't find Niche saying that, so the claim is gone and not replaced with its opposite. And we described GreatSchools' licensing revenue as the reason its scores appear on Zillow — the tax return doesn't say that, so we now report only what it does say.

See what your schools really look like