Search any school's acceptance rate and you'll get three different numbers on the first page. None of the sites is lying. They're answering different questions and labeling them identically.
Here's the anatomy, using a case we checked line by line.
The same school, four different numbers
For Cornell's Early Decision round we found, across sites reporting on the same admissions cycle: 11.6%, 17%, an admit count of 1,661, and Cornell's own filing showing 1,161 admits from 9,973 applications.
The 11.6% is 1,161 ÷ 9,973 — Cornell's published counts, divided. The others come from real sources too. They're just not the same source, the same year, or the same denominator.
The four things that actually differ
1. Which cycle. This is the biggest one. A school's Common Data Set for 2024–25 describes students who applied in fall 2024. Some sites publish that as "2026 acceptance rate" because those students enroll later; others label it 2024. Two sites can print different numbers, both correct, three years apart in meaning — and neither says which year on the page. This is why every figure we publish carries the cycle label next to it, not in a footnote.
2. Which pool. "Acceptance rate" can mean the whole first-year pool, or ED only, or everyone outside ED, or admitted-and-enrolled. At Cornell those are 7.8%, 11.6%, and other figures depending on how you slice. Comparing a school's ED rate against another school's overall rate is the most common error in this genre — and it looks like a real comparison.
3. Which document. The federal IPEDS survey, the College Scorecard, the school's own Common Data Set, and the school's press release are four different documents that disagree with each other regularly. They have different definitions, different collection dates, and different amounts of school discretion. A press release saying "most selective class in history" is a marketing artifact; C21 of the CDS is a filing.
4. Transcription. Then there's the boring one: someone copied a number wrong, and the next twelve sites copied that. The 1,661-vs-1,161 gap is what a digit swap looks like once it propagates.
How to check any number in ninety seconds
- Find the cycle label. If a page states an acceptance rate with no year attached to the data (not the URL, not "2026" in the headline), you cannot use it. That alone eliminates most results.
- Ask which pool. ED, RD, overall, or enrolled — if the page doesn't say, assume it's the flattering one.
- Go to the source. Search
[school name] common data set. Section C1 is the overall pool; C21 is early decision. The document is a PDF on the school's institutional research page, and it takes about a minute to read the two lines you need.
That last step is the whole game. We link the original filing on every figure we publish — partly for you, and partly because it keeps us honest: a linked number is one we can't quietly get wrong.
Why nobody does this
Because reading a 40-page PDF for one ratio is miserable, and because for most people the number is not worth ninety seconds. It becomes worth it when you're deciding where a binding application goes, or where a real tuition bill lands. At that point the difference between 11.6% and 17% is not trivia — it's the difference between two strategies.
Figures above from Cornell's published 2024–25 Common Data Set and public reporting we compared against it. Results are data-based analysis and do not guarantee admission.
See your list scored on sourced numbers — free.