How We Read the Common Data Set for You.
Original statistical analysis of the 2024–25 Common Data Set universe — 135 institutions read line by line — turned into the handful of patterns that actually change a STEM applicant's school list.
Original statistical analysis of the 2024–25 Common Data Set universe — 135 institutions read line by line — turned into the handful of patterns that actually change a STEM applicant's school list.
This is the single most important idea in the data. At a test-required school, every admitted student submitted a score, so the published 25th–75th percentile band is the real band of the admitted class. At a test-optional school, only some admits submit — and the ones who submit are precisely those whose scores help. So the published band reflects the strong half of the class, biased upward.
Across our universe of 75 top universities, test-required schools showed a median submission rate of 52% (mean 57%); test-optional schools, a median of 31% (mean 34%). Test-required submission rates run about 1.7× higher. At elite test-optional schools the gap is sharper still — WashU showed only 29% of admits submitting an SAT, Vanderbilt 27%.
Industry research puts the gap between submitters and non-submitters at test-optional schools at 30–80 SAT points. So a published "1500–1570" where only ~30% submit hides a true admit floor closer to 1430–1470. A strong STEM applicant at the printed 25th is usually above the real median, not at the bottom of it.
The practical rule we use: at a test-optional school, submit only if a score is at or above the school's submitter 50th percentile — roughly the midpoint of the published band. Not submitting at a test-optional school is read as "the score is probably below the band," so a real score anywhere inside the band should go in.
When a test-required (or high-submission) school shows a band, that band is honest. A 1530 means the 75th percentile of admits, not of self-selected submitters. For STEM applicants who test well, these are the most strategically legible targets. Schools with a submission rate at or above ~60% publish essentially the real class band:
| School | US News | SAT 25–75 | Submission | Math 75th |
|---|---|---|---|---|
| MIT | 2 | 1520–1570 | 83% | high |
| Brown | 13 | 1510–1560 | 61% | 800 |
| Georgetown | 24 | 1400–1540 | 78% | 780 |
| U Florida | 30 | 1320–1470 | 79% | 750 |
| Georgia Tech | 32 | 1370–1530 | 77% | 790 |
| FIU | 97 | 1070–1250 | 90% | 620 |
Several elite schools have reinstated test-required policies for recent cycles — among them MIT, Harvard, Brown, Dartmouth, JHU, UPenn, Caltech, and some Cornell colleges — and public flagships such as Georgia Tech, U Florida, and UT Austin require scores by mandate. For a student who tests well, these are where a real number is read at its real value.
Across 89 schools with both sub-scores, the Math 75th percentile runs a median of +20 points above EBRW; STEM-leaning schools cluster at +40 to +50. For applicants from a Chinese math curriculum who reliably score 780–800, that asymmetry is structural leverage — strongest at exactly the engineering-heavy schools many STEM applicants target.
| School | EBRW 75th | Math 75th | Gap | Policy |
|---|---|---|---|---|
| Wisconsin | 730 | 780 | +50 | TO |
| NJIT | 710 | 760 | +50 | TR |
| Georgia Tech | 750 | 790 | +40 | TR |
| Purdue | 720 | 760 | +40 | TR |
| UIUC | 750 | 790 | +40 | TO |
| CMU | 770 | 800 | +30 | TO |
At 17 elite schools the Math 75th is a perfect 800 — Princeton, Harvard, Stanford, MIT-tier and beyond — meaning ~25% of the class scored 800. There, an 800 is the median, not a distinguisher. The signal is saturated; the rest of the file has to carry the weight.
The implication: emphasize STEM-specialized schools where a math score is a positive signal rather than a baseline — Georgia Tech, CMU, Harvey Mudd, RPI, NJIT — and remember that rank within a major matters more than overall rank. CS at CMU, UIUC, or Georgia Tech outranks CS at most Ivies, regardless of the headline number.
For an international applicant, TOEFL is usually the binding constraint — a 1500 SAT means little behind a 92 TOEFL at a school that wants 105. But the folklore of "you need a 100" is wrong on the data: across 100 schools with a hard minimum, the median is 80, not 100. The minimum is real, but lower and far more school-specific than most families assume. We map it in five tiers.
Brown (105, with the dataset's highest IELTS minimum), Columbia, Northeastern, Pitzer (106).
The mainstream top-20 tier — UChicago, JHU, Cornell, CMU (102), Notre Dame, UCLA, Georgetown, NYU, and more.
Princeton, Harvard, Stanford, Yale, Duke, Northwestern, Dartmouth, UVA — operational threshold ~100–105, just unpublished.
WashU (90), Tufts, UNC, Lehigh, Case Western, Wake Forest, Purdue (88), RPI (88), and peers.
All 9 UC campuses (80), U Florida (80), UT Austin & Georgia Tech & UIUC (79), CU Boulder (75) — most public flagships.
Cheaper and faster, but not universal — UT Austin, UGA, Stony Brook, and TCU do not accept it; Georgia Tech only as a supplement.
A 110+ closes this gate entirely; 100–109 opens all but the very-high tier; a 90–99 still reaches the no-minimum elites plus the standard and accessible tiers. Note the new TOEFL scale launched January 2026 — many schools publish both old and new minimums, so verify which scale applies at submission time.
These archetypes assume an applicant with TOEFL 100+, Math 780+, EBRW within the listed band, strong-but-not-extraordinary activities, and competent essays. They are a starting frame, not a prescription — every real list is tuned to the student.
All nine University of California campuses are test-blind — a perfect 1600 cannot help. There the levers are course rigor, the four Personal Insight Questions, and recalculated GPA. Apply to three to five UCs strategically rather than all nine, and consider the official Transfer Admission Guarantee path at the six campuses that offer it.
A Common Data Set has ten lettered sections (A–J), each a standardized survey question every US college answers the same way — which is the point: it makes apples-to-apples comparison possible. It is a regulatory filing, not a marketing document; when the brochure and the CDS disagree, the CDS is the one to trust. Using Yale's 2024–25 filing as the worked example, four sections carry most of the information value.
Yale's undergraduate college is genuinely small — about 1,700 per class, the size of a large liberal-arts college — sitting under a graduate-heavy research university. Of 57,517 applicants, 2,227 were admitted and 1,554 enrolled: a 3.9% admit rate with a ~70% yield. Read the residency line too — Yale leaves the international breakdown blank, but unhooked international admit rates run closer to 2–3%.
The most important page for an applicant. Yale marks eight factors Very Important — course rigor, class rank, GPA, essay, recommendations, extracurriculars, talent/ability, and character — and rates standardized tests only "Considered" (a test-optional artifact). Crucially, level of applicant's interest is Not Considered: campus visits and opened emails do nothing. For an international applicant who cannot move geography or legacy, the essays and recommendations are the entire controllable game.
Yale's submitters showed an SAT composite of 1480 / 1530 / 1560 (25/50/75) — but 86% of admits submitted some test, so this band is close to the true class band, unlike a 29%-submission school. Aim for the submitter 50th (~1530) to be confident a score helps the file.
Often skipped, often the most consequential. Yale is need-blind for all applicants including internationals — one of only a handful of US schools — meets 100% of demonstrated need, and awarded an average international package of about $83,878. Only 12% of graduates borrow at all, averaging ~$7,265 total. For a family that genuinely cannot afford the ~$95k sticker, this is the number that matters more than any score band.
Read B1 + C1 (size, selectivity), then C7 (what the school values), then C9 (the band, with self-selection in mind), then H if you need aid, then I-3 (class sizes) and J (degree distribution — what the school actually is academically). Skip the bureaucratic A-sections and the definitions appendix. That captures ~80% of the value.
A condensed timeline for an applicant targeting a fall entry. The dates that quietly end lists are the ones discovered late — so we hold them from the start.
| Window | What happens |
|---|---|
| Jr. spring | First SAT and TOEFL attempts; AP exams; line up a substantive summer (real research or a project with visible output beats a brand-name camp). |
| Summer | The substantive activity itself; begin the Common App; open conversations with counselor and recommenders. |
| Aug–Oct | SAT / TOEFL retakes if needed; research current CDS bands; finalize a list of 8–12 with a 3-3-3 reach/match/safety split. |
| Nov 1–30 | Early Decision / Early Action deadlines (most Nov 1); UC application by Nov 30; CSS Profile for need-aware aid. |
| Dec | Early decisions release; if admitted to a binding ED, withdraw the rest; otherwise finalize Regular Decision. |
| Jan–Apr | RD deadlines (Jan 1–15); decisions release through late March; F-1 visa interview prep. |
| May 1 | Enrollment deposit due; then SEVIS fee, F-1 interview, I-20, orientation. |
TOEFL by November of senior year (scores valid two years); SAT by October at the latest (reporting takes 2–3 weeks); CSS Profile on the same day as the Common App; and book F-1 interview slots by April — they fill over the summer.
Blank data is not no data. A school that hides its band is telling you something too.
— The Tian2 EditorsThis analysis is built on the 2024–25 Common Data Set and shifts modestly year-over-year; we re-read the new filings each November. ← Back to STEM College Counseling