LiveVol. III · No. 04Cycle 2026
Cohortly
HARV · admit rate 3.4% YALE · reader avg. 12 min/file MIT · apps 25–26 38,940 STAN · admit rate 3.7% PRIN · narrative-tier 1 admits 78% IVY+ · profiles benchmarked 14,302 T20 · features modeled 42 HYPSM · median admit GPA 3.96 IVY · median admit ECs 9.2 IL. · free reports run today 417 HARV · admit rate 3.4% YALE · reader avg. 12 min/file MIT · apps 25–26 38,940 STAN · admit rate 3.7% PRIN · narrative-tier 1 admits 78% IVY+ · profiles benchmarked 14,302 T20 · features modeled 42 HYPSM · median admit GPA 3.96 IVY · median admit ECs 9.2 IL. · free reports run today 417
The Admissions Quarterly · Honest Questions
Updated · Cycle 2026
§ V · Honest questions

What this is — and isn't.

Everything we'd want a friend to ask before paying — what the score means, how refunds work, what we do with your data, and the limits of the model. If we missed yours, the contact desk goes to a real inbox.

§ I. The product

What this isand isn't.

We are not an admissions oracle. We are a benchmarking instrument. The difference matters more than people think.

Will this predict if I get in?

No, and anyone who claims to is selling something worse than this. Admissions is a committee process with too many degrees of randomness to predict. We score where you sit relative to admits, and tell you which levers actually move that.

Is it AI? Is it accurate?

Models do the pattern-matching, but every feature is benchmarked on real anonymized admit data. We publish error bars on each score and flag low-confidence reads explicitly. No false certainty.

How is this different from a Common Data Set chart?

The CDS gives you the median admit's GPA. We model 42 features, including narrative coherence and EC tiering — the things that decide your file in the actual reader's room.

What do I need to start?

A four-minute structured intake — stats, course list, activities, awards, demographics, and one optional essay sample. No credit card on the Free tier. Applicants are hashed at intake.

§ II. Pricing

The fine print, in plain English.

Pay once per profile. No subscription, no auto-renewal, no upsell to consulting. Refunds are honest.

Why $15 — and not a subscription?

Most users run one or two profiles in their cycle. A subscription would punish that. Pay once, get the report, leave.

Do I have to pay before I see anything?

Never. The composite score, percentile, strengths, and weakness are visible on Free with no card required. The sample report shows exactly what you get.

What if my profile changes?

Premium includes one free re-run within 30 days of purchase — finish that EC, retake the test, then re-score against the same cohort.

Is the Free analysis a stripped trial?

No. Free is its own report — a real composite, a real percentile, a real diagnosis. It just stops short of the levers (ranked ECs, narrative dissection, school-specific weaknesses, roadmap) that take serious modeling to produce.

How do refunds work?

Reports are non-refundable once generation begins, given the computational cost of producing them. Refunds are issued in three cases only: failed report generation, duplicate charges, or technical malfunction. Email billing with your order ID.

Are there bulk rates?

Yes — counselor 3-packs, practice 10-packs, cohort 25-packs, and quoted school bundles. See bulk pricing for the per-report ladder.

§ III. Methodology

How we read a profile.

Forty-two features, benchmarked on actual admits. Every score carries an error bar; low-confidence reads are flagged.

What is the reference set?

500+ verified accepted profiles, contributed voluntarily and anonymized at intake, plus public sources (Common Data Sets, IPEDS) for baselines. Cohort sizes for school-specific fits are disclosed on the Premium PDF.

Why 42 features, not more?

Adding features past the point where they have independent signal makes scores look precise without making them more accurate. Every feature in the schema has been ablation-tested for marginal predictive value against the cohort.

What does "narrative spike" mean?

A measure of how legibly your profile tells one story to a committee reader. High-spike files compound across activities, awards, and essays toward a single domain; low-spike files read as a checklist of unrelated achievements.

How current is the model?

The methodology version is shown on every report (currently v3.0 · Cycle 2026). When the model retrains or the schema changes, the version bumps and old responses remain reproducible for at least eighteen months.

§ IV. Data & privacy

Anonymized at intake.

Your file is hashed before it reaches the model. We don't sell data. We don't share intake content. Sub-processors are listed in the privacy notice.

What identifiers do you store?

Names, addresses, phone numbers, schools-of-origin, and free-text identifiers are stripped or hashed before scoring. The applicant record is keyed by an opaque hash; the model never reads identifiers. See the privacy notice.

Can I delete my account and data?

Yes. Email the privacy desk with subject data request and we export or delete within seven days. Deletion is irreversible and removes both the anonymized scoring record and the credential.

Do you sell or share my data?

No. We do not sell data, ever. We do not share intake content with third parties for advertising or training. Sub-processors (Clerk, Vercel, Stripe, model inference providers) are listed in the privacy notice.

Is the Premium PDF anonymous?

The PDF is generated from the anonymized record. Anything personally identifying is something you typed in yourself.

§ V. Advisors & institutions

For counselors and schools.

Bulk packs are self-serve up to 25 reports. Department-scale rollouts are quoted with a data-access agreement. The institutional API is in private development.

Do you offer counselor bundles?

Yes. Self-serve 3-, 10-, and 25-packs, plus quoted school and district bundles. Per-report rates ladder from $10 down to $6. See bulk pricing.

Are creator partnerships available?

Yes — affiliate revenue share, co-produced cohort reads, and sponsored composites for admissions creators. Standards and formats are on creator partnerships.

Can I integrate Cohortly into my own product?

Programmatic access is in private beta — endpoints, principles, and the waitlist live on data access. Aggregate cohort statistics only; never individual records.

Who built this?

A solo indie team. We don't sell consulting. The disclosures page covers independence and conflicts of interest.

Still have a question?

A real person reads every message. No ticket maze, no chatbot, no escalation tier — just the desk that fits your question.

Reach a real inbox See the sample report