No black box. Here's exactly how it works.

Your readiness score is NOT the output of an opaque, trained machine-learning model. It is a deterministic scoring engine: a published set of rules that map your academics (GPA, test scores, curriculum rigor), extracurricular depth, leadership evidence, research signals, and narrative coherence onto eight dimensions. The same inputs always produce the same score, and every dimension is shown to you with the reasons behind it.

Benchmarks come from a verified registry of university data points (median GPA / SAT / ACT, acceptance rates), each entry carrying its source URL and the date we verified it. Nothing in the scoring path is generated by a language model: AI writes narrative explanations on top of the scores, it never decides the scores themselves.

The scoring engine is guarded by a regression suite of golden student profiles (45+ tests) that runs on every code change. If a change would alter what a known profile scores or matches, the build fails and a human reviews it.

Every scholarship row carries a last-verified date, shown in the product. An automated hygiene job link-checks every active opportunity daily and flags anything that 404s, redirects, or changes — flagged rows are pulled from matching until re-verified.

New opportunities are added in curated, source-cited batches (most recently 23 programs on 2026-06-11; 53 active programs total). A nightly job verifies that every active row remains covered by the semantic search index, so matching quality cannot silently decay.

We deliberately track a smaller, verified catalog rather than scraping thousands of low-quality listings. Every program in the database is real, long-running, and checked — a 10,000-entry database that is 30% dead links would cost you application time we are not willing to waste.

A 'match' means your profile passes every eligibility rule we can verify: grade level and citizenship/country restrictions are HARD gates — if you fail either, the scholarship is never shown to you, no matter how well the rest of your profile fits. This is enforced in code and covered by the regression suite.

What we cannot verify (essay quality, interview performance, year-to-year jury taste), we do not pretend to. Each match shows its reasons and an eligibility score instead of a false certainty — a match means 'you clear the published bar and should consider applying', never 'you will win'.

Where we estimate probabilities (acceptance odds in the Future Simulator), the estimate is computed from published admit rates and benchmark distance, displayed as a calibrated range, and labeled as an estimate. We publish the method, not just the number.

Your data is never sold, never shared with advertisers, and never used to train third-party models. Documents you upload are used for exactly one purpose: extracting the achievements and academic signals that power your own analysis.

Storage is access-controlled per user (row-level security): your rows are readable by you and the service that builds your reports, nothing else. You can delete your data at any time — deletion removes your profile, analyses, documents, and derived signals.

The full legal detail lives in our privacy policy. If anything on this page and the policy ever disagree, the stricter protection applies.