What gets screened, how the score is built, and what the number does not mean.
Each issue is dated for a Monday and covers papers published from the prior Saturday through the Friday before it, by publication date, with online-first counting. Roughly twenty-six journals are screened each week, spanning general medicine, oncology, cardiology, neurology, surgery, radiology and translational science. Reviews, editorials, comments, case reports, errata, preprints, and meta-analyses without new data are excluded.
A paper reaches the ranked list only if it can be tied to a listed company. Preclinical work qualifies only when a public company owns the asset and the result defines a platform.
The Market Sensitivity Score runs 0 to 100 and is the sum of five components, each scored 0 to 20. Every published issue shows the breakdown, so the total can always be taken apart.
Two of these run against intuition. Attention scores high when a result is under-covered, because a finding already on every front page has less left to resolve. Novelty scores low when the data were already presented at a conference, even if the journal publication is new.
It measures how likely a paper is to matter to a listed company. It is not a view on whether that company is worth owning, and it carries no direction: a high score attaches equally to a result that helps a company and one that damages it. ResearchAlpha states no price targets, makes no recommendations, and does not predict which way a stock will move.
Each entry lists the companies the result bears on, with the role each plays — sponsor, competitor, supplier or partner — and how concentrated that company's value is in the asset. Exposure is marked high only when the asset is a top-three value driver.
An entry is marked high confidence only when the company link is evidenced: sponsor employees in the author list, or the company's own disclosure. Where the link is inferred — a device maker whose hardware the trial challenges, with no documented sponsorship — the entry is marked medium and the inference is stated rather than hidden. Anything weaker than that does not enter the ranked list; it goes to "also noted" or is dropped.
Every DOI is confirmed against publisher metadata, with the title and journal required to match. Every number quoted must appear in the abstract or the paper itself — no extrapolation and no rounding to a rounder number. Anything found only in a press release is labelled as such. A field that cannot be verified is left empty rather than guessed: missing is acceptable, wrong is not.
The archive builds forward from issue one, a week at a time. Issues are never written for weeks that have already passed. Catalyst proximity and attention are precisely the components hindsight corrupts, and a record fed by them would report that a paper was ranked highly and then moved, when the ranking was partly caused by knowing about the move. The publication launches thin rather than full and unfalsifiable.
Each entry names the benchmark it should be read against — XBI for biotech, IHI for medical devices — so that a company move can be separated from a sector move. A performance record measuring ranked entries against those benchmarks is planned but not yet published; when it exists it will live here, including the entries that scored high and resolved to nothing.