Every screener tells you who owns a healthcare name. This one tells you what kind of money it is.
An index fund, a multi-strategy pod running a three-week catalyst trade, a PIPE financier who bought at a discount, and a physician-founded specialist who spent four months on the mechanism all render as identical rows on a normal ownership table. They mean completely different things.
The money is classified first, and everything else — sorting, scoring, filtering, comparison — runs on that classification. Schedule 13D/G filings are carried alongside, because they land within days of a 5% crossing rather than 45 days after a quarter ends.
The register, by kind of capital
Every institution in the dataset carries one of eight classes and an explicit scoring weight — index money is weighted at zero, specialist conviction at full. Counts below are institutions, not dollars.
Classifies every holder
Eight classes of capital, each with an explicit scoring weight. Index money counts for zero. A pod counts for a quarter of a specialist.
Names the instrument, not the intent
A 13F files one line per instrument, so a manager holding common, a call and a put on one name files three. Each is labelled as filed. Ownership figures count the long claim only — a put is not read as belief — and nothing is badged “short” or “hedge”, because a put can be a hedge, a collar or a financing structure and a filing does not say which.
Shows what actually changed
Each fund’s book differenced against its own prior-quarter filing — opened, added, trimmed, and the exits most trackers throw away.
Carries a catalyst calendar
Every dated catalyst across every company in scope, grouped by when it is due and filterable by your own book. Each row on it cites the filing the date came from.
Says when a company has stopped trading
Deregistrations, delistings and acquisitions, taken from the issuer’s own filing index. Pending dates move to whoever bought the asset. A quote provider will serve the last trade of an acquired company indefinitely; this will not.
Surfaces Schedule 13D/G
The only ownership data that is not 45 days stale. A ⚡ marks anything filed after the quarter closed.
Finds lonely conviction
Large specialist positions almost nobody else holds — which a breadth-rewarding score systematically underrates, and the tool says so.
Flags crowding
Where specialists collectively own enough of an issuer that the door out is narrower than the position implies.
Shows its sources
97% of research profiles cite an SEC filing and none rests on a newswire, a republisher or an aggregator. Where something cannot be sourced, the record says so rather than reading as verified.
Refuses numbers it cannot stand behind
Where a reverse split has put holdings and quote on different share bases, ownership percentages are suppressed with a warning rather than shown wrong.
Reconciles every book to the filing it came from
All 36 tracked 13F books match their information table line for line and to the dollar on value. Getting there found two books short by $103m between them and an “exit” that never happened — a position missing from a transcription reads exactly like a position that was sold.
Keys every company to its SEC identifier
Each security carries the CIK EDGAR knows it by, linked to that company’s filing history. It is the one identifier here that stays valid after a company stops trading, which is why the filings, the price checks and the delisting gate all resolve through it rather than through a ticker.
Draws the events on the price
Five years of daily closes on every security page — 1,842 series and 2,024,491 sessions — with that company’s own dated, sourced events marked on the line, and 699 names carry both. Daily rather than quarterly for one reason: a 13F backtest is quarterly by nature, so quarter-end closes answer every question about return and none about a readout, because a result on 12 May and a quarter-end on 30 June are the same point. A security that has stopped trading shows its final per-share figure and the filing that established it, not a line.
Not here yet
This is in early beta, so it is worth being plain about what is missing rather than letting you find out. Everything below is designed and specified; none of it is half-built behind a flag. Next means the work is unblocked and scheduled; planned means it is waiting on a decision or on data the dataset does not yet hold.
Ownership over timenext
Five years and 2,536 filed books are already loaded and validated. Every past event on a security already shows the 13F either side of it. Still to come: how one manager’s position moved quarter by quarter, how the cohort rotated between therapeutic areas, and how long a position has actually been held.
Lonely conviction, over timenext
Today a lonely position is a single-quarter photograph. It cannot tell you whether a manager has been alone in it for eight quarters and adding, or is alone because two other books sold out from under them. Those are opposite situations.
Build a rule and replay itplanned
Compose a hypothesis out of the classification layer — “the top ten oncology specialists’ five most-overlapped holdings” — and replay it quarterly across the corpus. It would state what a rule would have owned, never that you should own it.
Long and short exposureplanned
The instrument is labelled today; the aggregate is not. This would show how much of a name the tracked cohort is long through common and warrants against how much it is short through puts — as a description of the positions, not a claim about anyone’s view.
Open it
An email address opens it. No password, no confirmation step, nothing to install.