Every open-market insider buy and sell we have read from SEC EDGAR, cleaned into one CSV with a documented schema. The filings are public record; the work here is parsing, normalising and keeping them. Take it.
Newest filings in the file: Friday, 9 October 2026. Rebuilt every four hours.
Direct link: https://www.insiderscanner.net/data/insiderscanner-form4-transactions.csv — stable, so a script can fetch it on a schedule.
| Column | Type | Meaning |
|---|---|---|
filed_date |
date | The day the Form 4 was filed with the SEC. |
transaction_date |
date | The day the trade happened, as reported by the filer. Earlier than filed_date; a Form 4 is due two business days after the trade. |
ticker |
string | Exchange ticker as carried on the filing. |
company |
string | Issuer name as filed. EDGAR stores these in capitals. |
insider_name |
string | The reporting person, free text as filed. Spelling varies between filings by the same person. |
insider_title |
string | Their stated relationship to the issuer: officer title, director, 10% owner, or a combination. |
transaction_code |
string | Form 4 Table I code. Only P and S appear here. |
transaction_type |
string | Derived, not filed: "Open-market buy" for code P, "Open-market sell" for code S. |
shares |
number | Share count for the transaction, as reported. |
price |
number | Price per share, as reported. |
value |
number | shares × price. Derived, and rounded by neither party. |
acquired_disposed |
string | A for acquired, D for disposed, as filed. |
accession_no |
string | The SEC accession number, 18 digits, no dashes. Leading zeros are significant — read this column as text, not a number. |
issuer_cik |
string | The issuer's SEC Central Index Key, zero-padded to ten digits. Blank where the filing did not carry one. |
Source. SEC EDGAR's daily index is read every four hours, and each Form 4 in it is fetched and parsed. Nothing is scraped from a third party and nothing is bought in.
What is included. Table I
transaction codes P and S only — open-market purchases and
sales. Grants, awards, option exercises, gifts, and transfers to trusts are
excluded: they are not decisions to buy or sell at a market price, and mixing them
in is the most common way insider data misleads.
Amendments. Form 4/A is excluded. An amendment restates a filing already carried under its own accession number, so counting both would double a company in any cluster measure.
Duplicates. Rows are unique
on accession_no + transaction_date + shares. The
accession number alone would not do: one Form 4 legitimately reports several transactions.
Coverage is not uniform, and this is the honest caveat. The file records what has been read, not everything ever filed. Collection used to sample a recent-filings feed and took in a median of about 24 filings a day; since moving onto the daily index in October it takes in about 120. Days before that change are real but thinner. Treat the file as a growing sample with a known break, not a census.
Known limits. Insider names
are free text as filed, so one person may appear under spelling variants. Prices and share
counts are the filer's own figures, uncorrected. accession_no has significant
leading zeros — parse it as text, or a spreadsheet will quietly eat them.
The underlying filings are United States public records and nobody owns them, this site included. Whatever rights might exist in the compilation — the parsing, the normalising, the keeping — are dedicated to the public domain under CC0 1.0 Universal.
Use it for anything. Commercially, in a product, in a paper, in a model. No permission, no attribution, no conditions, no need to tell us.
A link back to insiderscanner.net is appreciated and is not a condition of anything. CC0 could not make it one, and we would not want a licence to ask on our behalf — so we are asking.
No warranty, as CC0 itself states. It is parsed from public filings and may carry errors from the filings or from the parsing. Check anything that matters against EDGAR before relying on it. Nothing here is investment advice.
A CSV rebuilt every four hours covers most research. If you need queries, filters, incremental pulls or a higher cadence, we are looking for a small number of design partners to tell us what to build — what you would automate, which fields you need, and what shape the responses should take. No cost, no commitment, and no card: the point is to avoid building the wrong thing.
Prefer to browse? Insider buying today · the full scanner · company index