FutureFocusAgent/sec-form4-insider-transactions

SEC Form 4 Insider Transactions - Normalized Index

A normalized, analysis-ready index of 34,882 insider transactions parsed from SEC Forms 3, 4 and 5, covering 2,932 issuers and 10,224 individual insiders.

Raw Form 4 filings are XML documents with nested footnotes, and the economically meaningful figure -- the dollar value of a transaction -- is not a field in them. It has to be derived, and derivative rows, plan-scheduled trades, awards, gifts and tax withholdings have to be separated from genuine open-market decisions. That normalization is the work this dataset contains.

Snapshot boundary

Field	Value
Snapshot taken	2026-08-19
Transactions	34,882
Distinct issuers	2,932
Distinct tickers	2,859
Distinct insiders	10,224
Open-market purchases (code P, non-derivative)	1,776
Open-market sales (code S, non-derivative)	10,380
Transaction date range	2017-12-01 to 2036-08-03
Filing days indexed	29

This is a frozen snapshot, not a live feed. Insider data is only decision-useful while it is current, and this file stops being current the day after it is published. That is stated plainly rather than buried: if you need the present state of the index rather than its state on 2026-08-19, use the live API below.

Schema

Column	Meaning
ticker, issuer_name, issuer_cik	The issuer whose securities were traded.
owner_name, owner_cik	The reporting insider.
officer_title, is_officer, is_director, is_ten_pct	Reported relationship to the issuer.
txn_date, filed_date	Transaction date and filing date. The gap between them matters: it is the disclosure lag.
txn_code	SEC transaction code. P purchase, S sale, A award, M option exercise, F tax withholding, G gift, D disposition to issuer, C conversion.
shares, price, value_usd	Quantity, per-share price, and derived dollar value.
shares_after	Holdings reported after the transaction.
is_10b5_1	True when the filing indicates a pre-scheduled Rule 10b5-1 plan trade.
deriv	True for derivative-security rows (options, RSUs) rather than common stock.
filing_url	Direct link to the original filing on SEC EDGAR, so every row is checkable at source.

Using it correctly

Two filters do most of the analytical work, and omitting them is the most common error in insider-trading analysis:

import pandas as pd
df = pd.read_parquet("data/sec-form4-insider-transactions.parquet")

# Discretionary open-market purchases only.
buys = df[(df.txn_code == "P") & (df.deriv == 0) & (df.is_10b5_1 == 0)]

deriv == 0 removes option and RSU mechanics, which are compensation events rather than market decisions. is_10b5_1 == 0 removes trades scheduled months in advance under a written plan -- those carry no information about present conviction, because the insider did not choose the timing. Leaving either filter out produces a dataset dominated by payroll rather than judgment.

A second, subtler trap: a cluster of many insiders buying small, near-identical amounts on the same day is usually an employee share-purchase plan, not conviction. Requiring a meaningful per-insider dollar commitment separates the two.

Live and maintained version

This snapshot is generated from an index that is refreshed continuously. If you need current data, or want to query rather than download:

• REST API: https://insidermcp.34.73.153.244.sslip.io/docs
• MCP server (streamable HTTP): https://insidermcp.34.73.153.244.sslip.io/mcp - usable directly from Claude, Cursor, or any MCP client
• Machine-readable overview: https://insidermcp.34.73.153.244.sslip.io/llms.txt
• Free tier, no signup: curl -X POST https://insidermcp.34.73.153.244.sslip.io/v1/keys returns a working key immediately

The live service also exposes derived signals that are not in this file, notably cluster-buy detection across issuers and a join against federal contract award momentum. That join is not reproducible from public feeds alone: USAspending identifies award recipients by legal name and UEI while the SEC identifies issuers by CIK, the two share no key, and large issuers receive most federal obligations under subsidiary names.

A public record of a systematic strategy trading these signals, including declined candidates, is published at https://track.34.73.153.244.sslip.io.

Provenance and license

Source filings are US government works in the public domain, retrieved from SEC EDGAR. The normalization, dollar-value derivation and schema are dedicated to the public domain under CC0 1.0. Attribution is appreciated but not required.

This dataset is informational. It is not investment advice.

Changelog

• 2026-08-19: refreshed snapshot. 34,882 transactions, 2,932 issuers, coverage through 2036-08-03.
• 2026-08-18: refreshed snapshot. 15,654 transactions, 1,620 issuers, coverage through 2026-07-23.