How the insider-activity score is computed

Every insider-activity score on Tradune is a 0–100 summary of public SEC Form 4 filing activity for one stock. This page documents the exact formula — no black box, no hidden inputs.

The window: trailing 90 days

The score looks at Form 4 filings over the trailing 90 days only. Older filings age out of the score entirely, so a score can drop simply because time passed — that is by design: it describes current filing activity, not history.

What counts as a "buy": open-market purchases only

Only open-market purchases (Form 4 transaction code P) count as buys. Grants and awards (code A), option exercises (code M), tax withholding (code F), and gifts (code G) are excluded — those are compensation mechanics or transfers, not a decision to spend one’s own cash on the stock at the market price. Every code is listed in the Form 4 glossary entry.

The formula

score = min(distinctBuyers × 22, 66)
      + 17 if an officer bought
      + 17 if net value is positive
      (capped at 100)

Distinct buyers (up to 66 points). Each distinct insider who made an open-market purchase in the window adds 22 points, capped at 66 (three or more buyers). Breadth of buying dominates the score.
Officer involvement (+17). Added once if at least one buyer is an officer (CEO, CFO, other named officers) — people closest to operations.
Positive net value (+17). Added once if the dollar value of insider buys exceeds insider sells over the window.
Cap. The total is capped at 100.

Clusters and ratings

A cluster is flagged when two or more distinct insiders made open-market purchases within the 90-day window. The rating strings shown next to scores ("Strong cluster activity", "Cluster activity", and so on) are descriptive labels for score bands — shorthand for how much filing activity there is, nothing more.

What the score is NOT

Not predictive. It measures filing activity that already happened. We make no claim that a high score predicts future returns, and we have not published a returns backtest of this score.
Not a rating or recommendation. It is not a buy/sell/hold rating, not investment advice, and not a suggestion to follow any insider’s trades.
Not complete on its own. Insider activity is one public signal among many. Filings can be amended (Form 4/A), and sales may occur under prearranged Rule 10b5-1 plans.

10b5-1 status and change in reported holdings

Form 4s filed on or after April 1, 2023 carry a checkbox the filer uses to indicate that a transaction on the filing was made under a contract, instruction or written plan intended to satisfy the affirmative-defense conditions of Rule 10b5-1(c) (Form 4 General Instruction 10; SEC Release 33-11138). Company pages show “10b5-1 box checked” when it is checked and “10b5-1 box not checked” when it is blank. A blank box records what the filer disclosed; it does not prove that no plan exists. Filings older than the checkbox, or where it could not be read, show no tag. Held after is Form 4 Table I, column 5 — the shares the insider reported owning after the transaction — and change in reported holdings is the transaction’s shares divided by the implied holding before it (held after minus the shares for an acquisition, plus the shares for a disposition). It is computed only for non-derivative rows with a positive prior holding, from that one filer’s own direct or indirect line, and is not the change in total company ownership. Neither field affects the score.

Data source

All inputs come from public SEC EDGAR Form 4 filings (public domain), as filed. No licensed-provider figures are used in the score.

Rule 10b5-1(c) checkbox statistic: methodology

The checkbox statistic on the insider report counts how often insider open-market sales are filed with the Form 4 box stating that a transaction was intended to satisfy the affirmative-defense conditions of Rule 10b5-1(c). Current snapshot: 2026-10-16-90d, filings dated Jul 12, 2026 – Oct 9, 2026.

Which rows. Form 4 and Form 5 Table I (non-derivative) lines with transaction code S (open-market or private sale), from original filings — amendments (4/A) are left out because they restate lines the original already reported. Every Form 4 filer is a Section 16 insider (officer, director or 10% owner). One row per reported line, so a sale reported in several lots counts several times.
Three categories, nothing else. Checkbox disclosed: the filer checked the box. Checkbox not disclosed: the box is present and unchecked. Not comparable: the filing predates the checkbox requirement (filed before April 1, 2023), or the box is missing or could not be read. An unchecked box does not prove that no plan exists; it only records what the filer disclosed on that form.
The share. Share disclosed = checkbox disclosed ÷ (checkbox disclosed + checkbox not disclosed). Not-comparable rows are counted and shown separately and never enter the share. The value-weighted share uses shares × reported price, and only over comparable rows that reported a price; the number of rows without a price is shown beside it.
The box is per filing. A Form 4 has one checkbox, so every code-S line of a filing with the box checked is counted as checkbox disclosed, even if a footnote says only some of its sales were made under the plan. Footnotes are not read.
Roles. A filer with several relationships is counted once, in this order: officer, then director, then 10% owner (anyone else is “Other”). Officers are split by the title on the filing: CEO when it says “Chief Executive Officer” or “CEO”, else CFO when it says “Chief Financial Officer” or “CFO”, else other officer. Titles with “former”, “deputy” or “assistant” are never CEO or CFO, and a title naming both is CEO. “Co-CEO” and “CEO of” a subsidiary count as CEO; “Principal Executive Officer” or “Principal Financial Officer” without CEO/CFO wording count as other officer. On a joint filing by several reporting owners, the role is the first owner’s. Titles are free text, so this split is only as consistent as the filings.
Coverage. The 1,627 issuers in Tradune’s SEC Form 4 ingest at snapshot time (989 had at least one sale in the window) — not every US-listed company. The window is 90 days of filing dates and ends 7 days before the snapshot date, because our nightly ingest needs about a week to revisit every company.
Snapshots. A snapshot is taken on the 1st of each month and frozen: its rows never change after publication, and the CSV’s data rows hash to the SHA-256 printed on the report. Monthly 90-day windows overlap, and no trend is shown or claimed until at least 3 monthly snapshots exist.

Dates. The SEC’s amendments to Rule 10b5-1 (Release No. 33-11138) took effect on February 27, 2023; the Form 4 and Form 5 checkbox applies to reports filed on or after April 1, 2023. The statistic uses April 1, 2023 as its cutoff. Sources (checked October 15, 2026): SEC final rule, Insider Trading Arrangements and Related Disclosures (33-11138) and SEC, Modernizing Rule 10b5-1 Insider Trading Plans.

The statistic describes what was disclosed on past filings. It is not a signal about any stock and not investment advice.

← Recent insider buying Cluster signals → Guide: insider cluster buying →

This score summarizes SEC Form 4 filing activity over the trailing 90 days. It is descriptive, not predictive — not a rating, recommendation, or investment advice. Data: SEC EDGAR (public domain), as filed; amendments may revise figures. Tradune is not a registered investment adviser.

Data as filed; amended filings (Form 4/A) may revise figures. Sales may occur under prearranged Rule 10b5-1 trading plans.