How to Track Insider Buying Using SEC Form 4: 2026 Examples

Research cut-off: 3 November 2026. Financial data are filing-period figures; any market prices below state the exact date.

SEC Form 4 insider trading analysis begins by classifying each transaction code. An executive receiving shares as compensation is not the same as buying them in the open market; shares withheld for taxes are not necessarily a discretionary sale. Clarifo's Form 4 aggregation for Microsoft Corporation (Nasdaq: MSFT) covering 1 July–3 November 2026 counted 46 non-derivative transaction rows in the returned window, including 19 classified sales and no classified purchases. This is a dated dataset extract, not a claim about every insider transaction ever made.

Know what you are counting

Form 4 reports changes in beneficial ownership for specified insiders. The transaction-code column is central: open-market purchase and sale codes should be separated from equity awards, exercises, gifts and tax-related withholding. An automated alert that treats every share reduction as 'insider selling' can distort the size and intent of trades. Similarly, a share award is not an affirmative decision to invest cash in the stock.

What the Microsoft transactions actually show

The window included approximately $75.4 million of classified Microsoft insider sales across the returned rows. There were four distinct sellers and no distinct buyers in the summary output. As a concrete audit point, a Form 4 for Microsoft CFO Amy Hood records several sales dated 13 October 2026, with the filing dated 14 October. Another Form 4 records a share sale by Satya Nadella on 1 September. Each source should be read with its footnotes and reported ownership after transaction.¹²

Cross-check with a second issuer

For NVIDIA Corporation (Nasdaq: NVDA), Clarifo's same 1 July–3 November window counted 36 rows, including 22 classified sales, five tax-withholding rows, four awards and five gifts, with no classified purchase rows. One September Form 4 for Mark Stevens records an issuer-insider sale. These observations describe filings returned under the specified filters. They are not evidence that either company's future share price must fall.³

Why dollar totals can mislead

A transaction valued at millions of dollars may be small compared with an insider's holdings. Consider transaction size relative to ownership after the trade, whether it repeats, whether an automatic trading arrangement is disclosed and whether multiple genuinely independent executives acted in the same direction. Clustered discretionary purchases can be different from one routine compensation-related sale, but neither proves future returns.

Build a reproducible insider-trading screen

Record issuer ticker and CIK, reporting insider, role, transaction date, filing date, code, shares, reported price, direct or indirect ownership, security type and explanatory footnotes. Set a fixed lookback window; compare like with like across companies. For officers, directors and large beneficial owners, the relationships may overlap—avoid adding reporter classes together unless definitions are exclusive.

Use Form 4 as one layer, not an investment signal

Combine transaction patterns with the latest 10-K or 10-Q, capital allocation, share dilution, and company-specific events. Ask whether insider activity adds evidence to an investment thesis or merely reflects existing compensation and diversification arrangements. A careful AI answer includes links to the actual Form 4 filings, labels transaction types and explicitly states the observation window and coverage limitations.

Audit a Form 4 before treating an aggregate as a signal

A published aggregate is only as useful as its classification rules. In our Microsoft observation window, a Form 4 covering CFO Amy Hood includes transactions on 13 October 2026. The filing shows three sale rows at reported prices around $545 per share, including one for 991 shares at approximately $546.51. A second filing records Satya Nadella's September 2026 sale. These are concrete primary-source examples, not evidence that every row in a 46-transaction summary was independently verified by the examples alone.^{Form 4}

The summary separates 19 sales, 14 awards, 12 tax-withholding transactions and one gift. Their count totals 46, but a multi-line filing can generate more than one transaction row; it would be wrong to describe this as 46 distinct insiders or 46 independent investment decisions. The reported roughly $75.4 million sale total is an aggregation of those classified sale entries, not a measure of company-wide insider conviction.

Why the filing footnotes change interpretation

Transaction codes are a necessary starting point, not sufficient explanation. Review whether the security is derivative or non-derivative, whether shares were held directly or indirectly, whether the reporter had a pre-arranged trading plan and whether reported shares changed due to vesting or tax withholding. A Form 4 footnote may qualify the apparent price, ownership category or legal basis of a transfer. Do not invent those circumstances if the footnote is not supplied.

For NVIDIA, the same window showed 22 classified sales and no classified purchases in the returned summary. One filing records Mark Stevens's sale of 1,356,000 shares on 18 September 2026 for approximately $297.9 million. It illustrates why dollar size should be put in context: analysts need remaining beneficial ownership, other transactions and the trading-plan disclosures before assigning an investment motive.^{Form 4}

The strongest bullish and bearish inferences are both limited

A bullish analyst might note that there were no open-market purchase entries in the specified extract, but the absence of purchases is not automatically bearish and may reflect compensation practices, insider windows or other reasons not established here. A bearish analyst might emphasize large reported sales, but those also cannot by themselves distinguish discretionary pessimism from preplanned diversification. Neither interpretation follows mechanically from the table.

To make the result independently reproducible, the research output should provide the exact 1 July–3 November 2026 observation dates, issuer CIK, non-derivative filter, transaction-code mapping, the list of accessed Form 4 identifiers and any exclusion rules. A short summary with only two example source links does not independently prove the entire aggregate. We present the aggregate transparently as a Clarifo screening observation and use primary filings to substantiate individual rows.

The editorial test is whether a reader can reconstruct the observation window, distinguish 'sales' from withholding and awards, and see why the evidence supports no simple forecast of future share performance. That is a stronger use of insider data than ranking companies by gross sale value alone.