58% of Insider Selling Is Noise — Here Is How to Find the Part That Matters

Start With the Number That Breaks Most People's Intuition

Over the trailing 90 days, across 3,200 US-listed companies, insiders sold $69.9 billion of stock and bought $2.7 billion. That is a ratio of roughly 26 to 1. Selling touched 1,163 tickers; buying touched 462.

If you take that ratio at face value, corporate America is in a permanent state of panic. It is not. The ratio is an artifact of how executives are paid, and understanding why is the foundation of reading insider data at all.

Why selling always dwarfs buying

• Insiders receive stock; they purchase cash. A senior executive might get 70% of their compensation in equity. Converting some of it to money is how they pay a mortgage. Buying, by contrast, requires taking after-tax salary and voluntarily putting it at risk in an asset they are already massively exposed to.
• Concentration is a genuine problem for them. An executive whose net worth, salary, and pension all depend on one company has a risk profile no financial adviser would endorse. Selling is prudent diversification, and boards increasingly encourage it.
• Taxes force sales. When restricted stock vests, it is taxable income at vest. Many insiders must sell a portion simply to pay the resulting bill — these appear as “sell to cover” transactions and are purely mechanical.

So the base rate of insider selling is enormous and carries almost no information. The job is to find the exceptions.

How to Read a Form 4 Without Being Misled by It

Everything in this article is derived from one document type, and almost every popular misreading of insider activity comes from not knowing how that document is laid out. It takes about ten minutes to learn and it permanently improves how you read insider headlines.

The three forms

• Form 3 — the initial statement, filed when someone first becomes an officer, director or 10% owner. It is a snapshot of what they already hold, and it contains no transaction at all. A Form 3 is not a purchase, though it is regularly reported as one.
• Form 4 — the workhorse. Due within two business days of the transaction. This is where essentially all usable information lives, and the tight deadline is what makes the data timely enough to be worth screening.
• Form 5 — the annual catch-all, filed within 45 days of fiscal year end for small or exempt transactions and for anything that should have been on a Form 4 and was not. A transaction appearing on a Form 5 that belonged on a Form 4 is a late filing, and a pattern of them is a governance observation in its own right.

Table I and Table II are not the same thing

This is the distinction that matters most. A Form 4 has two tables, and conflating them produces nonsense.

The classic error: an executive exercises options and immediately sells the resulting shares. Table II shows the exercise, Table I shows the sale. A naive aggregator counts the Table I sale as a bearish open-market disposal, when in substance the insider converted compensation to cash and their economic exposure may be unchanged or higher than before. That is why our pipeline keys on the transaction codes rather than on the dollar totals.

One structural limitation to keep in mind: Form 4 covers officers, directors and holders of more than 10% of a class. It does not cover most institutional selling, which arrives quarterly and with a 45-day lag on Form 13F. Insider data is a narrow, fast window, not a view of who is selling a stock overall.

The 10b5-1 Plan: The Single Most Important Thing to Check

A Rule 10b5-1 plan is a written instruction an insider sets up in advance: sell 10,000 shares on the fifteenth of every month, or sell 25,000 shares if the stock trades above $80. Once it is adopted, the trades execute automatically. The insider does not decide anything on the day of the sale.

The point of the plan is legal protection. Because the instruction was written while the insider had no material non-public information, executing it later is an affirmative defence against an insider trading charge. For investors, this has a direct consequence:

Since the SEC's 2023 amendments, these plans carry a mandatory cooling-off period — generally 90 days for officers and directors — between adoption and the first trade, and directors and officers must certify they are not aware of material non-public information when adopting one. Form 4 now has an explicit checkbox indicating the transaction was made under a plan intended to satisfy the rule.

Where to find it on the filing

1. Open the Form 4 on SEC EDGAR for the company in question.
2. Look immediately below the signature and explanation area for the 10b5-1 checkbox. If it is checked, the transaction was pre-scheduled.
3. Read the footnotes. They are the most informative part of the document and almost nobody reads them. A footnote will often state the plan's adoption date — which lets you check whether the plan was adopted suspiciously recently.
4. Check the transaction code in the table. S is an open-market sale, P is an open-market purchase, M is an option exercise, A is a grant or award, and F is shares withheld to pay tax. An M followed by an S on the same day is an exercise-and-sell, which is compensation, not a view.

In our universe, applying the plan filter removes 58% of total sell value immediately. Everything that survives is a sale somebody actually chose to make, in the moment, knowing what they knew.

Why the 2023 Rule Changes Made This Analysis Possible

For twenty-odd years after Rule 10b5-1 was adopted in 2000, the plan defence had a well-documented weakness: it was easy to abuse, and nearly impossible to audit from outside. Plans could be adopted and begin trading the same week. They could be cancelled at will — and because cancelling a plan is not itself a trade, cancellation was effectively invisible. An insider could run several overlapping plans and keep whichever one happened to pay off. None of it had to be disclosed.

Academic work on this was consistent and unflattering: sales executed under 10b5-1 plans systematically preceded price declines, with the effect concentrated in plans adopted shortly before trading began and in single-trade plans. The defence was doing real work for people who were not entitled to it.

The SEC's amendments, effective through 2023, closed most of that. The five changes that matter for reading filings:

1. Mandatory cooling-off periods. Directors and officers must wait the later of 90 days after adoption or two business days after the next quarterly results are disclosed, capped at 120 days. Other employees wait 30 days. This single change removes the “adopt a plan on Monday, sell on Friday” manoeuvre that made the old regime hollow.
2. A good-faith requirement that runs continuously. The plan must not only be entered into in good faith — it must be operated in good faith. Cancelling or modifying a plan opportunistically can now cost the insider the defence on the trades that already happened under it.
3. No overlapping plans, and one single-trade plan per year. Directors and officers generally cannot run concurrent open-market plans, and may rely on the defence for only one single-trade plan in any twelve-month period. This closes the portfolio-of-plans strategy where only the winner is exercised.
4. A certification requirement. On adoption, directors and officers must certify they are not aware of material non-public information and are acting in good faith. It creates personal accountability at the moment the plan is written.
5. Quarterly disclosure. Item 408 requires companies to disclose in each 10-Q and 10-K whether directors or officers adopted or terminated a 10b5-1 plan during the quarter, with the material terms. Plan terminations are now visible — which was previously the biggest blind spot in the entire area.

There is a second-order effect worth naming. Because the rules made plans more rigid and more expensive to game, more insider selling now runs through plans than before — which is part of why the discretionary residual is only 42% of sell value. The reforms shrank the haystack. That is what makes the needle findable.

The Four Tests That Separate Signal From Compensation

Stripping out plan sales is step one. The remaining discretionary selling is still mostly routine. Four further tests narrow it to something worth a second look — these are the same conditions our screen applies across 432 sell candidates to arrive at a shortlist.

Test 1: Is it clustered?

One executive selling is a personal financial event. Four selling within thirty days, outside their plans, is a pattern. Independent personal circumstances do not synchronise. Clustering is by far the strongest of the four tests because it is the hardest to explain innocently.

Test 2: What percentage of their stake went?

This is the test almost every headline gets wrong. “CEO sells $20 million of stock” is unreadable without a denominator. If the CEO holds $400 million, that is 5% — routine diversification. If they hold $60 million, it is a third of their position, and it means something quite different. Always convert the dollar amount into a percentage of holdings before reacting to it.

Test 3: Is it a break in behaviour?

An executive who sells a steady tranche every quarter tells you nothing when they sell again. An executive who has not sold in two years and suddenly liquidates a third of their stake has changed their mind about something. The first sale in twelve or more months is a meaningful flag; the twelfth consecutive quarterly sale is not.

Test 4: Who is selling?

Role matters, and not in the way most people assume. In our 90-day snapshot, insider roles resolved as follows across 783 lookups:

Directors dominate the count, and their trades are the least informative — a non-executive director attends a dozen meetings a year and often knows less about operational detail than a divisional manager. The CFO is the role worth watching most closely, because the CFO sees the quarter's numbers forming before anyone outside the company does. A CFO breaking a long pattern of not selling is the highest-information single event in this dataset.

What Survives All Four Tests

Running those filters across the universe is brutally selective. From 432 sell candidates checked, 56 tickers raised at least one flag, and only 12 cleared the bar for a cluster-sell designation. Here are all twelve, ranked by a severity score that weights the number of unscheduled sellers, the share of each stake sold, and whether the sale broke a long quiet period.

Every row's “unscheduled sellers” figure is the detector's own count, quoted from its reason string. Two rows carry a parenthetical: for Elastic and Asset Entities the detector counted six unscheduled sellers, but we could only attach a verified role and stake percentage to five. We show the detector's number rather than the number we could profile, because the detector is counting the actual filings and the profiling step is a separate enrichment that can fall short — a name that cannot be resolved to a role is still a person who sold. Reporting the smaller number would quietly understate the cluster.

Notice that dollar value and severity are almost uncorrelated. Seagate (STX) has the largest discretionary total at $130.2M but ranks second on severity, because that sum came from only three sellers. Zscaler (ZS) shows just $2.9M and still scores 8.8, because four separate insiders chose to sell outside their plans. The number of independent decisions matters more than the size of the cheque.

The Ten-Minute Routine for Any Insider-Selling Headline

You do not need a screen to apply this. When a headline tells you an executive dumped stock, this is the sequence that resolves it, in rough order of how much time each step saves you.

1. Find the Form 4 and check the 10b5-1 box. If it is checked, you are done. Nearly three in five dollars of insider selling stop here, and the remaining nine steps are wasted effort on a pre-scheduled trade.
2. Read the transaction code. An M paired with an S on the same date is exercise-and-sell. An F is tax withholding and involves no decision whatsoever. Neither is an opinion about the business.
3. Divide by the post-transaction holdings. The last column of Table I gives you the denominator. Convert the headline dollar figure into a percentage before you form any view at all.
4. Pull the insider's prior filings. Is this the eleventh quarterly sale or the first sale in two years? A pattern of regular selling makes one more sale uninformative; a break in a long silence is the strongest single-person signal available.
5. Look for other insiders in the same thirty days. This is the test that does the heaviest lifting. One seller is a person; four unscheduled sellers is a pattern, because independent personal circumstances do not synchronise.
6. Check the footnotes for an indirect holding, a gift or a pledge. Any of the three can turn an apparent sale into something with no bearish content.
7. Then look for the boring explanation. A lock-up that just expired, a secondary offering, a closed acquisition, a divorce, a house. Rule these out before reaching for anything more interesting.

Steps one through three take about two minutes each and eliminate the large majority of insider-selling stories outright. The twelve companies in our table survived all of it — and for eight of them, the clustering test alone was the strongest reason they appeared:

Every one of those twelve primary reasons is the clustering test. Not one of the twelve reached the top of the list because of a single large sale, which is the shape of almost every insider-selling headline you will ever read. The thing worth screening for and the thing that gets reported are not the same thing, and that gap is the entire reason this kind of analysis is worth doing.

Five Mistakes That Make Insider Selling Look Scarier Than It Is

1. Reading the dollar amount without the stake

Covered above, and worth repeating because it is the most common error by a wide margin. $20M means nothing without knowing what fraction of the position it represents.

2. Counting option exercises as selling

Transaction code M followed by S on the same date is an exercise-and-sell: the insider converted an expiring option into cash. Frequently they had no real choice, because options expire worthless. Many aggregator sites report the sale leg and ignore the exercise leg, which makes compensation look like an exit.

3. Counting tax withholding as selling

Code F is shares withheld by the company to cover the tax bill on vesting. The insider never had a decision to make. These appear in “insider sold” totals constantly.

4. Treating 10% owners as insiders

A private equity fund or founder's family trust files Form 4 because of its ownership percentage, but it sells on a fund lifecycle, not on a view of next quarter. Our snapshot shows 55 such filers. Their selling is usually about their own fund mechanics — and in post-IPO companies specifically, it is often about an expiring lock-up. We cover that interaction in the lock-up expiry guide.

5. Expecting symmetry with buying

Buying and selling are not mirror images. Insiders buy for essentially one reason and sell for many, which is why “insiders are buying” feels like a stronger signal — even though, when we actually measured it across 929 cluster buys, it lost to the S&P 500 over twelve months. Insider data is better at raising questions than at answering them, in both directions.

Run These Filters Without Reading 432 Filings

Our insider dashboard applies every test in this article automatically: it parses Form 4 filings straight from SEC EDGAR across a 3,200-ticker universe, separates plan sales from discretionary ones, resolves each insider's actual role, computes what fraction of their stake moved, and flags the breaks in behaviour.

It also shows you the reasons behind every flag, so you can disagree with the screen rather than trust it.

The Bottom Line

Insiders sold $69.9 billion and bought $2.7 billion over 90 days. That 26-to-1 ratio is not a warning about the market — it is a description of how executives are paid. Reading it as bearish is the single most common mistake in this area.

Of that selling, 58% by value was pre-scheduled under 10b5-1 plans and contains no opinion whatsoever, by legal design. Of the discretionary remainder, 432 candidates produced just 12 companies where several insiders independently chose to sell outside their plans, in size relative to their holdings, after a long quiet period.

That shortlist — twelve names out of 3,200 — is the entire usable signal. Everything else is a headline about a mortgage payment.

All figures from a snapshot generated 7 October 2026. The underlying data refreshes on a schedule, so live numbers on the dashboard will differ.