Equity Strategy · 8 min read · 2026-05-04
How to filter Form 4 insider buying for real conviction.
Insiders sell for many reasons. They buy for one. The discipline is to filter the buys for genuine conviction.
Most insider buys are signal-free.
Approximately 80 percent of disclosed insider purchases on Form 4 are too small, too predictable, or too compromised by 10b5-1 plans to carry information. The remaining 20 percent — the cluster buys, the open-market large prints, the unusual-role buyers — produce the historical alpha that gives insider-buying its reputation. The screen separates them.
The nine indicators
The nine signals of a conviction insider buy.
Each signal is verifiable from EDGAR. The discipline is to filter the noise before drawing the conclusion.
Open-market purchase, not exercise of an option
Form 4 transaction code 'P'
Code 'P' is an open-market purchase. Code 'M' is an option exercise. Only 'P' transactions are signals; option exercises are mechanical.
Purchase size > 25% of insider's prior 12-month compensation
Threshold: trade $ / comp $ > 0.25
A meaningful buy hurts. When the trade size exceeds a quarter of the insider's annual compensation, the conviction is personal, not symbolic.
Cluster pattern: 3+ insiders within 30 days
Threshold: ≥ 3 cluster buys
A single CEO buy is anecdotal. Three or more insiders buying within a 30-day window is institutional. Cluster buys outperform single buys by 4–6× in historical samples.
Buyer is the CFO or controller, not the CEO
Role: financial officer
CFOs and controllers see the books in a way CEOs do not. CFO open-market purchases produce stronger forward returns than equivalent CEO purchases, on the historical record.
Not under a 10b5-1 plan
Form 4 footnote disclosure
10b5-1 plans are pre-set and mechanical. They are disclosed as such on the Form 4. A 10b5-1 buy is not a discretionary signal; an open-market discretionary buy is.
Stock down 20%+ from 52-week high at time of buy
Threshold: drawdown > 20%
Insider buys near highs are routine portfolio diversification reversed. Insider buys deep into drawdowns are conviction expressions.
Insider's prior buys outperformed the index
Track record: prior buys > S&P 500
Some insiders buy well; others do not. Historical track record on the same insider's prior open-market buys is the simplest filter on individual signal quality.
No simultaneous insider selling at the company
Net flow: net buying
When some insiders buy and others sell concurrently, the net signal is muted. Net-positive insider activity over a 60-day window is the cleaner setup.
Quiet-period compliance — buy outside the blackout window
Window: 4 weeks pre-earnings
The most informed buys happen outside of any blackout window — typically 4–6 weeks pre-earnings — because the insider had access to material information and chose not to trade until the window cleared. Such timing is a tell of process and conviction.
Why insider buys carry signal and insider sells often do not
Insider selling has many drivers. Tax planning. Diversification. Estate planning. Charitable giving. Forced sales under 10b5-1 plans. Margin calls on personal loans. Most retail commentators treat insider sells as bearish; the historical evidence is that, on average, they are noise. Insider buys are different. There is one reason an executive writes a personal check for company stock: belief that the stock is undervalued. The signal-to-noise ratio of insider buying is structurally higher than insider selling.
The nuance is that not all buys carry equal information. The discipline is to filter.
The cluster effect
A single insider buy is anecdotal. Three or more insiders buying within a 30-day window is a different category of evidence. The mechanism is informational: when multiple board members and executives independently reach the same conclusion about valuation in a short window, the joint signal is stronger than any single name's view. Cluster buys, in the historical sample going back to 1990, have outperformed the broad market by 8–14 percentage points over the subsequent 12 months. Single buys have outperformed by 1–3 percentage points.
The cluster screen is mechanical. Pull all Form 4 filings for the company over the trailing 30 days. Count distinct named insiders with code-'P' transactions. Three or more is the threshold. The cluster window resets monthly.
Why CFO buys outperform CEO buys
CEOs are paid in stock, talk about the company publicly, and have a reputational interest in being seen to buy the stock at every drawdown. CFOs do not. The CFO's compensation is more cash-weighted, the public exposure is lower, and the personal balance sheet is typically smaller. When a CFO writes a personal check for company stock at scale, the conviction is unusual.
Empirical studies confirm this. CFO open-market buys above $250,000 produce, on average, twelve-month forward returns 200–400 basis points higher than CEO buys of equivalent size, controlling for market cap and sector. The CFO sees the books. When the CFO is buying, the books are telling them something they cannot say publicly.
The 10b5-1 trap
10b5-1 plans were created in 2000 to give insiders a way to trade without violating insider-trading laws. The plan is set up in advance with a broker, specifying purchase or sale dates and prices, and the trades execute automatically. A 10b5-1 buy is not a discretionary signal; the insider committed to the trade weeks or months before, often with no current view of the stock.
The Form 4 indicates 10b5-1 status in a footnote. The screen excludes 10b5-1 trades. Only discretionary, open-market, code-'P' buys count toward the cluster and conviction signals.
The drawdown filter
Insider buys at 52-week highs are usually routine diversification reversed. Insider buys at 20%+ drawdowns from the 52-week high are conviction expressions. The drawdown filter alone, applied to the universe of cluster buys, sharpens the historical alpha by approximately 50%. The buy is a contrarian view on the company's valuation, made by people who know the company best.
Building the screen
EDGAR is free. Every Form 4 is filed within two business days of the trade. A simple weekly run pulling all code-'P' buys in the U.S. equity universe, filtering for 10b5-1 footnotes, scoring against the nine signals, and sorting by composite score produces a tractable shortlist of three to twelve names per quarter. That shortlist, in historical backtests, has outperformed the index by 7–11 percentage points over twelve-month forward windows.
The strategy is satellite, not core. Position size at 1–2% per name, cap the bucket at 8–12% of the portfolio, and rotate annually. The historical drawdown of the strategy in down markets is meaningful — insiders do not buy bottom ticks, and their buys often precede further drawdowns by months — so the strategy works best when paired with a regime overlay.
Get the nine, every Monday.
Free weekly digest. The only U.S. equity letter that publishes a name only when nine independent signals align.
Common questions
Questions.
Where do I find Form 4 filings?
EDGAR.SEC.gov is the primary source. Free aggregators including OpenInsider, Finviz, and InsiderArbitrage curate the data. Form 4 is filed within two business days of the trade.
Is insider selling ever a signal?
Mostly not, but extreme concentrated selling — for example, the CEO selling 50%+ of holdings outside any 10b5-1 plan — has historically had modest predictive value. The signal-to-noise ratio is much weaker than for buys.
Do board-member buys count?
Yes, but slightly less than executive buys. Independent directors are less informationally privileged than the CEO or CFO but still have meaningful access. Director cluster buys are signal-positive.
How long should I hold an insider-buy position?
Twelve months is the historical optimum holding period. Shorter horizons miss the diffusion of the information; longer horizons accumulate noise from the next cycle.
What about insider buying ETFs?
Several ETFs (KNOW, NFO, historically) implement insider-flow strategies. They tend to underperform the disciplined nine-signal screen because they lack the cluster, drawdown, and 10b5-1 filters that produce most of the alpha.
How does this interact with short interest?
Insider buying combined with high short interest is one of the strongest setups in the historical record. The shorts are forced to cover as the insider thesis plays out, producing a squeeze on top of the underlying re-rating.
One name. Sometimes weeks of silence. Always with conviction.
Free weekly digest. No credit card. Unsubscribe anytime.