Insiders

Each cell is one business day, counting the distinct holders who bought on the open market and the distinct holders who sold at their own discretion. Dividend reinvestment, employee stock purchase plans, 10b5-1 plan trades and same-day exercise-and-sell are excluded on both sides — measured, 71% of insider selling by value is mechanical, and including it would make this track payroll cycles.

Counted in holders, not dollars. One founder's $17m sale would outweigh fifty officers buying $100k each.

Insiders normally sell more than they buy, so a red-leaning calendar is the usual state, not a warning. What is worth noticing is a run of green against the rest.

Colour is direction; size is whichever measure you pick — value traded or holders trading, both counting buys plus sells that day. They are separate channels on purpose: encoding magnitude twice made each harder to read.

The two sizings disagree, and that is the point. By value, a single $4bn disposal makes its day the largest on the chart even though one person did it. By holders, that day is ordinary and a day on which forty officers each bought $50k becomes the large one. Neither is the true one; a day that is big on both is the one worth opening.

Each measure gets the scale its range needs. Daily value spans roughly 350,000-fold ($6k to $2.2bn), so it is drawn on a logarithmic scale — on a linear or square-root one the median day sits on the minimum size, indistinguishable from the quietest. Holders span only about 230-fold, where that same log curve crushes most days into the largest few sizes, so holders use a square root instead. Either way area is NOT proportional: size ranks days, it does not let you compare magnitudes by eye. Hover for the figures.

An almost-empty cell is a business day with no qualifying trades — a quiet day, not missing data.

1 earlier days are not shown at all: the Form 4 backfill works backwards and has not reached them.

Descriptive only. Not investment advice.

Open-market purchases only (SEC Form 4 code P) — cash spent at the market price. Share grants, option exercises and shares withheld for tax are excluded: they are compensation mechanics, not decisions to buy.

Purchases made under a pre-arranged Rule 10b5-1 plan are excluded too. Those were scheduled months in advance and say nothing about today's price.

So are plan settlements. Code P also covers dividend reinvestment, employee stock purchase plans and director stock-purchase plans — no money is chosen to be spent and no decision is made. These are identified from the filing's own footnotes, and excluding them matters: one quarter-end dividend reinvestment produced eleven “buyers” at a single company.

Buyers are counted as decisions. Affiliated entities that file jointly count once, shown as “co-filers”, because a shared filing is one investment decision and the form does not attribute a trade to any one of them.

Ranked by total value, then by number of buyers. Stake increase is shown separately and never folded into the ranking. Clusters below $50,000 combined are not shown — a few thousand dollars split between four directors is not evidence of anything.

744 open-market purchases across 207 companies in the window · records begin 2023-03-03.

Where a company carries one, our own four-question read is shown beside the buying. Quality and price use different words on purpose — a price is not good or bad, it embeds a requirement — and they are never combined into a score. Each question is answered independently, so a company can carry an earnings read and no price read. Measured over the 867 companies with open-market buying in the last 90 days, 557 carry at least one: earnings quality on 387, business quality on 263, price on 399. Where a question is missing it is because the filings did not support it — a reverse-DCF needs positive free cash flow, which many small caps and foreign filers on IFRS tags do not report.

Does insider buying predict returns? Not reliably, on our data. Measured over ~46,700 open-market purchases since January 2024: companies insiders bought beat comparable-sized companies by 4 to 15 points over the following year, but still returned less than the S&P 500 where they were small — micro-caps, 55% of all purchases, returned 10.6% against the index's 18.4%. More buyers was not better; in micro-caps it was worse. And the effect did not repeat in consecutive years. Treat this as a place to start looking, not as a reason to buy.

Insider buying is descriptive, not a recommendation. Nothing here is investment advice.

Open-market sales (SEC Form 4 code S) of common stock, grouped by company and newest first, where the company's discretionary selling totals more than $100k.

Sellers are counted as decisions, exactly as buyers are: affiliated entities that file jointly count once, shown as “co-filers”.

Rule 10b5-1 plan sales are excluded. They were scheduled months in advance, so the seller expressed no view on today's price.

Exercise-and-sell is excluded. An option exercise sold the same day is a compensation conversion, not a decision to reduce exposure. It is reported elsewhere as an ordinary sale.

“Filed Nd late” counts calendar days past the two-business-day deadline, allowing five days so that a Thursday or Friday trade crossing a weekend is never flagged. It is a governance observation about the company, not about the trade.

The two excluded groups overlap: a 10b5-1 plan often settles an option exercise, so the figures above do not sum to the total. Discretionary is counted directly rather than by subtraction.

Insiders sell for tax, diversification and life events. Selling carries far less information than buying, and nothing here is investment advice.

SEC rules require a Form 4 within two business days of the transaction. This counts calendar days and allows 5, so a Thursday or Friday trade filed after a weekend — or a public holiday — is never flagged. Everything counted here is late beyond argument.

Ranked by how OFTEN a company is late, then by how late it typically is. Those are different things: a company 19-for-19 late by one day and one 17-for-17 late by up to 176 days are both chronic, and only the second is serious. The two are shown separately and never multiplied into a score.

Companies with fewer than 8 filings in the window are excluded entirely. One late filing out of one is 100% and means nothing; ranking it first would make the least informative row the loudest.

This describes a company's filing discipline. It is not a statement about the trades, the insiders, or the share price, and it is not investment advice.