Editorial
How the programmatic desk builds an article
InsiderBuying.com reports and analyzes public filings. This article does not recommend any security.
The screens
Each screen is a versioned definition. Changing a threshold creates a new version, and an article keeps the version it was built from, so an article published under version 1 still describes how it was made after the threshold moves.
At least 2 distinct insiders with open-market purchases (code P) within 30 days. Aggregate value at least $250,000.
Used in: Cluster list articles
Closing price below $5 on the publish date. At least 1 open-market purchase by an officer or director in the last 180 days.
Used in: Under-$5 articles
Price down at least 20% from the 52-week high. Open-market purchase by an officer or director after the drawdown began.
Used in: Consumer, energy, financials
Holdings in the sector disclosed by a named official in a filing dated within 180 days.
Used in: Official holdings slots
At least 1 open-market sale (code S) by an officer or director filed under a Rule 10b5-1 plan in the last 90 days. Aggregate value at least $1,000,000.
Used in: Technology and AI slots
Defined by this desk rather than by the published specification, because the weekly rotation calls for a format the specification does not give a screen for.
Sector ETF return versus S&P 500 over the period, with the net insider flow for the sector's screened names.
Used in: Opening paragraph in every article
Weekly rotation
| Sector | Benchmark | Articles per week |
|---|---|---|
| Biotech and pharma | XBI versus the S&P 500 | 2 |
| Gold and metals mining | GDX and GDXJ versus the S&P 500 | 2 |
| Small cap | IWM versus the S&P 500 | 2 |
| Energy | XLE versus the S&P 500 | 2 |
| Financials | XLF versus the S&P 500 | 2 |
| Technology and AI | XLK versus the S&P 500 | 2 |
| Consumer and retail | XLY versus the S&P 500 | 2 |
| Industrials and defense | XLI versus the S&P 500 | 2 |
No more than 10 articles publish on any calendar day. A slot with no names passing its screen is held and the reason recorded, not filled with a weaker name.
What every article contains
- The dek is one sentence stating the screen and the period.
- Key points. Three bullets. Each bullet is one sentence with one number and one date.
- Two sentences. The first states how the sector and the benchmark moved over the period, with the source. The second states the screen result and why it matters.
- The backdrop. One paragraph placing the story in the economic backdrop, with a dated source (rates, inflation, payrolls, or sector policy).
- One sentence in a pull-quote block. It must be conditional and testable, and it must state what would invalidate it.
- One row per name (Section 5 fields).
- One short paragraph per name, in the same order as the table. Each paragraph covers the filings, the filer role, the price context and the benchmark comparison.
- What would change the view. Two items, each tied to a dated event or a data series.
- Methodology and sources. The screen ID, the data pull date, the filing accessions, and the benchmark definition.
Where the numbers come from
- Insider transactions
- SEC EDGAR Form 4, including transaction codes and the 10b5-1 flag
- Congressional trades
- House Clerk and Senate electronic financial disclosure periodic transaction reports
- Executive disclosures
- OGE Form 278e and periodic transaction reports
- Executive disclosures note
- The Office of Government Ethics publishes these as documents and indexes none of them, so each row on this site is typed from the filing it names rather than ingested from a feed.
- Institutional
- SEC Form 13F
- Prices
- Exchange closes for each name, plus XBI, GDX, GDXJ, IWM, XLE, XLF, XLK, XLY, XLI and the S&P 500
- Macro
- Federal Reserve rate decisions and Bureau of Labor Statistics releases, taken from a licensed economic calendar and used only once a figure has actually printed
United States listings only. Canadian insider filings are published on SEDI, which offers no machine-readable feed, so Canadian names are outside the screens.
How the figures are derived
- Period return
- Close on the end date divided by the close on the start date, minus one.
- Benchmark spread
- Name return minus benchmark return over the same dates.
- Net insider flow
- Purchase value minus sale value, using open-market codes only. Code P is an open-market purchase and code S an open-market sale. Codes A, M, F and G are excluded from the flow and listed separately.
Checks before publication
Every article clears an automated gate before it publishes: no em dashes, no words from the rejected list, every number matched against the stored data layer, a price date no more than one trading day old, and a re-checked close on any "under $X" headline. Articles are then read by an editor, and an article that has been read names its reviewer in its structured data.
Headlines and article text never use these words: bombshell, shocking, explosive, massive, huge, brutal, crash, skyrocket, must-buy, guaranteed, secret, game-changer, stunning, insiders know, sure thing.