Methodology

Signals are computed daily from stored public records. Magnitude is a 0–100 score. Keys are idempotent so re-runs do not duplicate events.

• Insider purchase: non-derivative Form 4 code P, acquired, value at least $50,000. Magnitude is min(100, round(40 + 15 × log10(value / 50,000))).
• Insider buying cluster: at least three distinct insiders with qualifying purchases in any 14-day window. Magnitude 80.
• Large insider sale: non-derivative code S, disposed, value at least $1,000,000. Magnitude is min(90, round(30 + 12 × log10(value / 1,000,000))).
• Federal contract award: a single USAspending transaction of at least $10,000,000. Magnitude is min(100, round(40 + 15 × log10(amount / 10,000,000))).
• Surge in federal contracts: last 30 days of obligations at least 3× the average 30-day sum over the prior 365 days, and at least $5,000,000. Magnitude 70.
• New lobbying issue: an issue code in the latest quarterly LDA filing set that was absent from the previous four quarters. Magnitude 50.
• Lobbying spend jump: latest quarter spend at least 2× the average of the previous four quarters and at least $100,000. Magnitude 60. Quarterly spend is outside-registrant income plus in-house expenses; amendments replace earlier filings for the same registrant and period.
• Public attention spike: 7-day mean Wikipedia pageviews at least 3× the 90-day median and at least 2,000. Magnitude is min(90, round(40 + 10 × log2(ratio / 3))).
• Material 8-K: items 1.01 (45), 1.03 (95), 2.01 (60), or 5.02 (50). One signal per filing using the highest item magnitude.

Sources: SEC EDGAR (company list, submissions, Form 4 XML, insider data sets), USAspending.gov, U.S. Senate LDA, Wikidata, Wikimedia pageviews. All are public-domain or CC0 for the fields we store.