Every congress stock trade filed under the STOCK Act passes through a 30-step daily intelligence pipeline that transforms raw government disclosures into ranked anomaly scores, corroboration signals from 8 independent federal data sources, and plain-English ARIA intelligence briefs — surfacing the trades that matter from the 32,000+ in the database.
We pull the official PTR XML index from disclosures-clerk.house.gov daily. Each filing gives us member identity, chamber, disclosure date, and a link to the underlying PDF.
Senate disclosures come from the Senate eFD system. Data is normalized to the same schema so House and Senate signals are directly comparable. Amended filings automatically supersede originals — no duplicate trades.
Each PTR PDF is parsed with a dual-pass extractor: structured table extraction first, then regex fallback on raw text. Produces ticker, transaction type, dollar range, and transaction date.
Eight independent federal data sources are cross-referenced against each trade. Each match adds a corroboration badge and increases the trade's anomaly signal weight.
Each ticker is classified using GICS sector data from yFinance. 85.9% of active tickers have sector assignments. Sector data enables committee-relevance scoring and corroboration matching.
Committee meetings, hearings, and bill markups are ingested daily from Congress.gov API v3. Each trade is matched to the nearest committee event for that member's committees. The result: a signed days_to_event value — negative means the trade occurred before an event (positioning), positive means after (reaction).
Each of 118 active trading members has a 24-month behavioral profile:
The baseline is the denominator for anomaly scoring. Without it, every Khanna trade looks like a signal.
Each trade receives an anomaly score (0–100) across five weighted dimensions, multiplied by a committee access factor (1.0–1.6×) based on whether the member's committee has direct oversight of the trade's sector. Committee chairs receive the highest multiplier.
For every trade with anomaly score ≥ 40, ARIA generates a 3–5 sentence intelligence brief using Claude API. Each brief synthesizes what makes this trade anomalous for this specific member, which corroboration sources confirm the signal, the member's baseline context, and the legislative proximity and committee access context.
220 intelligence briefs are currently live in the INTEL tab (PRO subscribers).
The remaining steps handle: conviction scoring (historical alpha weighting), bipartisan cluster detection, backtest preset refresh, AI summaries, member stats refresh, compliance watch (STOCK Act violation tracking), and daily quality checks across 22 dimensions before the pipeline completes at 07:30 UTC daily.
Five dimensions, each scored 0–100, combined into a weighted composite and multiplied by a committee access factor (1.0–1.6×). Minimum data requirements must be met for all dimensions — insufficient data defaults the dimension to 0.
All primary data is sourced directly from official U.S. government systems. We do not use commercial data vendors for the core trade feed.
Bipartisan clusters are the highest-confidence signals on the platform. The algorithm applies strict rules to minimize false positives.
The STOCK Act requires members of Congress to disclose trades within 45 days of the transaction. Our pipeline tracks every filing date against every transaction date and flags members whose late-filing pattern is statistically anomalous — not isolated delays, but a recurring pattern across multiple trades.
The /late-disclosures page publishes the most severely flagged members and weekly filing statistics — no account required. PRO subscribers get the full Compliance Watch table with all severity tiers, expanded statistics, and sortable columns.
220 ARIA intelligence briefs. 32,289 trades scored. 8 corroboration sources. The most rigorous congressional trade analysis available.