Scoring methodology

How FilingIQ turns raw SEC filings into ranked signals. Every model is calibrated against our full historical corpus, version-controlled, and backtested against out-of-sample windows before deployment.

Authored by Niklas Feldmann, Co-founder and Head of Research.

Composite Score (Form 4)

The Composite Score is FilingIQ's headline signal for SEC Form 4 insider transactions. It is calibrated against our full historical corpus of 3,410,722 transactions across 1,892,154 filings, and produces a 0-100 score for every open-market purchase or sale.

The model is a weighted combination of factor families spanning insider context, trade structure, and market environment. The factor families:

  • Insider context: seniority of the reporting person (CEO, CFO, 10% owners, directors), prior conviction history, and indirect / related-party reporting structure.
  • Trade structure: transaction code, plan-driven vs discretionary execution (Rule 10b5-1), and position sizing relative to the insider's existing holdings.
  • Market context: issuer-level valuation, trade timing within the prevailing range, and cross-signal alignment with other FilingIQ models.

Factor weights and threshold cutoffs are part of the calibrated model and not published in full. Each model release is backtested against held-out historical windows before deployment.

Institutional Flow Score (IFS)

IFS scores 13F-derived institutional accumulation across the active 13F universe, on a quarterly cadence. The current production model combines several institutional-flow components into a single 0-100 score per ticker per quarter, with each component converted to a cross-sectional percentile within the quarter.

IFS has iterated across multiple research generations; the current model is the result of comparing each candidate against held-out historical 13F sequences. Component weights and the exact composition are part of the deployed model and not enumerated here.

Congressional Political Score (CPS)

CPS scores Congressional stock trades disclosed under the STOCK Act. Inputs include committee-relevance context, sector concentration in the member's broader disclosed portfolio, and timing of the disclosed trade relative to surrounding congressional activity. Specific factor composition and weighting are part of the internal model.

Data refresh cadence

Source Refresh Latency from event
SEC Form 4 (insider) Continuous, real-time EDGAR poll Minutes after SEC publication
SEC Form 13F (institutional) Quarterly Up to 45 days after quarter end
House PTR (Congressional) Daily Up to 45 days disclosure window
Senate PTR (Congressional) Daily Up to 45 days disclosure window
SEO entity export to JSON snapshots Nightly plus on every deploy Up to 24 hours

Known caveats

  • Warrant and derivative leakage: a small fraction of derivative-instrument Form 4 transactions can score as conviction buys when the underlying derivative settlement metadata is incomplete. Affected tickers are reviewed manually; corrections are reflected on the next nightly export.
  • Dual-listed currency footnotes: a small set of dual-listed issuers (Australian, UK, German cross-listings) have historically reported share prices in their home currency without an explicit USD conversion. Where detected, totals are corrected; where not detected, the raw value is preserved with a footnote.
  • Ticker resolution gaps in Congressional disclosures: STOCK Act filings are free-text and a meaningful share of historical rows have security descriptions that do not unambiguously match a public-equity ticker. Unresolved rows are excluded from CPS scoring but remain visible in the raw feed.
  • Composite score availability: Form 4 transactions older than 3 years may not have a Composite Score if they predate the current scoring model's calibration window.

Primary sources only

Every signal on FilingIQ traces back to a public SEC or Congressional filing. The full data source inventory is at /sources. Not investment advice; see disclaimer.