Not all insider trades are equal
Thousands of SEC filings land every month. Most are routine. A few contain real information. Our scoring model isolates the ones that matter — separating signal from noise across 8 contextual factors.
Two-Stage Scoring Pipeline
Every filing passes through two stages before receiving a significance score
Stage 2 — Multiplicative Factors (runs first)
Multiplies the base signal up or down based on contextual factors — who traded, when, how much, and what else was happening at the company.
Stage 1 — Rules-Based Clamps (runs second)
Applies rules that can cap or floor the score based on known patterns that override context.
Suppressor Precedence
When a suppressor and a promoter both apply to the same filing, the suppressor wins. This prevents false positives where a structurally significant pattern is undermined by a noise-introducing factor.
The 8 Scoring Factors
Each filing is evaluated against these contextual factors
Insider Role
The seniority and information access of the filing insider.
CEOs and CFOs have broader company visibility than non-executive directors — the role reflects information asymmetry.
Cluster Detection
Whether multiple insiders are trading in the same direction within a short window.
Coordinated buying or selling across insiders is a stronger signal than any single trade.
Contrarian Signal
Whether the insider is trading against prevailing market sentiment.
Insiders buying when sentiment is negative, or selling when sentiment is euphoric, often reflects information not yet priced in.
Conviction
The size of the trade relative to the insider's existing holdings and compensation.
A $500,000 purchase by an executive earning $200,000/year carries more weight than the same purchase by a billionaire founder.
10b5-1 Plan Status
Whether the trade was executed under a pre-arranged trading plan.
Pre-planned trades were scheduled months earlier under different circumstances — they carry less timely information.
First-Time Buyer
Whether this is the insider's first open-market purchase in the company.
Initiating a new position often reflects stronger conviction than adding to an existing one.
8-K Timing
How close the trade falls to a recent material event disclosure.
Trades near 8-K filings (earnings, restructuring, M&A) carry elevated information relevance.
Cross-Filing Signal
Whether the same company appears in multiple filing types within a short period.
Form 4, 13D/13G, and 8-K activity converging on the same entity amplifies the signal through independent corroboration.
What We Do Not Score
The scoring model is filing-driven, not opinion-driven. We deliberately exclude:
- Company market capitalisation or size
- Post-filing stock price performance (returns after the trade)
- Analyst ratings or price targets
- Media coverage or social media sentiment
- Political or regulatory environment
The scoring model evaluates only what was filed with the SEC. A small-cap director buying $50,000 of stock scores on the same criteria as a large-cap CEO buying $5 million.
Score Output & Priority Tiers
Every filing receives a significance score from 1 to 10, mapped to a priority tier
Routine filing activity — no contextual signals present
Some contextual signals — worth noting, elevated attention warranted
Rare convergence of signals — filing activity with uncommon contextual weight
A Critical score does not mean buy or sell. It means the filing contains contextual patterns that, in aggregate, are rarely present in routine transactions.
Data Sources
Source filings: SEC EDGAR, ingested via the Thesma API.
Filing types covered: Form 4 (insider trades), 13D/13G (beneficial ownership), 8-K (material events), 13F (institutional holdings).
Data cleaning: The scoring model applies reduced weight to routine and non-discretionary transactions including pre-arranged 10b5-1 plan trades.
Powered by Thesma API
Akivus provides informational analysis of public SEC filings. We do not provide investment advice, recommendations, or opinions on any security. A high significance score reflects filing context — it does not indicate that a stock will go up or down.
Last updated: March 2026