SENTRA SIGNALS

How Sentra Signals Works

An insider-activity radar for the full Russell 3000. Surfaces the trades that carry real signal and filters out the noise.

The Problem

The most powerful signal in markets has always been hiding in plain sight: insiders. CEOs, CFOs, and major shareholders see earnings before they are announced, contracts before they are signed, product pipelines before the market has any idea. They have an edge that no analyst and no algorithm can replicate.

And by law, every time an insider acts on that conviction, they have to report it to the SEC within two business days. The data is public. The signal is sitting there. So why does it get ignored?

Noise. Insiders file thousands of trades every month: routine stock grants, scheduled sales on pre-set 10b5-1 plans, small compensation exercises. None of that means anything. Finding the trades that matter means telling the difference between an insider following a calendar and one acting on conviction. That is not easy to do at scale.

Sentra reads every Form 4 filing across the full Russell 3000 (~2,899 tickers), classifies each insider using a methodology grounded in academic research, and surfaces the trades where the signal is real.

CMP Classification

Three Classes of Insiders

Every insider in the Russell 3000 is classified using the Cohen, Malloy and Pomorski (2012) framework. The core insight: insiders who buy outside their historical calendar patterns generate meaningful returns. Insiders on predictable schedules generate essentially none.

Opportunistic

The insider bought outside their historical calendar pattern. This is the strongest class of signal. Opportunistic buyers tend to act when they have a genuine reason to buy, not because a scheduled plan told them to.

Unclassifiable

There is insufficient trading history to determine a pattern. The insider may be new, filing rarely, or joining the company recently. Carries signal but with less certainty than Opportunistic.

Routine

The insider trades on a predictable, calendar-driven schedule. These are typically plan-based or compensation-related transactions. The research shows they carry little forward-looking signal.

The Conviction Score

A Percentile Rank Within Each Classification

The color tells you the classification. The number (0–100) ranks that trade relative to other trades in the same classification. A score of 80 means this trade scored better than 80% of similar trades on the same signals. It is a description of trade characteristics, not a profit prediction.

Only purchases are scored. Insider selling carries almost no forward-looking signal: insiders sell for many reasons (taxes, diversification, personal liquidity). Sentra scores open market purchases only.

Scoring Factors (walk-forward validated)

F1 — Stake increase
30 ptsAll

How much the purchase increases existing holdings as a percentage. A large increase relative to existing position signals real conviction. New positions (no prior holdings) are scored by dollar size.

F2 — Price momentum 90d
40 ptsUnclassifiable and Routine only

The stock's 90-day price return before the trade. A pullback before a buy scores higher than buying into an uptrend. Not applied to Opportunistic trades (walk-forward IC is near zero for that group).

F3 — Role
12 ptsAll

10% owners and co-founders score highest (12 pts), followed by CEO, CFO, President, and Directors (9 pts), then other titles (7 pts).

F4 — Size tilt
8 ptsAll

Smaller-cap companies score higher. Pearson IC is modest (−0.11); Spearman is near zero. Labeled as a lottery tilt: adds high-variance upside potential, not a consistent edge.

Composite IC 0.05 to 0.09 across validation sample. Factors excluded from the model (noise or lookahead bias): trade size in dollars, insider cluster breadth, sector, 52-week range, and prior track record.

Six Signal Buckets

OPPORTUNISTIC50+Very high signalOpportunistic buyer with above-median conviction signals
UNCLASSIFIABLE50+High signalUnclassifiable insider with above-median signals
OPPORTUNISTICUnder 50Moderate signalOpportunistic buyer with below-median signals
UNCLASSIFIABLEUnder 50Low signalUnclassifiable insider with below-median signals
ROUTINE50+Low signalRoutine buyer, above median for its class
ROUTINEUnder 50No signalFormulaic or small routine buy

Academic Foundation

The Research Base

Cohen, Malloy and Pomorski (2012)
Journal of Finance
Decoding Inside Information
Insiders who trade outside their historical calendar patterns ("opportunistic" traders) generate significant market-adjusted returns. Insiders on predictable schedules ("routine" traders) generate essentially nothing. The classification is a better predictor of future returns than trade size, role, or any other single variable.
HOW SENTRA USES THIS
This is the core of the Sentra classification system. Every insider's trades are evaluated against their own historical pattern. The opportunistic/routine split drives both the color coding and the signal interpretation.
Jeng, Metrick and Zeckhauser (2003)
Review of Economics and Statistics
Estimating the Returns to Insider Trading
Insider purchases earn roughly 6% abnormal return over 6 months on a value-weighted basis. Insider sales show no significant predictive power.
HOW SENTRA USES THIS
This anchors the decision to score purchases only. Insider selling is noisy: insiders sell for taxes, diversification, and personal liquidity regardless of their view. Buying with personal capital is a cleaner signal.

Methodology

From Filing to Signal

01
Ingest SEC EDGAR Form 4 filings
Daily ingestion of Form 4 filings across the full Russell 3000 (~2,899 tickers). Every transaction is parsed, normalized, and stored with full insider metadata.
02
Classify each insider using CMP methodology
Each insider's trade history is analyzed for calendar patterns. Trades outside a predictable monthly or quarterly pattern are classified Opportunistic. Predictable, scheduled trades are Routine. Insiders without enough history to classify are Unclassifiable.
03
Score open market purchases 0–100
For each purchase, four validated factors are computed: stake increase, price momentum (Unclassifiable and Routine only), role, and size tilt. The raw composite is mapped to a 0–100 percentile rank within that insider's classification, calibrated from historical buys.
04
Display with classification color and signal bucket
Each trade is shown with its CMP color (green/yellow/red) and conviction score. Classification and score together determine signal strength using six buckets, from Very high signal (Opportunistic ≥50) to No signal (Routine under 50).

Sentra Signals provides informational data sourced from public SEC filings only. It is not a registered investment advisor and nothing on this platform constitutes investment advice. Users are solely responsible for their own investment decisions.