Home›Guide›Predictive Insider Signals

Predictive Insider Trading Signals: How to Identify the Moves That Matter

Not all insider transactions are created equal. The vast majority is noise — automatic plans, equity compensation, tax-related sales. This guide teaches you how to isolate the signals that have real predictive power over future stock price movements.

The Problem: Too Much Noise, Too Few Signals

Every day the SEC records thousands of insider transactions via Form 4 filings. But 80-90% of these transactions carry zero informational value:

After filtering all this noise, you're left with 10-20% of meaningful transactions — and these are the ones with predictive power.

Transactions to Include in the Analysis

The 6 Indicators Behind the WhaleSentiment Score

Our algorithm combines 6 indicators to generate a score from -100 (maximum warning) to +100 (maximum bullish signal):

Indicator Weight What It Measures
1. Cluster Score 25% How many different insiders transacted in the same period (30 days). More insiders = stronger signal.
2. Value Ratio 20% Transaction value / company market cap. A $500K purchase at a $2B company carries more weight than $5M at Apple.
3. Role Weight 20% CEOs and CFOs have more operational insight → their purchases carry more weight. Outside directors less so.
4. Conviction Score 15% Purchase value / insider's annual compensation. If a CEO earning $1M buys $500K in shares = extremely high conviction.
5. Price Context 10% Buying after a >20% decline = strong contrarian signal. Buying at all-time highs = weaker signal.
6. Historical Accuracy 10% The specific insider's track record: have their past purchases generated positive returns?

STRONG Signals (Score +40 to +70)

🟢 High-Value Single Purchase +40/+55

Pattern: A C-level executive (CEO, CFO, COO) buys shares on the open market for >$200K in a single transaction.

Why it works: There's only one reason a CEO spends hundreds of thousands of their own dollars: they believe the stock is undervalued.

Additional filters:

  • Transaction code = P (Purchase on open market)
  • Role = CEO, CFO, COO, President
  • Value > 10% of estimated annual compensation
  • Not preceded by sales from the same insider in the prior 60 days

🟢 Contrarian Post-Decline Buy +45/+60

Pattern: An insider buys after the stock has dropped >20% over 60 days.

Why it works: The insider is signaling that the market is overreacting. They have visibility into fundamentals the market isn't pricing in.

Historical average return: +8.2% at 6 months (Lakonishok & Lee, 2001)

VERY STRONG Signals (Score +70 to +100)

🟢🟢 Cluster Buy — Multiple Insiders Buying Together +70/+90

Pattern: 3+ different insiders buy (code P) within a 30-day window.

Why it's the strongest signal: If only the CEO buys, it could be a personal matter. But if the CEO + CFO + 2 directors all buy in the same month, they almost certainly share a positive view based on inside information (order pipeline, deals closing, costs declining, strong quarterly results ahead).

Historical average return: +12% at 12 months vs market (Insider Sentiment Research, 2019)

Scoring formula:

cluster_score = (n_insiders - 1) × 20 + sum(value_i / market_cap × 10000)

Example: 4 insiders buy, total $800K on a $3B company
cluster_score = (4-1) × 20 + (800000/3000000000 × 10000) = 60 + 2.67 = 62.67
→ Normalized: +82/100

🟢🟢 CEO First-Time Buy +75/+95

Pattern: A CEO who has NEVER bought shares in the past 3 years makes a voluntary purchase.

Why it's very strong: A behavioral change is far more significant than an established pattern. If a CEO who normally only sells (compensation) decides for the first time to buy with their own money, they're sending a powerful signal.

Bonus: Even stronger when combined with a recent stock decline.

🟢🟢 Outsized Purchase (Conviction Buy) +70/+85

Pattern: Purchase value > 25% of the insider's total annual compensation.

Why it's very strong: They're putting an enormous portion of their personal wealth into this bet. They wouldn't do it without extremely high conviction.

conviction_ratio = purchase_value / annual_compensation

If conviction_ratio > 0.50 → Score +85
If conviction_ratio > 0.25 → Score +70
If conviction_ratio > 0.10 → Score +55

WARNING Signals (Score -40 to -100)

🔴 Anomalous Cluster Sell -60/-85

Pattern: 3+ insiders sell (code S, NO 10b5-1 plan) within 30 days, and at least one is C-level.

Why it's a warning: Sales are normally noisy, but when MANY insiders sell together AND they're not programmed sales, they're likely anticipating bad news.

Critical filter: Exclude sales flagged as is_10b5_1 = true. Only discretionary sales count.

🔴 CEO Dumps After Lock-up or at All-Time Highs -50/-75

Pattern: CEO/CFO sells >25% of their holdings while the stock is within 10% of its all-time high.

Why it's a warning: The insider is taking profits at the top. They may know that growth is about to slow down.

🔴🔴 Buying Stops + Selling Begins -70/-90

Pattern: An insider who had been buying regularly suddenly stops buying AND starts selling.

Why it's the strongest warning signal: A change in direction is more significant than any single act. If a CFO who bought for 2 years suddenly reverses course, something has changed in their view of the company.

Transactions to IGNORE (Noise)

⚪ Sell-to-Cover (Code F) 0

The insider sells the minimum required to pay taxes on exercised stock options. Purely mechanical, zero information.

⚪ Exercise + Immediate Sale (M + S same day) 0

The insider exercises options and sells immediately. Typical of expiring compensation plans. No forward-looking view on the stock.

⚪ Scheduled 10b5-1 Sales 0

Automatic sales scheduled 3-6 months earlier. The current price is irrelevant — the insider decided to sell when the price was different.

⚪ Pharma Pre-Trial (The Moderna Case) 0

At pharma/biotech companies with pending clinical trials, insiders are in a blackout period. Their transactions (if any) are almost always pre-programmed automatic plans and don't reflect their view on the trial outcome.

Full Algorithm: WhaleSentiment Score

INPUT: Transactions from the last 90 days for a ticker, filtered (only P and non-10b5-1 S) STEP 1 — Classification: buys = transactions with code P AND acquired_disposed = A sells = transactions with code S AND is_10b5_1 = false STEP 2 — Cluster Score (25%): n_unique_buyers = count(distinct insiders with buy) n_unique_sellers = count(distinct insiders with sell) cluster_buy = min((n_unique_buyers - 1) × 25, 100) cluster_sell = min((n_unique_sellers - 1) × 25, 100) STEP 3 — Value Ratio (20%): buy_value_ratio = sum(buy values) / market_cap × 10000 sell_value_ratio = sum(sell values) / market_cap × 10000 (capped at 100 to avoid micro-cap distortion) STEP 4 — Role Weight (20%): For each transaction: CEO/President = 1.0, CFO = 0.9, COO/CTO = 0.8, VP = 0.6, Director = 0.4 role_score = weighted_avg(role_weights) × 100 STEP 5 — Conviction (15%): max_single_buy_value / estimated_annual_comp × 100 (capped at 100) STEP 6 — Price Context (10%): price_change_90d = (current_price - price_90d_ago) / price_90d_ago If buy AND price_change < -20%: context_score = +80 If buy AND price_change < -10%: context_score = +50 If buy AND price_change > +30%: context_score = +10 (less significant) If sell AND price_change > +30%: context_score = -60 STEP 7 — Historical Accuracy (10%): For each insider: % of past purchases that generated positive returns at 6 months accuracy_score = avg(hit_rates) × 100 FINAL SCORE: If net buy activity: score = cluster_buy×0.25 + value_ratio×0.20 + role_score×0.20 + conviction×0.15 + context×0.10 + accuracy×0.10 If net sell activity: score = -(cluster_sell×0.25 + value_ratio×0.20 + role_score×0.20 + conviction×0.15 + context×0.10 + accuracy×0.10) INTERPRETATION: +80 to +100: 🟢🟢 VERY STRONG SIGNAL — High probability of appreciation +60 to +79: 🟢 STRONG SIGNAL — Elevated probability of appreciation +30 to +59: 🟡 MODERATE SIGNAL — Worth monitoring with other indicators -29 to +29: ⚪ NEUTRAL — No significant signal -30 to -59: 🟡 CAUTION — Possible downward pressure -60 to -79: 🔴 WARNING — Elevated probability of decline -80 to -100: 🔴🔴 STRONG WARNING — High probability of decline

Practical Example: The Moderna (MRNA) Case — August 2026

📋 Analysis Using the WhaleSentiment Score

Transactions (90 days): 0 buys (P), 7 sells (S)

But watch the filters:

  • Bancel's (CEO) sales on August 5-6 are likely 10b5-1 → EXCLUDED
  • Hoge (President) sold on July 15 and June 15 with a regular monthly pattern → likely 10b5-1 → EXCLUDED
  • Klinger sells $173K on June 4 — small amount, possibly discretionary

After filtering: Almost no meaningful transactions remain.

Final score: ⚪ 0 (NEUTRAL) — No predictive signal in either direction. Consistent with a pharma company in a pre-trial blackout.

The positive news on August 19 (anti-melanoma vaccine) was NOT predictable from insider data. This is a case where the WhaleSentiment Score correctly generates no signal, because insiders are blocked from trading on MNPI (Material Non-Public Information).

When the WhaleSentiment Score Works Best

Sector Effectiveness Reason
Technology (mid-cap) ⭐⭐⭐⭐⭐ High information asymmetry, order pipeline, gradual growth
Industrials ⭐⭐⭐⭐ Multi-year contracts, internally visible backlog
Energy ⭐⭐⭐⭐ Insiders see production and costs before the market
Consumer Discretionary ⭐⭐⭐ Quarterly sales trends, geographic expansion
Financials (banks) ⭐⭐⭐ Credit quality, expected NIM
Mega-cap ($500B+) ⭐⭐ Market too efficient, insiders have less "edge"
Pharma/Biotech (pre-trial) ⭐ Blackout periods, binary events, 10b5-1 plans dominant

How to Use This Guide with WhaleSentiment

  1. Monitor cluster buys: On the homepage, stocks in the "Top Insider Buys" section with multiple active insiders are the strongest signals.
  2. Mentally filter the noise: On each stock page, ignore transactions with code F, M (unless followed by holding), and sales flagged as 10b5-1.
  3. Focus on mid-caps: Signals are more predictive for companies with $1-50B market cap.
  4. Use as a filter, not a standalone strategy: Insider trading is a powerful indicator but should be combined with fundamental and technical analysis.
  5. Timing: Abnormal returns after insider purchases typically materialize over 3-12 months. This is not a day-trading signal.

Monitor Insider Signals in Real Time

WhaleSentiment tracks insider transactions for 51 US-listed companies, updated daily from SEC filings.

→ Go to Dashboard

Academic References

Disclaimer: This guide is provided for educational and informational purposes only. It does not constitute financial advice, investment recommendation, or solicitation to buy or sell financial instruments. Past performance is not indicative of future results. Consult a qualified financial advisor before making investment decisions.