From 7,000+ stocks to the ones worth your attention.
Follow one stock through our entire pipeline — every filter, every factor, every decision — to see exactly how composite scores are built.
The Pipeline
Seven stages. One stock at a time.
Let's follow Apple (AAPL) through the pipeline. At every stage, we'll show you exactly what the system checks and what it finds.
Universe
7,000+ stocks
Elimination Filters
6 pass/fail checks
Factor Scoring
Quality · Value · Momentum
Multi-Track Scoring
Compounder, Mispricing & Efficient Growth
ML Refinement
Cluster models, VAE
Smart Money Overlay
13F institutional signals
Position Sizing
Composite tier × opportunity type
7,000+ stocks
6 binary checks
Quality · Value · Momentum
Compounder, Mispricing & Growth
Cluster models, VAE
13F signals
Tier × type matrix
The pipeline runs automatically after each market close. Scores typically refresh within 2 hours of the closing bell.
Stage 1 · Universe
Every day, the pipeline starts with 7,000+ stocks.
Every day, the pipeline evaluates 7,000+ US-listed equities. AAPL is one of them. To enter scoring, a stock must first survive six elimination filters. No cherry-picking, no pre-selection — the full universe of NYSE, NASDAQ, and NYSE American listings enters the pipeline. Financials and Real Estate are excluded because their capital structures make standard profitability metrics unreliable.
What’s included
- ~7,000+ US-domiciled equities
- 9 sectors: Technology, Healthcare, Industrials, Energy, Consumer Cyclical, Consumer Defensive, Basic Materials, Utilities, Communication Services
- All market caps above liquidity minimums
What’s excluded
- Financials — leverage-as-product breaks ROIC metrics
- Real Estate — REITs use different valuation frameworks
- OTC / Pink Sheet listings
- Foreign ADRs
Data freshness
- Full scoring cycle runs daily after market close (4:30 PM ET)
- Scores refresh within ~2 hours of the closing bell
- Each score carries a freshness label: Fresh, Stale, or Expired
Stage 2 · Elimination Filters
Six binary checks. One failure means elimination.
AAPL faces six binary pass/fail checks. One failure means immediate elimination — no exceptions, no overrides. All six run regardless of earlier failures so you see the full diagnostic, not just the first thing that went wrong.
6/6AAPL passes all six filters and advances to scoring.
Roughly 40% of the universe fails at least one filter and is eliminated before scoring begins.
Filter thresholds are sector-adjusted — a utility company and a tech company are held to different standards where appropriate.
Stage 3 · Factor Scoring
17 factors. Three pillars. Sector-neutral ranking.
AAPL passed all filters. Now it enters multi-factor scoring across three pillars. Each factor is ranked within AAPL's GICS sector — a percentile of 85 means AAPL scores better than 85% of its tech-sector peers on that factor.
Factor scores are converted to percentile ranks within each company's GICS sector, measuring relative strength against sector peers. These factor scores feed into a multi-gate scoring system where each track evaluates a different investment thesis — compounding, mispricing, or efficient growth.
Example: ACME Corp
TechnologyStage 4 · Multi-Track Scoring
Three independent tracks. Multiplicative scoring.
With factor scores in hand, AAPL enters the multi-track scoring system. Each track evaluates a different investment thesis through four gates. Scoring is multiplicative — one weak gate kills the score. A company can't compensate for a missing moat with cheap valuation.
Track A — Compounder
Identifies businesses with durable competitive advantages and strong reinvestment engines — the kind that compounds value over long holding periods.
Four gates — all must pass
Multiple structural signals of a durable competitive advantage
Retained earnings are being deployed at high incremental returns
Management allocates capital with discipline — buybacks, dividends, or reinvestment
The stock's price doesn't already reflect perfection — room for upside exists
Track B — Mispricing
Identifies stocks trading at a significant discount to intrinsic value with a catalyst to close the gap.
Four gates — all must pass
Multiple valuation methods agree the stock is cheap — not just one ratio
A floor exists on how much you can lose — asset backing or cash flow stability
Insider buying, institutional accumulation, or earnings momentum to trigger re-rating
Cheap for a reason doesn't qualify — a minimum quality bar must be met
Track C — Efficient Growth
Evaluates high-growth companies on unit economics, capital efficiency, and growth durability.
Four gates — all must pass
Rule of 40 score or strong revenue growth with high gross margins
Stable or expanding gross margins with operating leverage
Incremental returns on invested capital exceed cost of capital
Growth deceleration is manageable and addressable market has headroom
The efficient growth track runs for growth-style companies only. Investment style is classified by majority vote across four signals: valuation multiple, revenue growth rate, earnings acceleration, and R&D intensity.
Example candidate journey
7-stage pipelineAs a company's fundamentals improve and the market hasn't repriced, the composite tier rises. The engine tracks this progression automatically.
Stage 5 · ML Refinement
Deterministic first. Machine learning second.
The deterministic scores are now refined by machine learning models trained on the system's own prediction history. AAPL's factor profile is compared against patterns the models have learned from thousands of previous scoring cycles.
Weekly training cycle
Models retrain every Saturday at 2 AM UTC on 90+ days of accumulated scoring data. Short-lived market noise washes out; persistent patterns surface.
Quality gate
ML models are only activated when their prediction quality (rank IC) exceeds 0.15. Below that threshold, the system runs on deterministic scores alone.
Cluster + anomaly detection
Cluster models group similar stocks by factor profile. A variational autoencoder (VAE) detects anomalous factor patterns that may signal mispricing or risk.
Bounded adjustments
ML can adjust scores up or down, but adjustments are bounded and auditable. No black-box overrides — every change is logged with the model version and input features.
Graceful degradation
The system works fully without ML. If no qualified model exists, deterministic scores flow through unchanged. ML refines — it never replaces.
ML AdjustedWhen you see this badge on a score, it means the deterministic output has been refined by these models. The original deterministic score is always available for comparison.
Stage 6 · Smart Money Overlay
What institutional investors are doing with AAPL.
Finally, the system checks what institutional investors are doing with AAPL. This isn't a standalone signal — it's one input among many that feeds into the broader scoring framework.
13F filings
Institutional managers with $100M+ AUM are required to disclose equity holdings quarterly via SEC 13F filings. The system ingests these filings to track what institutional investors are doing with every stock in the universe.
Accumulation signals
Changes in institutional positioning feed into the catalyst strength score within the Momentum pillar. When multiple high-conviction managers are accumulating a position, that signal strengthens the overall catalyst assessment.
Curated manager list
Not all institutional managers are equal. The system maintains a curated list of high-conviction investors — managers with concentrated portfolios and strong long-term track records — and weights their activity more heavily.
45-day reporting lag
13F filings are due 45 days after quarter-end. Positions may have changed since the filing date. The system accounts for this lag — institutional signals are one input among many, never a standalone signal.
Institutional positioning is a confirmation signal, not a primary driver. The engine will never recommend a stock solely because institutions are buying it.
Stage 7 · Position Sizing
After all stages, AAPL receives its final output.
Seven stages of analysis produce a set of concrete, actionable outputs. No ambiguity — you see exactly why a stock scores the way it does, what composite tier it earns, and how much of your portfolio it warrants.
Example output — AAPL
Factor breakdown
Composite tier
How strongly the factor evidence supports the investment case. Ranges from NONE to EXCEPTIONAL based on gate alignment.
Opportunity type
Whether the stock qualifies as a Compounder (durable advantage), Mispricing (discount to intrinsic value), or both.
Suggested position size
An allocation percentage calibrated to composite tier strength and opportunity type. A higher composite tier earns a larger position.
Factor breakdown
The individual Quality, Value, and Momentum percentile scores — so you see exactly which dimensions are driving the composite tier.
Price vs. Margin Invest Value
When the current price falls below the buy price, the signal is Buy. Between buy and sell, it's Hold. Above the sell target, it's Sell. The margin of safety widens or tightens based on how much the valuation methods agree.
Next Steps
How to use these outputs.
The engine replaces the tedious parts of investment analysis — data gathering, normalization, cross-factor comparison, and ranking. The judgment call on whether to act is always yours. These guides walk you through practical workflows.
Transparency
We show our work because we trust our work.
Most investment tools hide their methodology behind vague descriptions. We publish ours in full. If you can't verify how a score was calculated, you shouldn't trust it.
Deterministic
Same inputs produce the same outputs. Enter AAPL today and tomorrow with the same data — the scores will be identical. No randomness, no hidden state, no human override.
Published formulas
Every formula used in scoring is documented in our guides. You can verify any factor calculation yourself with publicly available financial data.
Known limitations
The engine scores currently-listed equities only — delisted stocks are not included, which introduces survivorship bias in historical comparisons. Data can be delayed, restated, or incomplete. Conviction thresholds are theory-derived pending empirical backtesting validation. The engine cannot capture qualitative factors like management quality, regulatory changes, or geopolitical risk.
See the full pipeline in action. Score any stock and get the complete factor breakdown, composite tier, and position sizing.