ABI Analytics
Hedge FundResearch Products
Methodology

How ABI Analytics Builds Its Research Output

Every claim in an ABI deliverable should be auditable by a junior analyst in under thirty seconds. That principle drives every design decision in our pipeline – from how we ingest data, to how we score companies, to how we frame outputs as analytical observations rather than investment directives.

6
Scorecard Pillars
Each independently weighted
100%
Primary-Sourced
Every claim links back
5:00 AM ET
Daily Refresh
Scores, news, filings, insider
<30 sec
Audit Time
From claim to source

Three Principles

Primary-Source Citations

Every number in a published memo or company page links back to its origin: an EDGAR XBRL concept and fiscal year, a press release page reference, an earnings-call transcript timestamp. We do not synthesize numbers from prose.

Analysis, Not Recommendations

Our published views are asymmetry profiles (Favorable / Neutral / Cautious), probability-weighted fair-value estimates, scenario-trees with explicit weights, and triggers that would shift the view. We do not publish Buy / Sell / Hold ratings.

Human Oversight at Every Step

LLMs handle aggregation, synthesis, and language drafting. A human reviewer with buy-side training inspects every output that goes to a client. We document where AI is used and where humans intervene.

The AI Composite Scorecard

The scorecard you see on every company page is a 0–10 composite built from six factor pillars. Each pillar is scored independently and combined with explicit weights. Pillar inputs are quantitative (from structured XBRL data and market data feeds) and qualitative (from NLP analysis of 10-K text, earnings call transcripts, and management commentary).

PillarWeightPrimary Inputs
Business Quality & Moat20%Segment-level economics, switching-cost evidence, regulatory positioning, R&D intensity (XBRL: ResearchAndDevelopmentExpense), 10-K Item 1 narrative
Financial Health15%Leverage (Net Debt / EBITDA), liquidity (Current Ratio), cash flow durability (5-yr OCF / NI), debt maturity walls
Profitability & Returns20%Through-cycle ROIC, EBITDA margin vs. peer median, margin trajectory, FCF conversion
Capital Allocation15%5-yr capital deployment mix, incremental ROIC on acquisitions, capital-return discipline
Valuation15%Forward EV/EBITDA and P/E vs. peer median and own 5-yr history, DCF-implied fair-value range
Sentiment & Momentum15%Sell-side action flow, insider transactions (Form 4), 50/200 DMA position, short-interest dynamics
Composite Score100%Weighted average; reported 0–10 to one decimal
How to read the score: 8+ is a top-decile setup across all pillars; 6–8 is institutionally interesting; 4–6 is balanced with specific flags; below 4 indicates concentrated weaknesses worth surfacing. The score is a navigation aid, not a directive – the underlying pillar detail is where the real analytical insight lives.

The Asymmetry Framework

For every covered name we publish a probability-weighted scenario tree with explicit bull / base / bear cases. Each case carries an assigned probability and an implied scenario value.

Base Case Fair Value

Our most-likely-case analytical estimate of intrinsic value. The anchor of the scenario tree, typically weighted 40–60%.

Bull & Bear Scenarios

Each with explicit probability (typically 20–30% each). Bull case if the structural thesis accelerates; Bear case if execution or valuation breaks.

Up/Down Ratio

The ratio of bull-case upside to bear-case downside. 0.5x means downside is 2x the upside reward. 2.0x is the reverse. We characterize as Favorable / Neutral / Cautious.

Data Sources

Structured Data

EDGAR XBRL companyfacts for fundamentals (every covered name, every available fiscal year). S&P Capital IQ for sell-side estimates, comparables, ownership, and transcripts. Market-data feeds for ratios and consensus consolidation. Yahoo Finance for live prices, 52-week ranges, and beta. PDUFA database for FDA catalyst dates. SEC Form 4 for insider transactions. 13F filings for institutional ownership.

Unstructured Data

10-K and 10-Q filing text for forensic diff analysis, risk-factor language drift, and quality-of-earnings checks. Earnings call transcripts (Capital IQ + supplementary sources) for management tone analysis and say-do scorecard construction. Press releases and 8-K filings for catalyst and event tracking. Sell-side research firm action announcements (publicly disclosed only).

Where AI Is Used (and Where It Isn't)

LLMs Handle

  • Aggregating structured data into prose
  • Comparing filing language across periods to surface material changes
  • Summarizing earnings call commentary
  • Suggesting initial scorecard pillar scores subject to review
  • Drafting standard product output templates

Humans Handle

  • Final scorecard pillar weights
  • Calibration of probability assignments in scenario framework
  • Judgment on whether a flagged change is material
  • Sign-off on every output before client delivery
  • Methodology adjustments and weight re-calibrations

We don't dress this up – there's a clear division of labor. The LLM is the analyst's faster colleague. It is not the analyst.

Update Cadence

D

Daily, 5:00 AM ET

Composite scores refresh; sell-side action ingestion; insider trade detection; pre-market news scan.

E

Earnings Event-Driven

Earnings reaction notes ship within 30 minutes of release. Slack-native delivery to subscriber workspaces.

F

Filing Event-Driven

10-K and 10-Q diffs run within 4 hours of EDGAR filing. Material-change alerts within the same window.

Q

Quarterly

Sector deep-dives, peer-comp recalibrations, asymmetry framework reviews.

What We Don't Do

Audit Any Claim | We Mean It

Open the MTRN sample memo or live company page. Click any primary-source link. Read it for yourself.

Open the Product Showcase