How We Build Research

This document describes the analytical architecture under which Vector Research originates investment research. It is intended for readers who require an audit trail of the process behind a published view, and for counterparties evaluating the integrity of our research infrastructure. The framework below is the reference architecture; an individual report may apply a subset of these lenses with depth proportionate to the call being made.

1. Investment Philosophy

Vector Research operates on a single premise: institutional-grade equity research is breadth-bound, depth-bound, and editorial-bound, and the practice that closes all three constraints simultaneously tends to produce non-consensus signal. Breadth is the universe an analyst can see; depth is the rigor with which any single name is examined; editorial is the discipline of declaring a conviction-coded view rather than recycling consensus. Each constraint is binding in isolation. Our objective is to be bound by none.

Operationally, this means: computational systems carry breadth, analysts carry depth, editorial process carries discipline. We do not regard quantitative and fundamental research as competing paradigms; we regard them as inputs into the same Bayesian update on a thesis.

Vector Research operates a human-led, computationally- augmented research process. Analysts set the analytical agenda and frame the thesis; computational systems — including large language models and structured data pipelines — gather, organize, and surface candidate signal from primary-source material; analysts then review, weigh, and decide what is published. The analyst sits at both ends of the loop: originator of the question and author of the conclusion. The machine works in between.

2. Coverage Universe and Selection

Our coverage universe is opt-in, not survey-based. We do not aspire to publish on every constituent of any benchmark. We initiate coverage on a name when at least one of the following holds:

• Mispriced thesis — consensus has, in our read, not absorbed a structural change in the issuer’s economics, competitive position, or addressable market.
• Coverage asymmetry — the name is under-followed by qualified analysts, or follower coverage is structurally compromised (sell-side conflicts, siloed analyst beats, captive distribution).
• Information asymmetry — primary-source data (filings, transcripts, court records, regulatory dockets, foreign sources) is materially under-utilized in the public record.
• Catalyst opacity — identified upcoming catalysts (earnings, regulatory milestones, capital events, structural transactions) are inadequately mapped in the published record.

We do not initiate solely because a name is liquid, popular, or within a benchmark. We do not initiate at the request of issuers, promoters, or any party with an economic stake in our coverage. The initiation decision is a discretionary editorial judgment by the analyst, subject to internal review.

3. The Analytical Architecture

A Vector Research report is, in its general form, a structured triangulation across seven independent analytical lenses. The lenses are designed to be complementary: each addresses a class of question for which the others are weak. A thesis that survives confirmation across multiple lenses is, as a rule, more durable than a thesis that depends on any single lens in isolation.

• Lens 1 — Fundamental: what the business actually earns, owns, and owes; what its cash conversion looks like; what the equity is worth on disclosed economics.
• Lens 2 — Technical & Price-Action: what the tape says about the marginal-buyer / marginal-seller balance, trend integrity, and key reaction levels.
• Lens 3 — Volatility & Options-Implied: what the options market is pricing about the distribution of forward outcomes, dealer positioning, and tail risk.
• Lens 4 — Positioning & Flow: who already owns the security, who is selling it, and what their incentives are.
• Lens 5 — Cross-Sectional & Comparative: how the name prices relative to its peer cohort, sector, and structural comparables.
• Lens 6 — Catalyst Mapping: the calendar of identified events that will, in our read, force re-pricing.
• Lens 7 — Value-Chain & Industry Architecture: where the issuer sits in the rent-capture stack of its industry, and how that position is defended or eroded.

4. Lens 1 — Fundamental Analysis

Fundamental analysis is the foundation; everything else weights evidence around it. Our fundamental work follows the conventions of institutional sell-side research and CFA Institute guidance, with operational extensions noted below.

4.1 Financial-Statement Analysis

• Three-statement reconciliation across multiple fiscal quarters and years (annual and interim filings, including foreign equivalents where applicable), normalized for non-recurring items, accounting changes, and restatements.
• Quality-of-earnings review: cash conversion, accruals dynamics, capitalization policy, working-capital behavior, and deferred-revenue and deferred-cost patterns.
• Segment reporting where disclosed, cross-checked against management commentary, capital-allocation patterns, and geographic or customer concentration.
• Off-balance-sheet review: lease structure, purchase commitments, performance guarantees, contingent consideration, litigation reserves, and material unconsolidated entities.

4.2 Cash-Flow Modeling

• Free-cash-flow architecture: FCFF (firm) and FCFE (equity) constructions, with explicit treatment of stock-based compensation, working-capital normalization, and reinvestment assumptions.
• Discounted-cash-flow modeling under a multi-stage construction (high-growth → transition → terminal), with explicit decomposition of value attributable to each stage. Where the terminal value exceeds 70% of total enterprise value, we treat the model as a terminal-value model and label it accordingly.
• Sum-of-the-parts (“SOTP”) constructions for diversified issuers, with each segment’s implied multiple tested against pure-play comparables. SOTP is preferred where segment-level economics differ materially.
• Sensitivity and scenario analysis: bull / base / bear scenarios with defined trigger conditions and explicit deltas in WACC, terminal growth, and key operating drivers.

4.3 Cost of Capital

We construct a discount rate using a CAPM-anchored cost-of-equity with size and quality adjustments where reasonably justifiable, and a market-rate cost-of-debt informed by current credit spreads and observed issuance. We disclose the rate constructed and its components. We do not regard the discount rate as a parameter to be tuned to a desired conclusion; sensitivity to it is presented explicitly.

5. Lens 2 — Technical and Price-Action

Where applicable, we read the tape. Technical analysis is not used to determine the thesis; it is used to characterize the marginal-buyer / marginal-seller balance, identify the levels at which the read of the tape changes, and inform an analyst’s understanding of how positioning has accumulated around the current price.

• Trend structure: moving averages (50-day, 200-day; 20-week and 50-week for swing horizons), regression channels, higher-highs/higher-lows discipline. We use moving averages as state variables — above-or-below tells us which cohort owns the name — not as predictive signals.
• Support and resistance: high-volume levels identified via volume-by-price (VAP) histograms, prior swing extremes, and round-number psychology. Levels are presented with the rationale that produced them.
• Momentum oscillators: RSI(14), MACD(12,26,9), rate-of-change. We use oscillators principally to flag divergence between price and momentum, not as standalone buy/sell triggers.
• Volatility envelopes: Bollinger Bands (20-period, 2σ), ATR(14), Bollinger band-width history as a regime indicator (compression vs expansion).
• Volume profile: volume-weighted price action, high-volume nodes (HVN) and low-volume nodes (LVN), session opens and prior-period reference levels (PVAH, PVAL, POC).
• Relative strength: relative-strength lines against sector and benchmark to characterize whether the name is a leader or laggard within its cohort.

6. Lens 3 — Volatility and Options-Implied

The options market prices a forward distribution. Where the listed options market is liquid enough to be informative, we read it. The purpose is not to construct trades; it is to extract what the marginal options participant is pricing about the name and its forward catalyst path.

• Implied volatility term structure: the curve of ATM IV across expirations. Backwardation (front high, back low) reads as event-driven; contango as steady-state. We compare observed term structure to historical regime.
• Volatility skew and risk reversal: the relationship between OTM put IV and OTM call IV at constant delta (typically 25Δ vs 75Δ), and the standard 25Δ risk reversal. Persistent put-skew expansion against historical norms is a positioning signal we weight.
• IV percentile / IV rank: the realized location of the current IV against its trailing 1y / 2y distribution.
• Realized vs implied: trailing realized volatility (close-to-close, Parkinson, Yang–Zhang) against ATM IV to characterize whether the options market is paying for or selling realized risk.
• Dealer positioning and gamma exposure (GEX): where reasonably estimable from open interest, an estimated dealer GEX profile that helps characterize the stock’s likely behavior in different price regimes (positive-gamma pinning vs negative-gamma reflexivity).

7. Lens 4 — Positioning and Flow

We treat positioning data as a structural prior on price behavior: who already owns the name, who is unwilling to add, and who is forced to sell.

• Form 4 insider activity: rolling timeline of officer and director transactions (purchases vs sales, discretionary vs 10b5-1, dollar-weighted), with cluster analysis for buying or selling concentration.
• 13F institutional holdings: quarter-over-quarter changes in qualified institutional holders, with attention to new initiators, full exits, and concentration shifts among long-only versus hedge-fund cohorts. Lagged data, treated accordingly.
• 13D / 13G filings: beneficial-ownership accumulations above the 5% threshold, distinguishing passive (13G) from active (13D) postures.
• Short interest and days-to-cover: reported short interest, days-to-cover, borrow rates and availability; historical squeeze potential.
• ETF flows: creation/redemption pressure on relevant ETFs and the implied flow into or out of the name.
• Buyback and issuance cadence: announced and executed buyback patterns, share counts over time, ATM offerings, and dilutive transactions.

8. Lens 5 — Cross-Sectional and Comparative

No equity is priced in isolation. Cross-sectional analysis sets a relative-pricing frame against pure-play comparables, sector, and the broader equity universe.

• Peer cohort construction: a curated cohort of three to six structurally comparable issuers, justified by revenue mix, end-market, capital intensity, and growth phase — not by GICS classification alone.
• Multiples panel: P/E (TTM and forward), EV/EBITDA, EV/Sales, P/B, P/FCF, with attention to definitional consistency across the cohort and the issuer’s own historical range.
• Quality panel: gross margin, operating margin, ROE, ROIC, FCF yield, leverage, and growth durability against the cohort.
• Relative-strength matrix: short-, intermediate-, and long-horizon price returns against each peer to characterize leadership rotation.
• Correlation matrix: rolling correlations within the cohort to identify regime breaks.

9. Lens 6 — Catalyst Mapping

We map the next twelve months’ identified catalysts as a structured calendar. Each catalyst is annotated with: (a) date or date window, (b) information content (what the market does not yet know), (c) base-case expected outcome, (d) range of plausible outcomes, and (e) directional implication for the central thesis. Catalyst categories typically include:

• Earnings releases and pre-announcement windows (and their associated quiet periods).
• Regulatory milestones (FDA action dates, antitrust review windows, FCC and FERC orders, foreign equivalents).
• Industry conferences and primary-research touchpoints.
• Capital events: refinancings, lock-up expirations, secondary offerings, tender offers, M&A close dates.
• Index inclusion / exclusion and benchmark reconstitution.
• Litigation milestones: class-certification, summary-judgment, settlement, appeal windows.

10. Lens 7 — Value-Chain and Industry Architecture

Where the thesis depends on industry structure, we decompose the value chain into its constituent layers, identify where economic rents accrue, and assess the durability of the issuer’s position. The framework draws on standard strategy-research conventions and operationalizes a layered value-chain decomposition and rent-capture audit in our industry work.

• Layer decomposition: from raw inputs through intermediate processing, integration, distribution, and end demand.
• Rent-capture mapping: for each layer, which incumbents capture economic rent, what mechanism produces the rent (network effect, switching cost, scale economy, regulatory moat, IP), and how durable the mechanism is under stress.
• Moat-quality assessment: wide / narrow / none, labeled with confidence per Section 4 of our Editorial Standards.
• Strategic 2×2 framing: where helpful, orthogonal axes that locate the issuer against competitors on dimensions that are dispositive for the thesis.

11. Computational Infrastructure

Our computational stack is research infrastructure. It does not form theses or pass editorial judgment. Its function is to lower the cost of breadth so the analyst can spend marginal time on depth.

• Document ingestion: structured extraction from SEC EDGAR, foreign regulatory equivalents, earnings transcripts, investor presentations, and court records, with provenance retained for citation.
• Quantitative screening: multi-factor screens across the public-equity universe, with explicit safeguards against the standard backtesting biases (see Section 14).
• Pattern recognition: structured similarity search across historical analogs, comparable transactions, and prior cycles for an issuer or sector.
• LLM-augmented synthesis: large language models are used for retrieval, drafting scaffolds, summarization, and adversarial framing. They are subject to known failure modes (hallucination, retrieval drift, training-data bias) and their outputs are verified against primary sources before any factual claim derived from them appears in a Report.
• Editorial guardrails: automated checks for citation discipline, numerical consistency, restricted-list conflicts, and personalized-advice language.

12. The Editorial Pipeline

A Report typically progresses through the following stages. Stages may be compressed or adapted for time-sensitive material.

• Intake: the analyst proposes coverage initiation, articulates the candidate thesis in falsifiable form, and identifies the lenses on which the thesis principally relies.
• Research: primary-source review, model construction, cross-lens synthesis, and adversarial pre-mortem.
• Draft: the Report is written to publication standard, with confidence labels and sources attached.
• Review: per Section 6 of our Editorial Standards, including thesis test, evidence audit, devil’s advocate pass, numerical sanity pass, and compliance pass.
• Certification: the analyst affirms that the views expressed reflect their own independent judgment.
• Publication: the Report is released, the trading embargo lifts subject to anti-scalping policy, and the catalyst map governs subsequent monitoring.

13. Conviction Coding

Each published thesis carries a conviction level (High / Moderate / Exploratory) per our Editorial Standards. Conviction encodes the analyst’s posterior probability that the thesis is correct, not a target return, sizing, or recommendation. Conviction is set at publication and does not roll forward in the absence of an updated Report. The full framework is set out in our Editorial Standards, Sections 3 and 4.

14. Limits of Methodology and Known Biases

The architecture above is constructed to mitigate, not eliminate, the following known sources of error in research output:

• Look-ahead bias: the use, in retrospective analysis, of information that would not have been available contemporaneously. We constrain backtests and analogs to point-in-time data where reasonably practicable.
• Survivorship bias: the systematic exclusion of delisted, bankrupt, or acquired issuers from a comparable cohort, which causes ex-post performance to overstate ex-ante opportunity. Cohort construction is documented and reviewed for survivor inclusion.
• Data-snooping bias: the discovery of patterns by repeated re-specification of the same dataset. We separate hypothesis-formation from hypothesis-testing data where the conclusion is statistical.
• Sample-selection bias: the construction of a comparable set whose membership criteria pre-load the result. Peer cohorts are justified on structural grounds, not on outcome similarity.
• Time-period bias: the dependence of a conclusion on a particular historical window. Where a finding is window-dependent, the dependence is disclosed.
• Confirmation bias: the tendency to weight confirmatory evidence over disconfirmatory evidence. The devil’s-advocate pass and pre-mortem are designed to counter this; they do not eliminate it.
• Hindsight bias: the tendency to reinterpret prior conclusions in light of realized outcomes. Closed-position reviews are conducted against the contemporaneously published thesis.
• Model risk: the risk that a model’s structural form misrepresents the underlying economics. We stress models for sensitivity to key assumptions and document assumption ranges.

We accept that no methodology eliminates these biases. We commit to identifying them, structuring the workflow to reduce their probability, and disclosing where a conclusion is materially dependent on a methodological choice that a reasonable analyst could make differently.

15. Continuous Improvement

The methodology described above is a working architecture, not a terminal one. We conduct periodic reviews of:

• Walk-forward validation of any quantitative signal we rely on, with out-of-sample windows that postdate the signal’s specification.
• Closed-thesis post-mortems: for any thesis we subsequently retract or supersede, an internal review of whether the error was in the evidence, the inference, the editorial process, or the world.
• Source-quality review: ongoing assessment of the reliability of secondary sources we cite, with downgrades and exclusions where reliability deteriorates.

16. Application to Reports

Not every Report deploys every lens. The proportion of analytical weight assigned to each lens is a function of the thesis: a balance-sheet-driven distress thesis weights Lens 1 heavily and may deploy Lens 2 only as an entry-level reference; a positioning-driven squeeze thesis weights Lenses 3 and 4 heavily; an industry-structure thesis weights Lens 7. The Report itself identifies the lenses principally relied upon. A Report that depends materially on a single lens is identified as such, and its conviction is calibrated to the narrower evidentiary base.

17. Updates

We may revise this methodology document as the architecture evolves. The version operative as to any Report is the version posted to this page on the publication date of that Report. Material revisions are reflected in the “Last updated” date above.