Skip to content
StockMarketAgent
Discover
Methodology v2.5

How every report is produced.

Every covered company is analyzed the same way, every month — no exceptions, no favorites. The same disciplined sequence runs on a mega-cap compounder and a small-cap cyclical alike: we decide what kind of business it is, assemble the evidence, value it through several independent lenses, argue the bear case before the bull, score it against a fixed rubric, and only then synthesize a rating. What follows is the approach in plain terms — what each step does and what it produces. The specific parameters, weights and recipes that calibrate it stay in the engine.

6
Phases per report
5
Valuation lenses
~950
Tickers covered
18
Languages
v2.5
Methodology
Two frameworks

One discipline, two levels of resolution

Two public research frameworks sit behind the platform: one for individual companies and one for the broader market. We show the principles, the reader-facing checks, and the version history here while the internal orchestration, calibration controls and private benchmarks stay proprietary.

The sequence

Six phases, run in the same order, on every name

The order is the point. Framing comes before evidence; the bear case comes before the rating. Each phase below is named and described conceptually — enough to judge the rigor, without the internal thresholds that would let it be gamed.

View
Detail
01
Phase 01 · Framing

Classify the company

Before a single number is modeled, we decide what kind of business we are looking at.

Every company is sorted into one of our coverage archetypes — durable compounders, cyclicals, financials, capital-intensive operators, hyper-growth names and others. The archetype is the lens that follows the company through the rest of the analysis: it decides which valuation methods carry weight and how demanding each assumption should be. A deep cyclical and a wide-moat compounder are never judged on the same template.

What comes out
archetypeapplicable lensesnarrative frame
02
Phase 02 · Evidence

Gather the evidence

We assemble one complete, current picture of the business before drawing any conclusion.

The multi-year financial record, the latest quarter and forward expectations, the competitive position and sources of any moat, and the comparable set the company trades against are reconciled into a single evidence base. Nothing downstream is allowed to run ahead of the evidence that supports it.

What comes out
financial recordforward expectationspeer setmoat read
03
Phase 03 · Valuation

Build multiple valuation lenses

No single model decides fair value. We value every business several independent ways.

Intrinsic, scenario and relative methods are each built from the same evidence base, then triangulated. We pay as much attention to where the lenses disagree as to where they agree — wide dispersion is itself information. The result is a composite fair-value range rather than a single false-precision number.

What comes out
composite rangeper-lens estimateskey swing factors
04
Phase 04 · Risk

Stress-test the thesis — bear case first

We write the bear case before the bull case, on purpose.

The most credible ways the thesis could be wrong are documented first, then pressure-tested across downside scenarios, a 5×5 sensitivity grid, and a fixed accounting-quality gate. Only the conviction that survives that scrutiny is allowed to inform the upside view — a deliberate guard against the optimism that creeps into most research.

What comes out
bear casedownside scenariosquality-gate result
05
Phase 05 · Scoring

Score against a fixed rubric

Every company is graded on the same scorecard, so a rating means the same thing across the coverage universe.

Business quality, balance-sheet resilience, valuation and risk are scored consistently from name to name. The rubric is identical for a mega-cap compounder and a small-cap cyclical — there is no bespoke grading and no soft pages for names the firm happens to like. A score is comparable across the whole universe.

What comes out
category scorescomposite scoreconfidence read
06
Phase 06 · Synthesis

Synthesize a rating

Finally, the lenses and the scorecard are reconciled into one published view.

The composite fair value, the margin of safety the archetype demands, and the score combine into a single rating, a fair-value range and a stated confidence level — with the full reasoning written out in the report rather than hidden behind it. Every reader sees how the conclusion was reached.

What comes out
ratingfair-value rangeconfidencewritten thesis
Phase 03, expanded

The models we triangulate

Fair value is never one model’s answer. Five lenses are built from the same evidence base and reconciled into a composite range — each answers a different question, and each is documented in the per-stock methodology.

Intrinsic
Discounted earnings

What is the company’s future profit stream worth today?

Projects the SBC-adjusted earnings a business can sustainably generate and discounts them back at cost of equity — the anchor intrinsic estimate for stable, profitable companies.

Open the model
Scenario
Scenario forward earnings

What could it be worth across bull, base and bear?

Carries the business forward under three distinct futures and values each, so the range — not just the midpoint — is explicit. Especially useful where the path is uncertain.

Open the model
Implied expectations
Reverse DCF

What does today’s price already assume?

Inverts the usual question: instead of solving for value, it solves for the growth the current price implies — a clean test of whether the market’s expectations are reasonable.

Open the model
Cash floor
Owner earnings

What would an owner actually pocket in cash?

Strips reported profit back to the cash an owner could withdraw without starving the business (net income plus D&A, less maintenance capex) — a conservative floor that keeps accounting optics honest.

Open the model
Relative
PEG-adjusted peer multiples

How is it priced against comparable businesses?

Places the company against a like-for-like cohort on growth-aware multiples (forward P/E relative to growth), so the intrinsic work is sanity-checked against what the market pays for similar economics.

Open the model
Weights are not fixed across the board. How heavily each lens counts toward the composite depends on the company’s archetype — a financial leans on different lenses than a hyper-growth software name. The blend is part of the engine; the lenses themselves are open above.
Why it holds up

Two principles the process is built around

A quality gate every name must pass

Before valuation conclusions stand, each company runs through a fixed accounting-quality gate that checks whether reported profit is backed by cash (OCF/NI), whether accruals and the Beneish read look clean, and whether returns on capital are real and durable. A company that does not clear the gate is not failed outright — it is flagged, and the margin of safety we demand widens to match the doubt. The gate is the same for every ticker, which is what makes a clean pass meaningful.

The bear case is written first

Most research reaches a conclusion and then reaches for the evidence. We invert that. The bear case is documented before any bullish synthesis, so the downside is argued on its own merits rather than as an afterthought. The upside view has to survive a case that was built to defeat it — and the worst credible outcome anchors the floor of the fair-value range. Certain hard-fail conditions block a bullish rating outright.

Versioning

Methodology v2.5

Methodology v2.5 · current

The method is versioned, and every report is stamped

The methodology is treated as a living document under version control. Each published report records the version it ran under, so a number from one cycle can always be read against the rules that produced it. When the approach changes materially, the version increments and the change is logged — analysis is never quietly altered beneath you.

Methodology changelog
versionv2.5cadencemonthlyrendered18 languagesreviewanalyst-approved

Figures, tickers and units are invariant across languages — only the prose is translated. This page describes the approach and what it produces; it does not disclose the internal parameters, thresholds or weights that calibrate the analysis. Educational analysis only — not financial advice.