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.
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.
A structured company-level workflow: business context, evidence gathering, financial-quality review, valuation triangulation, risk review, sensitivity, and final synthesis. Built for one company at a time.
Read the full per-stock methodologyA monthly outlook on the broader market: regime scorecard, breadth, sector rotation, valuation dispersion across the index, and a probability-weighted scenario range. One report for the macro picture, refreshed every month.
Read the market-level methodologySix 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.
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.
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.
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.
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.
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.
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.
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.
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 modelWhat 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 modelWhat 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 modelWhat 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 modelHow 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 modelTwo 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.
Methodology v2.5
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 changelogFigures, 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.