Editorial guidelines

Coverage decisions are made by the research desk and are not influenced by issuers, advertisers, or any third party with an economic interest in the conclusion. We do not accept paid placements, sponsored coverage, or compensation tied to a rating or fair-value range.

Analysts may not initiate, change, or remove a rating on a name in which they hold a personal position. Personal-position disclosures are kept on file for every active analyst and apply for the duration of coverage plus a thirty-day cooling-off window.

Every report is built on the same nine-phase methodology (currently v2.4), described in detail at /methodology. The framework is deliberately archetype-aware so a hyper-growth software business and a mature compounder are not forced into the same model; the archetype is named explicitly in the assumption ledger of every report.

Fair value is always expressed as a range, not a point estimate, with a confidence number attached. The range is the convergence band of multiple independent valuation models; the confidence score reflects data quality, model agreement, balance-sheet integrity, and accounting-quality gates.

External claims that quote a number cite the source and the retrieval date. Primary sources (SEC filings, company earnings press releases, central-bank releases) are preferred over secondary aggregators where they exist. Numbers that cannot be traced to a primary source are either reconstructed from disclosed components or omitted.

Reports do not attempt to forecast macro variables (GDP, inflation, interest rates) outside the scenario window disclosed in the assumption ledger. When a thesis is contingent on a macro outcome, the contingency is named explicitly in the risk section.

Every report is structured bear-case-first. The risks section and the explicit kill scenarios are written before the bull case so anchoring on the bullish narrative cannot suppress disconfirming evidence. The order is editorial discipline; the substance is the same in either order.

When a factual error is found in a published report, the correction is logged on the public record at /corrections with the date the error was identified, the section affected, and what changed. The original report carries a “last reviewed” timestamp that updates on every correction.

Errors are categorized as material (changes the rating, fair value, or a load-bearing claim) or non-material (typo, broken link, formatting). Material corrections trigger a ratings review; the rating may stand, may change, or may be withdrawn, and the post-correction state is recorded explicitly.

We are not a signal service, a copy-trading platform, a personalized advisor, or a newsletter. Every published report is independent equity research; readers must apply their own judgment about whether the thesis matches their portfolio, risk tolerance, and time horizon. Nothing on this site is a personal recommendation.