Methodology
Every company evidence library starts from a required primary-source checklist, with retained sources and material retrieval gaps disclosed at the page's information cut-off. Review depth varies by page; a source pack, full spread or independent review is claimed only when that artefact exists. Public evidence pages contain no equity ratings, price targets, fair values or buy/sell/hold recommendations. Independently derived credit views, where present, are explicitly labelled as non-official.
The pipeline aims to keep every load-bearing figure traceable to retained evidence and to label what cannot be fully closed. New to the pages? Start with how to read an evidence library.
The pipeline
- Source checklist. Issuer and exchange announcements, annual and interim reports, relevant circulars, debt and rating documents, and regulatory or industry sources are checked to the stated cut-off. The retained pack and material retrieval failures are named; a configured monitor is not treated as proof that the tape was completely scanned.
- Spread. Where the analytical scope requires it, financial statements and segment data are re-keyed from primary filings rather than imported from an aggregator. The page or assurance note states the period and perimeter actually covered. Aggregator figures, where used, are labelled (E) and are not load-bearing.
- Segment economics. Segments are analysed on capital efficiency (return on segment assets and on net capital employed) and complementarity — how the segments feed each other — not just margin.
- Curious pass and skeptical pass. One pass hunts for what the accounts quietly show (inter-segment flows, fee-derived volume proxies, language deltas in successive outlook statements); a second pass attacks the emerging view — every analytical claim must survive a named best-alternative explanation.
- Quality control. Cross-footing, cash-flow or balance-sheet reconciliation and claim-to-source review are applied to the artefacts the page carries. Observable controls are reported on the quality dashboard; their presence does not certify accuracy or source completeness.
Evidence grading
| Tag | Meaning | Example |
|---|---|---|
| R | Read directly from a primary document | A segment table from an audited annual report |
| D | Derived arithmetically from R inputs; the derivation is shown | Return on net capital employed, recomputed from the segment note |
| G | Guidance stated by management, dated to the source | A volume, margin or capital-spending range in a results announcement |
| E | External or secondary source; treated as unverified colour | Press citations of ministry statistics that the ministry itself does not publish |
| O | Opinion or judgement; flagged as such at the point of use | An assessment of management bench depth |
| F | An analyst forecast, distinct from company guidance | A modelled forward estimate shown with its assumptions |
Industry economics before comparison
The industry economics map groups companies by a shared causal mechanism, not by a ready-made valuation peer set. Each company is mapped through demand, pricing, capacity, cost or funding and the next observable. A comparison is used only after period, perimeter, unit, accounting, ownership and business mix are reconciled; otherwise it remains a question to test rather than a number transferred into the company.
For funded lenders, the map separates customer or pledge volume from average ticket and collateral-price effects. For finite-life assets, it separates current cash generation from asset ageing, replacement and funding. For project businesses, it follows order book through margin, working capital and cash. This is why a large market, rising asset value or growing backlog does not by itself prove value creation.
Data and news scanning
Each company has a dated information cut-off. A reliable update scan distinguishes issuer, exchange, debt, rating, regulator and causal-industry sources, and records the newest item seen plus any access or pagination limitation. “Monitoring” means the page remains in the maintained library; it does not mean every later filing has been processed. The quality dashboard separately shows which advertised pages currently have an automated official-tape source configured.
Falsifiers
Where a page carries an analytical hypothesis or forecast, it should name the future disclosure or threshold that would confirm or refute it and, where knowable, the expected date. If a claim cannot name a falsifier, it should be narrowed or labelled opinion. This makes the claim gradable when the next relevant evidence arrives.
What the public site deliberately does not do
- No public equity ratings, price targets or fair values. The public evidence pages describe businesses, their economics, their funding and their risks. They do not say what an equity security is worth or what a reader should do. Private working materials used in the author's own process are outside the public product and are not offered, published or distributed. A public page may include an independently derived indicative credit view; where present, it is explicitly labelled non-official and is not a buy, sell or hold recommendation.
- No personalised advice. Company questions, factual corrections and stock ideas are welcome through the contact page. Requests about an individual's circumstances, portfolio or trading decision are not answered as investment advice, and a personal reply cannot be guaranteed.
- No paid coverage. No issuer, broker or sponsor pays for anything on this site, and there are no affiliate links.
- No silent edits. Corrections are dated and described. Every note states its information cut-off.
Independence, conflicts and how the work is produced
- No paid coverage, no sponsors, no affiliate links. No issuer, broker or sponsor pays for anything on this site, and coverage is not offered for a fee.
- The author may hold or transact in securities mentioned. Public pages carry no view on any security, so no page can be read as a recommendation in either direction; the private working view is not distributed.
- AI-assisted and source-controlled. AI may assist document collection, spreading, drafting and adversarial review, but it is not evidence. Load-bearing claims should remain traceable to retained sources and derived figures should identify inputs or method. Multiple role or model passes using the same evidence are challenge checks, not independent factual assurance. An independent review is claimed only where a page-specific, dated assurance note records its scope, findings and limitations.
- Coverage selection. Companies are chosen for being undercovered relative to the depth of their public filing record — typically Singapore and Asian listed small and mid caps with little or no sell-side coverage. Reader suggestions are welcome through the contact page; neither coverage nor a reply is guaranteed.
- Update cadence. Each page states its information cut-off and may identify a next scheduled update. The library is reviewed on a scheduled and event-led basis, not continuously, and no turnaround promise is implied. Current automated official-tape coverage is disclosed on the quality dashboard. Dated changes are listed in the Atom feed.
Common questions
- Does SMID Research publish ratings, price targets or fair values?
- Not on any public page. The public evidence libraries describe a business, its economics, funding and risks from primary filings and name the disclosures that would settle the open questions. Separate access-controlled working research exists and is not available for distribution.
- Is this investment advice?
- No. Everything on the public site is general-circulation information. It does not take into account anyone's objectives, financial situation or needs, and it makes no recommendation to buy, sell or hold any security. Readers should consider suitability with a licensed financial adviser.
- Where do the figures come from?
- Principally from retained issuer and exchange filings, supplemented by identified debt, regulatory and industry sources. The exact period and source scope varies by page and material retrieval gaps should be named. The R/D/G/E/O/F taxonomy distinguishes reported, derived, guidance, external, opinion and forecast claims where the page uses those labels.
- How do I know when a page was last updated, and what changed?
- Each page states its information cut-off date in its status line. Pages with a corrections section record substantive changes there, and the Atom feed lists library changes newest first. The quality dashboard shows corrections-section coverage across advertised pages.
- Can I ask whether to buy, sell or hold, or about my own holdings?
- No. Questions about a company, its filings or the research method, factual corrections and suggestions for companies to cover are welcome. Anything touching an individual's own position, portfolio, timing or trading decision is not answered, however it is phrased.
- Is the research produced with AI?
- Yes. AI assists collection, extraction, reconciliation, drafting and adversarial review, but it is not a source. The intended discipline is that load-bearing figures remain traceable to retained evidence and limitations are stated. Review depth is page-specific; an independent review is claimed only where a dated assurance note records it. Errors still happen and are corrected with a dated note.
- Can I cite or reuse the material?
- Yes, for search, reference and AI-grounding use, with a link to the canonical page and the page's stated information cut-off. Figures are as-filed by the issuers and may be superseded by later filings. Model-training use is not permitted; the crawler policy is in llms.txt.
Corrections
Errors, once identified, are corrected in place with a dated correction note describing what changed and why. Where a prior figure was materially wrong, the wrong figure is shown struck through alongside the correction — the record of having been wrong is part of the record.
Questions, corrections and stock ideas
Readers can contact SMID Research with company questions, primary-source corrections and ideas for undercovered listed companies. Messages may inform a dated update to the evidence library, but neither an individual reply nor future coverage is guaranteed.