How to read a company dossier

A dossier is a file of claims, not a verdict

An OpenDossier counterparty file is organised as sections (identity, ownership, people, financial, legal, intellectual property, web and documents, feedback) and each section is a list of claims. A claim is one checkable statement: incorporated in Delaware on 9 February 2022, no federal civil cases found as a party, SOC 2 Type II report supplied under NDA. Every claim carries its own evidence chain and its own verification status. Nothing is averaged.

This structure is deliberate. A counterparty file is read by different people for different reasons: procurement wants security evidence, finance wants capital and insolvency signals, legal wants litigation and sanctions, a board wants to know who controls the company. Each reader needs to see the underlying record and its date, not a blended opinion. The dossier therefore reads like a well-kept case file: short statements, each with a citation, each with a note on what the source does and does not establish.

Read the summary first for orientation, then go to the sections you are responsible for, and always open the chain on any claim you intend to rely on.

The evidence chain, link by link

Each claim shows seven fields. They answer the questions a careful reviewer would ask before relying on a record.

A claim with a thin chain, for example a source but no original location, should be treated as a lead and checked at the authority before use.

Verification status labels

Status describes how the claim was established, not how good or bad the fact is.

• Verified: read directly from an official or primary record (registry, court, regulator, IP office, filing) and matched to the entity.
• Company-stated: supplied by the company through its claimed Trust Workspace profile, such as a capitalisation summary or a policy. It is attributed, dated and signed, but not independently confirmed.
• Third-party: from a licensed or reputable non-official source, for example licensed review data or a news report, with the source named.
• Open: a question the analyst could not close; the dossier says what was asked and of whom.
• Disputed: the company or another party has lodged a reply that contradicts the record; both are shown.

Absence of a record is also a claim. No cases found is reported with the index searched, the search terms, the date and the known gaps in coverage. It never means no risk; it means a specific search, on a specific day, returned nothing.

Completeness, freshness and open questions

Three indicators sit at the top of every dossier. They describe the file, not the company.

Completeness is the share of the standard section checklist for that jurisdiction and company type that has at least one verified or company-stated claim. A private Japanese 株式会社 cannot reach the same completeness as a US public company on public sources alone, because shareholder lists are not public in Japan; the gap is shown rather than hidden.

Freshness is the age, in days, of the oldest retrieval date among the claims that change most often: registry status, officers, charges, litigation, sanctions. A freshness figure of 12 days means those sources were re-read within 12 days; it says nothing about slow-moving records such as a 2019 patent grant.

Open questions counts items the analyst asked the company or an authority to resolve and that remain unanswered. A high count on a new file is normal; a high count on a mature file where the company has claimed its profile and not answered is itself information.

Human review status

Every dossier records whether an analyst has reviewed it and when. Automated collection is good at reading structured registers and matching identifiers; it is weaker at judgment calls, such as whether two same-name entities are the same business, whether a court record describes an allegation or a finding, and whether a company-supplied document actually says what the company says it says.

The review line (for example Analyst-reviewed 21 Oct 2026) means a named analyst read the chains, resolved contradictions where possible, wrote the open questions and signed off the summary. Reviews are repeated when a monitored event changes a material section, and at a set interval for files in active use.

If the review line reads automated only, the file is still usable for orientation, but entity matches and litigation interpretation should be re-checked before the file supports a decision. The distinction is shown precisely so that readers do not give machine output more authority than it has earned.

Why there is no single trust score

Readers sometimes ask for one number. OpenDossier does not provide one, for four reasons.

• A score hides the chain. The useful question is not 72 or 68 but which record, which date, which gap.
• Risk is relative to the deal. A pending contract dispute matters to a supplier but not to a buyer of a one-off licence; a missing ISO 27001 certificate matters to a data-processing engagement and not to a marketing agency.
• Scores invite gaming. A company that learns the weights can optimise appearance rather than substance; a chain of dated records is harder to dress up.
• Weighting is a legal and commercial judgment that belongs to the reader, not to a data vendor. OpenDossier assembles evidence; it does not give advice.

Instead the dossier gives the indicators above, a short written summary that states what was found and what was not, and the ability to export the chain so that your own risk committee can apply its own policy.