What reaches a company, and whether it matters.
Exposure does not forecast. It reads what governments publish, keeps what reaches the company through something the company itself disclosed, and tests each change for materiality.
Nearly four rows in five are one document returned by more than one search. Of the 7,607 documents, one in 36 reaches the company; one in five of those is assessed in writing; six make the front page.
Four rules that override everything else.
They are written down, they are the same for every company, and every edition shows where each one was applied.
- 1
Nothing is invented.
No source, locator, date or quantity. What is absent is printed as not disclosed, not monitorable or human review required, together with what would close it.
- 2
A retrieval asserts its own completeness.
A count that looks fine is not proof. A company’s own site once returned 9 of 109 records with no error and HTTP 200.
- 3
A scan of nothing is not a clean scan.
Every source is checked against a record known to be in it; if the check fails, the source is marked failed, not trusted. An official sanctions file: 5,708 records, not one name.
- 4
A conflict is printed, not resolved.
When sources disagree, both figures appear with their sources. One company publishes four different counts of itself.
Eleven stages, with the checks built into them.
Code decides; models only propose. A stage that fails its own test stops the edition, and a person is asked at fixed points, never in between.
Who it is, where it operates, and how much of it sits in each place.
- Identify the legal entity and the group around it.
- Footprint: every site to the city, customers and suppliers, and the company’s own named risks, word for word.
- Weigh four separate shares: revenue, headcount, facilities and assets, and supplier spend. A person confirms them.
- Deep scan: research on the company’s sector, countries and themes, from peers, think tanks, consultancies and academia.
- Categories: the kinds of exposure the company has, and those it provably does not, each with evidence.
- Scope: who can act on the company, and where. A place with no source is marked not monitorable.
What governments published, and whether it can reach this company.
- Sweep the official sources in code, in their own languages. A source older than seven days is printed as stale.
- One shared reading per law: whom it binds, what it does, from when. It is made once, reused for every company, and each quote is checked character by character.
- A first judge asks whether each law could reach the company through one of eleven channels, by a named object.
- Triage: keep, or set aside with a reason. A second, blind judge re-reads a sample of both, and a person settles every disagreement.
- Materiality: five tests choose the front page; the rest go to the watch register.
What it does to the company, in its own units, checked against the law.
- Translate: one sentence per item, in the company’s own units. Text that would fit another company fails.
- An action to consider per item, with a sourced date and the function it falls to, or none, with the reason.
- Verification against the law’s own text: every quote, date and number is checked against the law itself. One false claim stops the edition.
- Attack: a hunt for defects in a frozen copy. Any defect that blocks release holds the edition until it is fixed.
- Render the dashboard, the brief in three lengths and the board pages, with a coverage statement drawn from the source log.
Five tests separate a change that matters from news.
An item reaches the front page only by passing all five. Every other assessed change goes to the watch register, carrying the name of the test that held it back. A test that cannot measure says so; it never passes by default. No override exists.
- 1
Weighed
Does it touch a place or a relationship the edition has weighed?
- 2
Channel
Does it reach the company’s economics through one of eleven named channels?
- 3
Size
Is the weight behind it above a threshold derived from the company’s own figures?
- 4
Actionable
Is there a date, a window or an eligibility test the company can act on?
- 5
Judgement
Would its absence change the reader’s judgement, or is it outweighed by another item of the same kind?
The eleven channels, a closed list
- Trade instrument
- Export or import control and licensing
- Sanctions and financial access
- Regulation and standards
- Ownership, screening and expropriation
- Input price and availability
- Physical access and logistics
- Demand and market access
- Capital, currency and payments
- People and talent
- Reputation, licence to operate and litigation
A change that reaches the company through none of them is not shown.
Four weights, five bands, and no single score.
- Four weights
- Revenue, headcount, facilities and assets, and supplier spend, each at the detail its source publishes and never finer. A share that is not published is a blank with a named request, never an estimate.
- Five-level bands
- Severity and likelihood are each rated on five levels. No composite exposure score is ever computed.
- Bands, not forecasts
- Likelihood is given in bands, never as a point probability. Likelihood against impact appears only as a labelled preview.
- Status by date
- Act when a date has passed, falls within 30 days, or carries major or severe impact within 180 days; then Prepare and Watch.
Registered, scored probabilities belong to Strature Briefs and forecasting engagements, which use the six frameworks. Exposure shows where a company is exposed now; the six frameworks ask what may come next.
Every edition prints what it set aside.
The counts are from the RBC edition of 28 Sep 2026. RBC is the subject of the analysis, not a client.
- 34,333rows readRead the official sources.
- 7,607distinct documentsCount each document once.
- 210reach the companyKeep what can reach the company.
- 41assessed changesWrite each one: what changed, why it reaches the company, an action to consider.
- 6front pageRank by impact and date; the rest go to the watch register.
What was set aside, and why
- Read
- Fifteen official sources for the company’s footprint, from the Canada Gazette and OSFI to the US Federal Register and EUR-Lex.
- Deduplicate
- Duplicates fold into one record, and the totals are made to add up in print.
- Match
- Each document is tested against the company’s own disclosures. 3,196 were set aside with a reason; 4,219 are not yet judged, never counted as clean.
- Assess
- Each assessed change quotes the official text and its date. 31 of the 41 were verified; the rest say why not.
- Rank
- Six made the front page; 35 sit on the register with the reason they are not in front.
Source for every count: Strature Exposure, RBC edition of 28 Sep 2026, the coverage funnel, coverage tiers, summary counts, the sources table and the rank rule.
It cannot prove completeness, and says so.
Each edition states what was swept and which gaps are declared. It never claims to hold every instrument. Actions to consider are illustrations, not instructions, and not legal, tax, accounting or investment advice.
Six more limits, printed with the work
- Some sources cannot be monitored, among them some legislatures, municipal councils and several US states. No Chinese government source is used, by licence policy.
- Revenue by country is not shown where the company’s accounts do not report it, and supplier geography needs the client’s own payables data.
- When a run’s budget binds, the changes it did not reach are printed as not judged or not written, never counted as clean.
- The second judge reads a sample. The recall figure it gives is a floor, not a proof.
- Zero is not zero exposure. A place with no items is a coverage statement, printed per place. Corrections are printed, not overwritten.
- Client material never leaves its client. It is kept per client, never pooled, and never used in a published brief.
Research is produced with AI-assisted tools under human direction.
Tell us the decision you are facing.
The Principal reads every message and replies from the firm’s own address. No mailing list, no tracking.