Cask publishes research anchored in primary documents. This page sets out where the material comes from, what will not appear here, how conflicts are handled, and what happens when we get something wrong.
Every substantive piece is anchored to something a reader can check: a named filing with its date, a regulatory document, an earnings call, a speech, a dated data release, or a named person speaking at a named venue.
Cask analyses that material. It reads language changes across successive filings, traces a disclosure to its consequence, and says what a document implies. The rule is that the analysis begins in a document rather than in a view.
Where a piece states what would resolve an open question, it names the specific document or disclosure that would resolve it.
Primary sources first, without exception. A filing, transcript or release is ingested and read before any outside commentary on it is consulted. Sell-side notes, news coverage and other analysts' work are used to enrich or challenge a primary read. They are never used in place of one, and they never anchor a piece.
Figures that are not read directly off a single document are identified as derived, and their constituents and sources are named. An aggregate across four companies says which four, which periods, and which filings.
Quotations are verbatim and attributed to a named person. Cask does not publish a quotation attributed to an unnamed employee of a named organisation.
Cask is built on a system that captures, structures, indexes and surfaces source material. A person decides what is worth a reader's attention and writes it. Nothing here is published automatically.
When a piece runs an argument against a forthcoming print, filing or data release, Cask publishes beforehand the specific result that would mean the argument is wrong. A condition that would falsify it, rather than one that would merely qualify it.
Publishing the condition in advance is the point. A note written privately before an event and produced afterwards proves nothing to a reader, because its timestamp is set by the machine that wrote it. A public statement carries a date the reader can check.
Where the schedule does not allow the condition to be published before the event, the piece says so, and says when the condition was set.
The result is then published either way, at the same length and in the same slot, whichever way it lands. A piece is not moved up the schedule because an outcome was confirming, and it is not allowed to slide because an outcome was not.
Cask Advisory research is written by Sebastien Davies, CFA, FRM, who is a partner at Primal Capital and is involved in building Brila and Elara. As a CFA charterholder he is bound by the CFA Institute Code of Ethics and Standards of Professional Conduct.
Cask has interests. This section sets out how they are handled.
Every piece is classified before publication:
Where the subject of a piece is a company in which Sebastien or Primal holds a position or a role, the piece is never classified as independent. Where a subject is a named competitor or counterparty of such a company, the classification is reviewed before publication rather than assumed.
Classifications are recorded at the time the determination is made, together with the basis for it, and are not reconstructed afterwards.
If Cask misstates what a document says, the correction is published promptly, on the original item and in the archive record. Corrected material is marked rather than removed, so the record of what was said and when it was said remains readable.
An error about a source is treated as more serious than an error of judgement, because the first is the thing Cask exists to get right.
Neither of these reached publication. They are included because a standard that shows its own caught errors is worth more than one that asserts rigour.
An aggregate that did not reconcile. A figure of $388bn for signed but not yet commenced data centre leases was circulating and was going to anchor a piece. Checked against the filings, it did not reconcile. The four companies disclose the item directly, in those words, and at 30 June 2026 they came to $830.5bn between them, or $898.5bn including leases signed in July. The likely origin of the smaller figure is a single company's commitments table, where one column for obligations due beyond five years, covering all categories rather than leases alone, reads $386,806m. The circulating number was dropped and the disclosed figures were used.
A period that was wrong. A draft described a company's remaining performance obligations rising from $138bn to $638bn "in one quarter." The filing shows that change across the fiscal year to 31 May 2026. The draft had compared it to another company's quarterly step, so the error made two different rates look like one phenomenon. It was caught in review and the piece was rebuilt at the correct frequency.
Cask Advisory publishes general information and research about markets, companies and public disclosure. It is not tailored to the circumstances, objectives or holdings of any person, and it should not be relied on as a recommendation to buy, sell or hold any security.
Cask Advisory is not registered as an investment adviser or dealer in any jurisdiction, and nothing published here should be taken as advice given in the course of such a business. Readers making investment decisions should take their own advice from someone registered to give it.