Methodology

This page sets out how we handle evidence: what we treat as established, what we treat as inference, and where those distinctions appear in a report. It applies to reports issued from 15 September 2026.

1. Evidence classes

Every factual claim in a report carries one of four classes.

[D] DISCLOSED. The company or protocol has stated this itself — in filings, documentation, official communications, or materials provided to us.

[E] EXTERNAL. A third party has reported or estimated it — press, research firms, data providers, on-chain analytics.

[V] VERIFIED. We have confirmed it against an independent source that does not depend on the first.

[H] HYPOTHESIS. Our reading, inference, or model. It is ours, and it is not a fact.

Two of these are routinely conflated. That a company has published an ARR figure is [D]. That the figure is accurate is [V], and it is a separate question with a separate answer. We keep them apart.

In screening work, [V] is the least common of the four. That is not a defect of the method; it is the output. A claim that cannot be independently confirmed at this stage is a claim that has to be confirmed later, and identifying which claims those are is most of what a screen is for.

2. Where the tags are carried

Evidence classes are not confined to footnotes.

A claim tagged [H] in the analysis stays tagged [H] where it appears in the summary, and again where it appears in the Diligence Agenda. If a conclusion rests on inference, that is visible at the point the conclusion is read rather than several pages away.

Where a section’s conclusion draws on a mix of classes, we say which parts of the evidence carry it.

3. Valuation language

Terms that assert a market judgment — undervalued, fair value, valuation floor, arbitrage, re-rating — do not appear in our reports as statements of fact.

Where the underlying idea is worth raising, it is written as a hypothesis with its assumptions and its downside sensitivity attached, and tagged [H].

This is narrower than it may sound. We are not withholding a view. We are keeping a view labelled as one.

4. Comparison discipline

When a report compares companies or assets:

• One comparison date applies to every name, and it is stated on the table.
• Revenue is defined explicitly — ARR, GAAP revenue, or run-rate — and named as such rather than left as “revenue”.
• Currency is stated, and the date of the source appears alongside the figure.
• Each figure is defined once in the report. Every other mention refers back to that definition rather than restating it.

The last rule exists to prevent the same number appearing in two sections with two values. Restating a figure invites drift; referring back to it does not.

5. What each pillar closes with

Each of the five pillars ends with three lines:

1. What the evidence supports.
2. The largest question the evidence leaves open.
3. The finding that would change the investment decision.

The third line is the test of the section. If we cannot name a finding that would change the decision, we have described a company rather than analysed one.

6. As-of dating

Every report carries a publication date and an as-of date for the data beneath it.

Past thirty days from issue, a report is either reissued against a current as-of date or circulated with its original date displayed. The sample reports on this site are published as written, each with its date shown.

7. Scope of our work

We do not conduct primary research. Customer reference calls, expert interviews and financial reconciliation sit downstream of this work. What we deliver is the agenda those steps should follow.

This is a boundary rather than a gap. A screen earns its place by being finished before that work is commissioned, and by narrowing what it has to cover. A report that has already carried out primary verification is a different product on a different timetable.

This standard applies to reports issued from 15 September 2026.