Methodology

This page fixes the definitions. The point of writing them down before the archive grows is that a figure published in 2026 should still mean exactly the same thing when it is read in 2031.

How we research

The starting question is not whether a business is good. It is what the current price appears to assume, and whether our reading of the evidence differs from that. A company can be excellent and fully priced, which is not interesting here, or unremarkable and badly misunderstood, which is.

Work begins with the company's own disclosure and stays there as long as possible. Annual reports, regulatory filings, investor presentations, earnings calls and official announcements come first. Third-party material supplements primary sources; it does not replace them. Every article lists its sources with the primary ones separated out, so a reader can see how much of the analysis rests on the company's own words.

We do not forecast markets, interest rates or sentiment, and we do not publish price targets with dates attached. What we publish is a view of what a business is worth under stated assumptions, and the arithmetic that follows from it.

Valuation

Fair value is derived from the economics of the business: revenue, margin, earnings and the multiple those earnings deserve. Each article states its valuation basis explicitly in the disclosure at the foot of the page.

The distinction that governs this entire site is between what isauthored and what is derived.

No return figure is ever stored in an article. Every one is computed from the assumptions and the cash flows, by a single shared module, so the published numbers cannot drift away from the analysis that produced them.

Bear, Base and Bull

Every equity thesis publishes three cases. They are not confidence intervals and they are not probability-weighted. They are three coherent sets of assumptions, each internally consistent, describing what the business is worth if things go badly, ordinarily, or well.

Each horizon carries its own three cases. A three-year view and a five-year view rest on different earnings and different assumptions, so they are authored separately rather than one being extrapolated from the other. Where the analysis only supports a three-year view, only a three-year view is published.

The base case takes a subtle emphasis in the tables. That marks it as the central assumption, not as a prediction.

Expected returns on equity

For a normal equity thesis we publish Expected 3Y CAGRand Expected 5Y CAGR, one for each horizon the article supports.

Each is the annual rate at which the publication price would have to grow to reach that horizon's fair value, over that horizon. It assumes the gap closes on schedule. That assumption is doing real work and is worth stating plainly: if the market takes six years to agree with a five-year view, the realised return is lower than the published expectation, even though the valuation was right.

Expected returns on special situations

A liquidation has no terminal fair value to converge toward. It has a schedule of payments. So special situations publish three figures instead:

• Expected Total Return. All expected proceeds against the price paid, expressed as a single percentage.
• Expected Annualized Return. The same proceeds expressed as an annual rate, taking their timing into account.
• Estimated Duration. How long the base case expects the situation to take.

All three are derived from an authored schedule of expected payments, where each case carries its own amounts and its own timing. A slower outcome usually pays less as well as later, and modelling those separately would understate the downside.

Realised returns on closed theses

When a thesis closes, three figures are published together and are never shown apart:

• Total Return. Everything received, including distributions, against the price paid. This is the primary measure of whether the thesis worked.
• Holding Period. How long it actually took.
• Annualized Return. The same result expressed as an annual rate.

The price used is the price recorded when the research was published, and it is never revised. Neither is anything else in a published article: when facts change, a dated update is added and the original analysis stays readable as it was written.

Dividends and distributions

Cash received during the holding period counts toward the return. Dividends, capital returns and liquidation instalments are all treated the same way:

• per share, in the currency of the thesis;
• gross of withholding tax, because the rate depends on the holder;
• not reinvested, because reinvestment assumptions imply a precision the underlying analysis does not have;
• recorded with the date received, which matters for the calculation below.

Currencies are never converted. Each thesis is analysed and reported in its own currency, and every figure is displayed with its currency code attached. For the same reason there is no aggregate return across the archive: adding a Norwegian krone result to a euro result would produce a number that means nothing.

How Annualized Return is calculated

One name, because the reader should not have to learn a vocabulary to read a research archive. Two calculations underneath it, because one would be wrong.

When there is one entry and one exit, the annual rate is the rate at which the amount paid compounds into the amount received over the actual number of days held.

When there are several dated cash flows, for example a liquidation paying in instalments, or an equity holding that paid dividends along the way, the calculation uses those dated cash flows directly. It does not pretend there was a single entry and a single exit. Money received early is worth more than the same money received late, and a calculation that ignored the dates would understate a thesis that returned capital quickly.

The two methods agree exactly when there are only two cash flows. That is deliberate and it is tested: it means the archive never contains a step change caused by the calculation switching methods rather than by a result actually differing.

Day counts are actual days over 365.

Short holding periods

Annualizing a short holding period produces a large number.A position held for three weeks that returns 4% annualizes to well over 100%. That figure is arithmetically correct and it is not a claim that anything similar can be repeated fifty times a year.

This is why Total Return, Holding Period and Annualized Return are always published together, and why the annualized figure is never shown on its own anywhere on this site. Where a thesis was held for less than ninety days, the annualized figure carries a marker pointing here.

Short holding periods are still annualized rather than suppressed, because the time a result took is part of the result. Six months at 10% is genuinely different from three years at 10%, and hiding that would flatter the slower outcome.

Research status

A thesis is in one of three states.

• Active. The position is open and the thesis stands as published.
• Thesis changed. Something material has changed and the original argument no longer holds in the form it was written. The article says what changed and when.
• Closed. The thesis has ended and the realised result is published.

Research is never silently rewritten. Updates are dated and added; the original text is not edited to match what happened.

Update policy

This is the publication's own editorial policy. It is not a description of anyone's regulatory requirements, and it is not advice about what any publication is obliged to do.

An article is updated when a fact it relied on changes, not when its price moves. Results, guidance, corporate actions, transaction timetables, regulatory decisions and anything that would have altered the original analysis are grounds for an update. Share price movement, on its own, never is.

• Updates are additive and dated. They are appended to the article with the date they were written. The original analysis is left exactly as published, including the parts that turned out to be wrong.
• Corrections are marked as corrections. Where a published figure or statement was simply incorrect, the correction says so plainly rather than quietly replacing the text.
• The reference price is never revised. It is one observation on one date and it stays that observation forever.
• An update never converts a thesis into a different thesis.Where the argument itself no longer holds, the status changes to "thesis changed" or the thesis is closed, and the reason is stated.
• Silence is not an update. If nothing material has changed, no update is published, and the absence of one is not a restatement of the original view.

There is no commitment to cover any company continuously, and no commitment to publish a further piece on any issuer. Research is published when there is something to say.

Outcome labels

Every closed thesis carries one of three labels. These are judgments, not thresholds calculated from the return, and they are about whether the reasoning worked rather than whether the number was positive.

• Successful. The thesis worked broadly for the reasons given.
• Partial. The result was acceptable but the reasoning was substantially wrong, or the reasoning was right and the result was not.
• Unsuccessful. The thesis did not work.

The archive

Closed research stays published permanently, at the same URL, including research that lost money. Unsuccessful theses are listed in the same archive as successful ones, with the same figures shown in the same way and no additional friction to find them.

This is not modesty. A research record that keeps only its successes tells a reader nothing about the researcher, and the most useful documents in any archive are usually the ones that explain what went wrong.

Every thesis, open and closed, is listed on thetrack record. That page publishes no combined return, no average and no success rate. The theses are in different currencies, were never sized against one another and do not describe a portfolio anyone held, so a single summary figure would have nothing behind it.

What every article discloses

Each report carries a disclosure block at the foot of the page. It states who wrote the research and in what capacity, when the analysis was completed and when it was published, the reference price and when it was observed, the valuation basis, and the version of this methodology it was written under.

For research about a real company it also states three things directly: whether the author holds a position and in which direction, whether the issuer has paid or provided anything of value for the research, and whether the issuer saw the research before publication and, if so, whether anything changed as a result. Any other material conflict is listed, and any earlier research on the same issuer in the preceding twelve months is linked.

Position direction is disclosed. Position size is not recorded anywhere on this site, in any form: not the number of shares, not a portfolio weighting, not a cost basis. The author's dealing rules are published in the repository as TRADING_POLICY.md.

Sample research is the one exception, and it is marked as such on the page. A fictional company cannot be held, cannot pay for research and cannot review it, so those three fields are absent rather than invented.

Dates and time zones

Every timestamp on this site is published in the time zone it was recorded in, with the offset shown. "5 Sept 2026, 18:30 UTC+02:00" means half past six in the evening in Madrid, not in the reader's own zone and not in UTC.

This matters more than it looks. A price observed shortly after midnight, republished in UTC, would be dated to the previous day, and the record of when a piece of analysis was finished relative to when it was published would drift by an hour twice a year.