Method

Every essay here tries to explain a product or technology from the moment the idea was conceived to what it is today, or to why it was abandoned, through the decisions that shaped it, and to make the philosophy and psychology inside those decisions understandable. This page sets out how that is done, so that readers can judge the work by its own standards.

How an essay is researched

Primary sources first. Original papers, company filings and annual reports, court judgments, regulators’ decisions, official announcements and the published accounts of the people involved come before secondary histories and journalism, which are used to cross-check and to fill gaps.
Facts are checked, not remembered. Dates, figures and quotations are verified against a source before they are used. Where sources disagree, the essay uses the better-supported figure or leaves it out, and says so.
Quotations are short, exact and attributed, with a footnote to where they can be read. Everything else is paraphrased.
Every essay ends with its evidence: footnotes linking to the sources used, a selected bibliography, and a research-limitations note listing what rests on a single source, where accounts conflict, which claims come from participants themselves, and which cases or events were still unresolved when it was written.
Nothing is invented. No sources, quotations, statistics, motives or personal experiences are made up. Where something is not known, the essay says so.

Fact, claim, interpretation and uncertainty

The essays try to keep four kinds of statement distinct. A fact is something documented and checkable. A participant’s claim is what someone involved said about their own reasons or actions, reported as their account rather than as established truth. An interpretation is the essay’s own reading, including every connection drawn to a philosopher, economist or psychologist; these connections are offered to make ideas clearer and are labelled as interpretation, since the people who built the products rarely set out to answer those thinkers. Uncertainty is stated where the evidence runs out.

How decisions are judged

Each essay, memo and road not taken includes the decisions that shaped its subject, and each decision is judged twice. The call judges what happened. The process grade judges how good the decision was given what could reasonably be known at the time, apart from luck. A decision can be well made and turn out badly, or badly made and turn out well; collapsing the two, judging choices only by their results, is the most common error in learning from experience, and the separation is the point of the exercise. The Decision Atlas shows both grades for every decision on the site.

The call: what happened

Paid off: The decision achieved what it was meant to, and its benefits clearly outweighed its costs.
Mixed: It achieved some of its aims and failed at others, or its benefits came with serious costs.
Backfired: It produced the opposite of what was intended, or its costs clearly outweighed its benefits.
Too early: Not enough time has passed, or the evidence is not yet in, to judge the outcome.

The process: how well it was decided

Strong: Well reasoned on the information available: real alternatives considered, risks understood, the choice coherent with the situation.
Sound: Reasonable given what could be known, with gaps that were understandable at the time.
Weak: Significant, foreseeable risks or alternatives were missed or discounted, by the standards of the time.
Poor: Contrary to evidence or warnings that were available and should have been heeded.

Three rules apply to every judgement. Decisions are judged in the context of what was knowable then, not with hindsight. Decisions are attributed to the company, team or institution that made them, not presented as the failings of individuals. And no wrongdoing is alleged beyond what courts, regulators or the organisations themselves have documented; where a case is unresolved, the essay says so.

Patterns and lessons

Each judged decision ends with a lesson and is linked to one or more patterns, recurring lessons that appear across different products and eras. A new pattern is added only when the same lesson genuinely recurs in more than one place, so that the patterns describe evidence rather than slogans.

The philosophy and psychology

Each essay has a long section that examines the ideas beneath its subject one at a time. For each idea it explains the idea in plain terms, introduces the tradition of thought it belongs to with accurate attribution, shows where it appears in the history, gives the strongest objection to it, and says what it means for people deciding today. Ideas that recur across essays are explained once on the Ideas pages. The aim is understanding, not name-dropping: a thinker is cited only when their argument is explained.

Cases, memos, papers and books

Decide cases present real decisions without naming the company or the outcome, so that readers judge the situation as it looked at the time. The reveal is hidden until the reader has decided.
Memos and founding papers are original commentary. They quote briefly, attribute every quotation, and link to an official or lawful copy of the original, never reproducing it.
The library arranges books on fifteen shelves, from strategy and decision making to philosophy and the good life, rates each out of five, and links each book to the essays, ideas and reading paths that draw on it. Reading paths order eight books around a single question, with a reason for each step. The best-rated books have notes. Each note takes the form that suits the book (a case to follow, a practice to use, a story of decisions, a way of seeing), paraphrases rather than quotes, uses short quotations only from public-domain editions or lawful copies, and lists what its facts were checked against. Book pages describe books; they do not criticise their authors, though they do report where later evidence has changed the picture.

The Decision Record

Every judged decision on the site is also coded by its kind, such as a business-model change, an acquisition or a technology bet, and by the position of whoever made it. The Decision Record publishes the codes, the base rates they suggest and the full dataset, together with its limits: it is a record of decisions the site has written about, not a random sample, and its process grades were assigned with knowledge of the outcomes.

Expert review

Essays are written to be checked. Where a specialist in the field has read an essay and agreed to be named, the essay says so under its byline, with the reviewer’s role and what they reviewed. Reviewers check facts, sources and fairness; the interpretations and judgements remain the author’s, and so does responsibility for any error. No essay names a reviewer who has not read it. Specialists willing to review an essay in their field are welcome to write, using the address on the corrections page.

Citing and printing

Every essay, idea, memo, founding paper, road not taken, case and strategy question ends with a ready-made citation that includes its dates. Essays can be printed or saved as a PDF from the button under the byline; the printed edition opens every decision and prints the address of every external link, so it stands on its own.

Revisions

Essays are revised as new evidence appears or events move on. Each essay shows when it was published and, if it has changed since, when it was last updated. One product has one essay: revising a subject means improving its essay, not adding a new one.

Errors of fact are a different matter from revisions. When a date, figure, quotation, attribution or claim turns out to be wrong, it is fixed and the fix is recorded, with its date and what changed, in the corrections log. Corrections are not made silently.

What this site does not do

It does not report news, review products, rank companies or pass verdicts on people. It is not investment, legal or professional advice. Its judgements of decisions are arguments, made openly with their evidence, and readers are invited to disagree with them. Reuse of the writing is described on the copyright and reuse page.