Business writing is dominated by lessons from successful companies, and the methodology producing those lessons has a structural problem that is worth naming clearly.
Survivorship
Studying successful companies and identifying common features cannot establish that those features caused success.
Because the same features may have been present in companies that failed, and those companies are not in the sample.
Which means every quality identified — boldness, persistence, focus, adaptability — describes both survivors and casualties, and the analysis cannot distinguish them.
Testing this properly requires comparing successful companies against matched unsuccessful ones, which is rarely done because the failures are harder to study.
The regression problem
Companies identified as exemplary at a point in time frequently perform worse afterwards.
Which has happened to the subjects of several famous business books, sometimes dramatically.
Part of this is regression to the mean — extreme performance in one period is partly luck, which does not repeat.
Which suggests that what was identified as excellence was partly the luck component, and the analysis could not separate them.
Narrative fallacy
Successful outcomes are explained retrospectively with coherent stories.
Which are constructed after the fact from what happened, selecting the decisions that worked and framing them as strategy.
Founders' own accounts are subject to this, and the accounts change over time as the narrative settles.
Contemporaneous evidence, where it exists, frequently shows considerably more uncertainty than the retrospective account conveys.
Context
Practices that work depend on conditions rarely stated.
Market structure, capital availability, regulatory environment, labour market, technological moment.
Which means a practice that worked at a particular company in a particular decade may not work elsewhere, and the advice generally omits the conditions.
Culture practices from companies with enormous margins and abundant capital are the clearest example, since the margin funded the practice rather than resulting from it.
Scale
Advice derived from large organisations frequently addresses problems small ones do not have.
Coordination, standardisation and process exist to solve problems of scale, and applying them to a small team adds cost without benefit.
The reverse is also true — practices that work for a small team frequently do not survive growth.
What is more reliable
Failure analysis, which identifies things that reliably go wrong.
Because the causes of failure are more consistent than the causes of success, and avoiding known failure modes is more tractable than replicating success.
Controlled research where it exists — hiring, pricing, incentives, team effectiveness — which has been studied experimentally in ways that case studies cannot match.
And base rates, which describe what typically happens rather than what happened once.
How to read a case study
Ask what else was true of the company that is not true of yours.
Ask whether companies that did the same thing and failed exist, and whether anyone looked.
Ask whether the account is contemporaneous or retrospective.
And treat it as a source of hypotheses rather than of conclusions, which is what a single case can legitimately provide.
Consultants and frameworks
Frameworks organise thinking and are frequently presented as producing answers.
Which they do not — a framework structures analysis and the judgement remains.
Their value is in ensuring nothing important is omitted, which is genuine, and in providing a shared vocabulary, which is also genuine.
Treating framework output as a decision is the error, and it is encouraged by how they are marketed.
Benchmarks
Industry benchmark figures are widely used and frequently derived from small self-selected samples.
Which means the median in a benchmark report may not describe the industry, and comparing against it can drive changes that are not warranted.
Checking the methodology and sample of any benchmark before acting on it takes minutes and is rarely done.
What to do instead
Run small experiments in your own business, which produces evidence about your context specifically.
Which is slower than adopting a practice and produces knowledge that actually applies.
Personal experience
Advice from someone who ran one business describes one business.
Which is valuable as a source of specific observations and unreliable as generalisation, and the confidence with which it is delivered is unrelated to how far it transfers.
The most useful version is descriptive — here is what happened and what we did — rather than prescriptive.
What to take from it
Mechanisms rather than conclusions.
Understanding why something worked allows judging whether the conditions apply, which copying the action does not.
Which is the difference between learning from a case and imitating it.
Where advice is genuinely reliable
Areas that have been studied experimentally rather than through case study — selection, incentives, pricing, team processes.
Which is a narrower set than business writing covers and is considerably more trustworthy within it.
Preferring evidence from controlled research over anecdote is a straightforward filter and it eliminates most of the genre.
Which is a modest and defensible position, and it is considerably less satisfying than a set of principles to follow.