·General Thinking & Meta-Models
Section 1
The Core Idea
Groups spend the most time on the decisions that matter least.
C. Northcote Parkinson identified the pattern in 1957 in
Parkinson's Law, and Other Studies in Administration. His example was a finance committee reviewing three agenda items: a £10 million nuclear reactor, a £350 bicycle shed for the office staff, and a £21 annual coffee budget. The reactor was approved in two and a half minutes — none of the committee members understood nuclear engineering well enough to contribute, so they deferred to the experts and moved on. The bicycle shed consumed forty-five minutes. Everyone understood bicycle sheds. Everyone had an opinion about the roof material, the colour, the placement. The coffee budget consumed over an hour because every committee member also had strong feelings about suppliers, delivery schedules, and whether biscuits should be included. The pattern was inversely proportional: the less consequential the decision, the more time the group spent on it.
The phenomenon acquired a second name in 1999 when Poul-Henning Kamp, a Danish software developer, posted a message to the FreeBSD mailing list explaining why a minor proposal about the sleep() function had generated an outsized thread. He called it "bikeshedding" — a direct reference to Parkinson's bicycle shed — and the term entered the permanent lexicon of software engineering, where it became the default shorthand for disproportionate debate over low-stakes design choices. Any engineer who has watched a team spend thirty minutes debating variable naming conventions in the same meeting that approved a new microservices architecture in five minutes has witnessed the law in action.
The mechanism is straightforward. Complex, high-stakes decisions — a merger, a product strategy, a database architecture — require domain expertise that most participants lack. Engaging meaningfully feels risky: you might ask a naive question, expose ignorance, or challenge someone with deeper knowledge. The rational response is silence or deference. Simple, low-stakes decisions — the colour of a button, the location of an off-site, the wording of a press release — are accessible to everyone. Contributing feels safe, even gratifying. Competence is easy to perform. The result: the difficulty of contribution determines the volume of discussion, and difficulty is inversely correlated with importance.
The inversion is so reliable that it functions as a diagnostic tool. If you want to measure the health of an organisation's decision-making process, observe the ratio of discussion time to decision magnitude across a dozen meetings. In dysfunctional organisations, the correlation is negative — the most consequential items receive the least debate. In well-structured organisations, the correlation is positive or at least neutral. The ratio is more predictive of organisational effectiveness than headcount, budget, or reported satisfaction scores, because it directly measures whether the institution's scarcest resource — collective analytical attention — is being directed at the problems that will determine its future.
This is not laziness. It is not incompetence. It is a structural feature of group deliberation that appears in corporate boardrooms, open-source projects, government committees, academic departments, and neighbourhood associations with equal reliability. The U.S. Congress routinely passes trillion-dollar defence authorisations with less floor debate than naming resolutions for post offices. Apple's early board meetings, before Jobs's return, reportedly spent more time on conference logistics than on the product roadmap that would determine whether the company survived. Academic hiring committees spend weeks debating the format of campus visit itineraries and minutes on the substantive question of which candidate will best serve the department for the next thirty years. The pattern doesn't require incompetent participants. It requires only that some decisions are complex enough to be intimidating and others are simple enough to invite universal participation.
The law scales with the size of the organisation. A five-person startup is partly immunised because everyone understands the full problem space — the competence gap that drives the inversion doesn't exist when everyone is a generalist. A five-hundred-person company is vulnerable because the gap between specialists and generalists widens at every level. A fifty-thousand-person company is structurally guaranteed to produce the effect, because the number of people qualified to contribute to any consequential decision shrinks as a proportion of total headcount while the number qualified to opine on trivial decisions remains near 100%. This is why bikeshedding is a disease of scale. The larger the committee, the wider the competence gap, the stronger the inversion.
Jeff Bezos's "two-pizza rule" — no team should be larger than two pizzas can feed — is partly a response to this scaling problem. Smaller teams reduce the competence gap because every member is closer to the core problem. They also reduce the social dynamics that produce bikeshedding: in a room of six, there's less need to perform engagement and less safety in hiding behind silence on the hard topics. The organisational design of team size is, in this light, a structural intervention against the law of triviality — one that operates at such a basic level that most people don't recognise it as such.
The cost is not the time wasted on the bicycle shed. The cost is the scrutiny that never reached the reactor. Every hour of institutional attention is a zero-sum resource. When a committee exhausts its energy on trivia, the hard decisions receive less challenge, less debate, less adversarial testing — and consequently more errors. Parkinson's law of triviality is, at root, a model of institutional attention misallocation. The organisation isn't failing to decide. It is deciding where to focus, and it is consistently choosing to focus on the thing that feels most comfortable to discuss rather than the thing that most needs discussing.
