“Any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes.”
Charles Goodhart, 1975 [1]
I have seen this scene more than once. A dashboard improves. The metric turns green. The meeting ends with a sense of progress.
Then someone closer to the work explains what actually happened: the team closed tickets faster by transferring difficult cases, the deadline was met by pushing risk into the next phase, or the new process removed a step that looked pointless only because its original purpose had been forgotten.
This is why I am cautious when a workplace idea fits perfectly on a sticky note.
Parkinson’s Law. Goodhart’s Law. The Peter Principle. Brooks’s Law. They are useful precisely because they compress a pattern into a memorable phrase. But compression has a cost. A mental model can help me see a mechanism I would otherwise miss. Treated as a universal law, it can become a substitute for judgment.
The eleven models in this article are not legal norms. They are not rules that dictate what an organisation must do. They are different kinds of intellectual tools: some began as empirical research, some as observations about organisations, some as philosophical heuristics, and some as aphorisms that became more famous than their original context.
I use them through four questions:
What mechanism does this model help me notice? Where did it originally apply? What evidence would make it inapplicable? And who is harmed if I treat it as fate?
That shift matters. It turns productivity slogans into a more demanding practice of management.
1. A diagnostic grammar, not a box of hacks
The eleven ideas answer different questions. Some help explain why projects drift. Others reveal how incentives distort measurement. Others ask us to slow down before drawing conclusions or tearing down an old process.
| Model | The useful question | The mistake to avoid |
|---|---|---|
| Parkinson’s Law | Is slack being filled by avoidable scope, meetings or revisions? | Equating a shorter deadline with equal-quality work. |
| Hofstadter’s Law | What does comparable past work tell us about the estimate? | Treating overruns as inevitable. |
| Brooks’s Law | Is the remaining work genuinely parallel, or will new people add coordination cost? | Using it as a ban on hiring. |
| Pareto Principle | Is outcome concentration visible in the actual data? | Assuming every system follows an 80/20 ratio. |
| Goodhart’s Law | What behaviour will this target create? | Abandoning all metrics because they can be gamed. |
| Peter Principle | Does excellence in the current role predict success in the next one? | Labeling people as permanently incompetent. |
| Hick-Hyman Law | Can the next decision be made clearer without hiding necessary options? | Optimising speed at the expense of safety or accuracy. |
| Hanlon’s Razor | Is there a non-malicious explanation that the evidence supports? | Ignoring repeated harm, fraud or discrimination. |
| Dunning-Kruger effect | How well calibrated is confidence against performance? | Using a research label to humiliate someone. |
| Occam’s Razor | Does a simpler explanation fit the evidence just as well? | Assuming the simplest story must be true. |
| Chesterton’s Fence | What function does this rule or process currently serve? | Defending the status quo indefinitely. |
I find this table more helpful than a list of clever quotations because it makes each model actionable. The value is not in reciting the phrase. The value is in using the right question at the right point in a decision.
2. Time is not one thing, and neither is capacity
The fastest way to make a workplace less intelligent is to confuse avoidable sprawl with real uncertainty.
Parkinson’s Law is often reduced to a brutal productivity prescription: shorten every deadline and the task will shrink to fit it. C. Northcote Parkinson’s original 1955 Economist essay was more specific and more satirical. It observed the tendency of bureaucracy to generate work and staff even when the underlying workload declines.[2] It was not a licence to turn every project into a fire drill.
The practical insight still holds. When a task has no clear boundary, slack may become optional analysis, more meetings, an extra round of revisions or a new stakeholder who arrives after the work should have closed. I have seen strategic work become less rigorous, not more rigorous, because nobody defined what “done” meant.
A better use of Parkinson’s Law is a scope diagnostic. For a decision memo, I can define the decision, the evidence threshold, the owner, the page limit and the review date. If someone needs to expand it, that expansion should be visible and approved. The point is not to make the team hurry. It is to make the trade-off between scope, quality and time explicit.
Hofstadter’s Law corrects the opposite mistake. Douglas Hofstadter’s recursive aphorism, first published in Gödel, Escher, Bach in 1979, says that work takes longer than expected even when we account for the law itself.[3] It is memorable because complex work contains complexity that an early plan cannot see: dependencies, rework, feedback loops, integration and unfamiliar decisions.
But it is an aphorism, not a forecasting method. If I respond by adding an arbitrary 30% buffer to every plan, I have learned nothing. The useful companion is the planning-fallacy research of Roger Buehler, Dale Griffin and Michael Ross. They found that people can forecast their own tasks too optimistically, while an explicit connection to relevant past experience can reduce the bias in one of their studies.[4]
So the workplace ritual I prefer is simple. Before committing to a date, I ask: what did comparable work actually take, what dependencies did it include, and what quality threshold are we protecting? A range with named risks is often more honest than a single confident date.
Then comes Brooks’s Law, perhaps the most abused sentence in project management: adding people to a late software project can make it later. Frederick Brooks used the phrase in The Mythical Man-Month, based on the real problem that new contributors require onboarding while existing people must divide work, communicate, integrate and test.[5]
Brooks himself called the maxim an intentional oversimplification. That qualification matters. Adding people early can help. Adding specialists to an independent workstream can help. Adding a fully contextualised person can help. What does not help is throwing unfamiliar people into a late, tightly coupled project and pretending that human attention is infinitely divisible.
These three models are not contradictory. They describe three different failures of planning:
- Parkinson warns me about work that grows because it has no boundary.
- Hofstadter warns me about work that exceeds the boundary because reality is more complex than the plan.
- Brooks warns me that extra headcount may increase the complexity of the recovery itself.
The mature response is neither permanent urgency nor permanent delay. It is bounded scope, outside-view estimation and a real map of dependencies before changing the date or the team.
3. Every organisation begins to distort what it can see
If time is the first management illusion, measurement is the second.
The Pareto Principle is usually presented as a magical ratio: 80% of results come from 20% of effort. Vilfredo Pareto’s original work was about the concentration of income and wealth in the late nineteenth century, not an instruction manual for office productivity.[6] The modern workplace lesson is narrower: many systems have concentration, but the shape of that concentration must be measured rather than assumed.
Suppose a customer-support team receives 10,000 tickets. A Pareto analysis can rank issue categories and show whether a handful of defects generates most demand. That can be valuable. It may reveal that one broken onboarding step creates a large share of customer confusion.
But frequency is not the same as importance. A rare security incident, safety risk or compliance breach may matter more than the hundredth routine request. The 80/20 principle is useful when it tells me where to investigate. It becomes dangerous when it tells me what I am allowed to ignore.
Goodhart’s Law then asks what happens after I make the metric into a target. In its original monetary-policy context, Goodhart warned that an observed statistical regularity could collapse when pressure was put on it for control.[1] The popular wording, “when a measure becomes a target, it ceases to be a good measure”, is a later and stronger formulation, often associated with Marilyn Strathern.[7]
The distinction is important. Goodhart did not prove that all targets become useless. He showed why proxies become fragile once people have an incentive to optimise the proxy rather than the underlying outcome.
Give a support team a hard target for average handle time and the dashboard may improve. Agents may also transfer difficult cases, end calls early or discourage follow-up. The number is not necessarily false. It is incomplete, and the system has adapted around it.
That is why I prefer metric constellations to single metrics. Handle time can sit alongside first-contact resolution, customer-quality review, repeat-contact rates and employee feedback. None is perfect. Together they make gaming more difficult and reality more visible.
The Peter Principle adds a human version of the same mistake. Laurence J. Peter’s 1969 idea was that people in a hierarchy tend to rise to their level of incompetence. In more contemporary terms, an organisation may promote someone for performing brilliantly in their current job even though the next job requires a different set of capabilities.[8]
A top salesperson does not automatically become a strong sales manager. A great designer does not automatically become a good design director. A gifted engineer may not enjoy or excel at coaching, delegation, conflict resolution or building a team.
The research is more nuanced than the slogan. A large study of sales organisations found evidence consistent with firms promoting high current performers even when other observable characteristics better predicted later managerial performance.[8] Yet apparent performance decline after promotion can also reflect regression to the mean rather than a pure failure of promotion logic.[9]
The practical conclusion is not that people should stop being promoted. It is that promotion needs a separate theory of the destination role. If the next role requires coaching, judgment, commercial context or people leadership, those capabilities should be assessed directly. Current output is evidence. It is not the whole case.
The organisation becomes less intelligent when it mistakes a proxy for the thing itself: ticket volume for customer health, individual excellence for leadership, a green dashboard for genuine progress.
4. The cost of a simple explanation
The next set of models concerns how people decide, interpret and judge each other.
Hick’s Law, more accurately the Hick-Hyman law, comes from controlled research on choice reaction time. As the information uncertainty in a set of possible responses increases, the time needed to choose a response tends to increase in an approximately logarithmic relationship.[10] [11]
The workplace version is useful but limited. In an incident-response flow, a person facing twenty unclear actions may freeze. A well-designed interface can first offer three clear next steps, then reveal detail when it is needed. That reduces immediate uncertainty.
It does not follow that fewer options always produce a better decision. In medical, legal, safety or people matters, an “efficient” menu may hide the exception that someone needs. Hick’s Law helps me design the first decision. It does not give me permission to eliminate complexity from the world.
Hanlon’s Razor offers a different kind of simplification: do not attribute to malice what can be adequately explained by error or incompetence. The named formulation was first printed in Arthur Bloch’s 1980 Murphy’s Law, Book Two, though similar ideas are much older.[12]
At work, this can be an antidote to paranoid storytelling. When a colleague omits a stakeholder, delivers a flawed report or misses a handoff, I should first test ordinary explanations. Was the brief unclear? Was the template outdated? Did the person lack a necessary piece of information? Did the process make the error likely?
That makes collaboration less theatrical and diagnosis more precise. But it is not a pardon. In harassment, discrimination, retaliation, fraud, safety or security matters, intent may be relevant and evidence must be preserved. Repeated “mistakes”, concealment and clear incentives can change the explanation.
The Dunning-Kruger effect is perhaps the most casually weaponised model in management. The original 1999 studies by Justin Kruger and David Dunning examined how people assessed their performance in specific tasks such as humour, grammar and logic. They proposed that people who perform poorly may also lack some of the knowledge needed to recognise their errors.[13]
This is not a diagnosis of “stupid people”. It is not a personality type. It is not a managerial licence to dismiss junior colleagues.
The research itself has been debated. Critics argue that some apparent patterns may arise from statistical features of noisy self-assessment and the way groups are formed, while later work has defended aspects of the original interpretation.[14] For a leader, the debate leads to a better practice than name-calling: ask people to state their expected result and confidence before an assessment, compare it with observable outcomes, and give specific feedback.
That turns calibration into a learnable skill.
Occam’s Razor completes this chapter. The principle is usually translated as “the simplest explanation is probably the right one”. That is too crude. Ockham’s actual philosophical legacy is a preference against adding unnecessary assumptions when competing explanations account for the evidence equally well.[15] [16]
When a project date slips, I should test the explanation with the fewest unsupported assumptions. Perhaps a known dependency was late. Perhaps the original estimate omitted integration work. Perhaps the team never agreed on the acceptance criteria.
But simplicity is not a licence to ignore inconvenient evidence. Complex systems can have complex causes. The purpose of Occam’s Razor is to keep me from inventing a story too early, not to force reality into the story I find most elegant.
5. Before removing the fence
Chesterton’s Fence may be the model I return to most often in transformation work.
The popular phrase is simple: do not take down a fence until you know why it was built. The phrase is a later shorthand for a longer passage by G. K. Chesterton in his 1929 book The Thing.[17] Its real lesson is not “preserve every old rule”. It is understand the function before deciding whether to remove, redesign, replace or retain it.
That is a crucial distinction for any digital transformation programme.
I have encountered approval steps that looked like pure bureaucracy. Sometimes they were. But sometimes they existed because a previous failure had created a legal, financial, safety or brand risk that had since disappeared from institutional memory. I have also encountered legacy procedures whose original purpose was valid but whose burden had become disproportionate. In both cases, removing the process without investigation is lazy. Keeping it without investigation is equally lazy.
Chesterton and Occam work well together.
Occam asks me not to invent five causes when one evidence-based cause is sufficient. Chesterton asks me not to remove a control merely because I cannot immediately see its purpose. One prevents needless explanation. The other prevents needless destruction.
Together with the other models, they form a responsible intervention loop:
- Observe the system before naming the problem.
- Use an outside view before promising a date.
- Map incentives before turning a metric into a target.
- Test a parsimonious explanation against evidence.
- Understand the function of a rule before removing it.
- Run a reversible change where possible, then watch for second-order effects.
This is less seductive than a productivity hack. It is also how organisations learn without repeatedly rediscovering the same mistakes.
The purpose of workplace mental models is not to make us sound clever in a meeting. It is to make us harder to fool, including by our own preferred explanations.
I will leave myself, and you, with the question that matters most:
When I call a workplace practice irrational, am I seeing the system clearly, or merely demanding that it be simple enough to fit my preferred story?
References
[1]: https://www.rba.gov.au/publications/rdp/1990/9013/conference-volumes.html “Reserve Bank of Australia: Papers in Monetary Economics, Volume 1”
[2]: https://www.economist.com/news/1955/11/19/parkinsons-law “Parkinson’s Law”
[3]: https://www.basicbooks.com/titles/douglas-r-hofstadter/g%C3%B6del-escher-bach/9780465026562/ “Gödel, Escher, Bach: An Eternal Golden Braid”
[4]: https://web.mit.edu/curhan/www/docs/Articles/biases/67_J_Personality_and_Social_Psychology_366,_1994.pdf “The Planning Fallacy: Why People Underestimate Their Task Completion Times”
[5]: https://web.eecs.umich.edu/~weimerw/2018-481/readings/mythical-man-month.pdf “The Mythical Man-Month: Essays on Software Engineering”
[6]: https://archive.org/details/fp-0148-1 “Cours d’économie politique, Volume 1”
[7]: https://doi.org/10.1017/S1062798700002660 “Improving Ratings: Audit in the British University System”
[8]: https://academic.oup.com/qje/article/134/4/2085/5550760 “Promotions and the Peter Principle”
[9]: https://www.gsb.stanford.edu/faculty-research/publications/peter-principle-theory-decline “The Peter Principle: A Theory of Decline”
[10]: https://doi.org/10.1080/17470215208416600 “On the Rate of Gain of Information”
[11]: https://doi.org/10.1037/h0056940 “Stimulus Information as a Determinant of Reaction Time”
[12]: https://archive.org/details/murphyslawbooktw00bloc “Murphy’s Law, Book Two: More Reasons Why Things Go Wrong”
[13]: https://doi.org/10.1037/0022-3514.77.6.1121 “Unskilled and Unaware of It: How Difficulties in Recognizing One’s Own Incompetence Lead to Inflated Self-Assessments”
[14]: https://www.bps.org.uk/psychologist/dunning-kruger-effect-and-its-discontents “The Dunning-Kruger effect and its discontents”
[15]: https://plato.stanford.edu/entries/ockham/ “William of Ockham”
[16]: https://plato.stanford.edu/entries/simplicity/ “Simplicity”
[17]: https://catholiclibrary.org/library/view?docId=/Contemporary-EN/XCT.165.html&chunk.id=00000011 “The Thing: Why I Am a Catholic, The Drift from Domesticity”