all models are wrong
expressionops cultureprogramming
George Box's statistics proverb: every model simplifies falsely - the question is whether it is useful anyway.
Monitoring dashboards, capacity forecasts, risk matrices, network diagrams: all wrong, some useful. The operational discipline is remembering which simplifications each model made, because that is exactly where it will mislead you.
All models are wrong, but some are useful. George Box's line is about statistics and transfers directly to every abstraction in computing, because every model deliberately discards detail and the discarded detail is what eventually matters.
The OSI layers are a model that no real network implements exactly, and they remain the most useful vocabulary the field has. A threat model is a simplification of everyone who might attack you. A capacity model is wrong about the workload it will actually see. A network diagram omits the things that will cause the outage. Each is valuable, and each is valuable precisely because it left things out.
The discipline is knowing your model's boundary conditions, since a model applied outside where it holds produces confident wrong answers rather than obvious failures. That is why the useful question during an incident is not what does the diagram say but what is the diagram missing, and why an engineer who can articulate where their mental model breaks down is considerably more reliable than one who cannot, regardless of who has the better model.
Also known as: all models are wrong, but some are useful