I don’t like to design to a detailed specification. In many cases designing to specification leads to mediocre value.
In other words, specs leave money on the table. My problem with specifications is that they are usually expressed as an all-or-nothing value proposition: If the spec calls for a maximum weight of 150g, then a design that comes in at 151g fails and has no benefit, whereas a marginally lighter one at 149g passes with full benefit. I would argue that in most cases this is too much of a simplification. In reality there will be a smooth benefit curve between “too heavy” and “impossibly light” – sometimes steep and sometimes shallow.
The blue step-function in the graph shows the value-parameter curve for a pass/fail specification. In reality it might be a somewhat smoother curve (red), a very weak dependency (yellow), or a curve that shows an optimal value (green). Not shown in the graph, but curves can also be inverted – as in the example of weight above.
It might look like there is a lot of user preference data required to draw up these curves, but in effect no more research is required than if you had specified a pass/fail specification: How else would you know where to put your threshold?
Assuming it is possible to create such a benefit curve based solely on user preference, a designer also needs to take into account the cost curve, which is usually equally smooth – but can only be determined once the design process has sufficiently advanced. The difference between these curves represents the value to the manufacturer, and it is this value that the designer should maximise. The law of diminishing returns dictates that at the higher end user benefits will rise much slower than additional costs, so the manufacturer value curve will have a maximum.
The obvious examples of specifications where pass/fail value judgement is appropriate include standards and regulations. A product that almost but not quite fulfills CE-marking criteria has no value, and a product that does not adhere to industry standards might not find a wide market.
What is your take on specs vs. curves?