On quality

When does quality matter? Is it worth paying for? It depends. Things are either used

  • by end-users – pizza, music, pencils, books, or software applications – or
  • in systems – screws, microchips, conveyor belts, or software libraries.

Systems can be part of other systems and so on; eventually there’s an end user. Things could be used by end-users or in systems regardless of the intention behind their design (wood, stone, fire, flowers, music).


Poor quality manifests as

  • Poor performance:
    • Higher fuel consumption in cars.
    • Slower computations in computers.
    • Less tasty and healthy tomatoes.
    • Harder to understand web app.
  • Higher risk of failure:
    • Lightbulb goes out.
    • Plane crashes.
    • Chicken fillet has salmonella.
    • Heater stops heating.

Poor quality costs

In end-user things, poor performance can be wasteful, annoying, expensive, and distracting to the end-user. Those are all bad things and I think the cost of distraction in particular is usually underestimated. Higher risk of failure can increase mental load – if you have to remember to check for failure, make plans in case of failure, or be interrupted when it fails. Replacement cost of frequent failure can be high. Using the thing might be unacceptable for certain tasks because the cost-weighted risk is too high.

In system things, poor performance is usually amplified in the system, which also amplifies the costs. Drafty windows in a cold country probably cost more in heating bills over a few years than fixing the windows would. Higher risk of failure of individual parts of a system increases the risk that the system as a whole fails; how much depends on the system’s resilience characteristics.


Since systems can amplify both poor performance and cost, it is usually worth paying for quality. System costs not borne by the builders create incentives for compromising on quality even when the overall economics favour high-quality: rental/pre-built housing, government processes, and software monopolies are examples of this.

You may think poor performance in end-user things is not amplified therefore unproblematic, but I think poor quality is costlier than we realise. The things an end-user interacts with or relies on – including procedures and routines – can be considered a system and the accumulated costs of low-quality things – reduced focus, time, and productivity – can easily become significant.

So when in doubt, choose quality.