Specifications and test data: proving it works

Saying your design is 'strong' or 'easy to use' doesn't convince anyone. Saying 'it held 120 kg in three trials' does. Specifications and test data turn opinions into evidence.

What is a specification?

A specification (spec) is a list of what a design must do, written so you can test it. Many designers sort their specs into three groups:

Tolerances

Nothing is made perfectly. A tolerance is how far off the target a measurement can be and still pass. A bench seat 450 mm high, plus or minus 10 mm, passes at 445 mm but fails at 470 mm.

Fair and repeatable tests

A fair test changes only one thing at a time and keeps everything else the same. A repeatable test gives similar results each time. Run each test three times. If one result is very different, check what went wrong rather than just ignoring it.

Why the median?

The median is the middle result when you put them in order. It is less affected by one odd trial than the mean. If a shelf held 9 kg, 10 kg and 3 kg (because a screw was loose), the median is 9 kg, a fairer summary than the mean of 7.3 kg.

Round one, round two

Testing twice shows whether your changes actually helped. Put both rounds side by side in a graph. If round two is better, you have evidence. If it isn't, that's evidence too, and it tells you where to look next.