Psychometrics

Reliability (KR-20)

How consistently an exam measures — often reported for objective tests as the KR-20 coefficient.

What is reliability?

Reliability is the extent to which an exam measures consistently — if a candidate sat an equivalent version, would they get a similar result? It's a property of the whole test, not a single question, and it's one of the two pillars of a sound assessment (the other being validity).

For objective tests where each question is marked right or wrong, reliability is commonly reported as the KR-20 (Kuder–Richardson 20) coefficient. Like many reliability measures it runs from 0 to 1, and higher is better — values around 0.8 and above are generally considered good for a high-stakes exam.

Why it matters

A high-stakes result you can defend depends on a reliable exam. Reliability rises when questions are well-written and discriminate well, and when the exam has enough of them. It's exactly the kind of evidence regulators and awarding bodies expect to see.

Synap reports whole-exam reliability as KR-20, alongside item-level psychometrics. See it on the standardised testing page.

See rigorous assessment in action

Book a tailored demo and we’ll show you how Synap puts these concepts to work in real exams.