Situational judgement tests: what they are and when they earn their place.
An SJT puts candidates in front of the dilemmas the job will bring. Built well, it is a fair, realistic early test of judgement. Built badly, it is a quiz about common sense.
On this page
What an SJT measures, the four steps to building a good one, and when it is worth the effort.
What an SJT is
A situational judgement test presents a series of realistic situations from the job – a customer who is angry about something that was not your fault, a colleague who is missing deadlines, a manager’s instruction that conflicts with a safety rule – each with a set of possible responses. Candidates judge the responses. Depending on the design they rate how effective each one is, rank them, or pick the most and least effective. Scenarios can be written, animated or filmed.
Answers are scored against a key, usually agreed in advance by experienced people in the role. Results are typically reported against a small number of competencies, for example dependability, leading others, handling change and building working relationships for a managerial test, so the output is a profile rather than a single mark.
What it measures
An SJT measures judgement about how to handle work situations. Where an SJT is built to measure a specific skill, it is most often leadership or interpersonal skill, and less often teamwork or job knowledge (Christian, Edwards and Bradley, 2010). The wording of the instruction matters. Asking what a candidate should do tests what they know about effective behaviour, and scores tend to relate more closely to cognitive ability. Asking what they would do tests their behavioural tendency, and scores tend to relate more closely to personality (McDaniel, Hartman, Whetzel and Grubb, 2007). Decide which you want before you write a single scenario.
The research evidence is that SJTs predict job performance, and that they do so best when they are developed from an analysis of the job and when what they target matches the performance you care about (McDaniel and colleagues, 2001 and 2007; Christian and colleagues, 2010).
When an SJT earns its place
- High-volume recruitment, where you need a consistent, fair first step that hundreds of people can take
- Graduate, apprentice and entry-level roles, where candidates have judgement but little track record
- Roles where how people treat customers, colleagues or the public matters as much as what they know
- When you want candidates to see what the job is really like, and to rule themselves out if it is not for them
- When your organisation’s values should show up in day-to-day decisions, and you can write scenarios that test them
When it does not
- Small numbers of candidates, where the cost of building a good test outweighs the benefit
- Technical roles where the real question is specialist knowledge or skill; use a knowledge test or work sample
- When an off-the-shelf test’s scenarios look nothing like your work
- As the only method, or as the final decision
- Where you cannot check the results in a supervised setting later on
How a good SJT is built
Four stages, each built from the job itself.
Collect real situations
Ask people who do the job, and people who manage them, for examples of situations where the way someone handled it made a real difference, for better or worse. These critical incidents become the raw material for scenarios.
Write scenarios and plausible options
Each scenario needs a clear dilemma and four or five responses that a reasonable person might choose. If one option is obviously right and the others obviously wrong, the item tests reading, not judgement. Include responses that are well meant but less effective; real mistakes usually look like that.
Agree the scoring key with experts
Experienced job holders and managers rate the options independently. Keep items where they broadly agree, and rewrite or drop items where they do not.
Trial it and check it
Trial the test with people already in the role, check that scores relate to how well they do the job, and check whether results differ between groups of candidates. Review the scenarios every few years, because work changes.
What AI changes
A 2024 study in Scientific Reports gave five widely used AI chatbots a social situational judgement test, ten times each, and compared them with 276 pilot applicants. All five performed at least as well as the applicants – two were statistically level with them and three scored significantly higher (Mittelstädt and colleagues, 2024). The test was multiple choice, in text, with expert-rated options – the format of many online SJTs. So a candidate who pastes scenarios into a chatbot during an unsupervised test may do well. The authors also note that a good score on hypothetical scenarios does not guarantee good behaviour in real situations.
That does not make SJTs useless. It changes how to use them. We suggest treating an unsupervised SJT as an early sift rather than a decision, telling candidates clearly what help is and is not allowed, and building a supervised step later in the process where candidates discuss how they would handle a situation, so the judgement you saw on screen can be tested in person. More on this in hiring in the AI era.
Reading the results
Use the competency profile, not just the overall score. A strong overall result with a clear weakness in, say, handling change tells you where to probe at interview. Be careful with cut-off scores: set them from evidence about the role and the candidate pool, check what they do to the make-up of your shortlist, and do not treat a narrow miss as a clear fail. And feed the results forward: the scenarios a candidate found hardest make good interview questions.
To see where an SJT fits among other methods for your role, try the selection method chooser.
Drawn from great{with}talent’s situational judgement test design for graduate, managerial, retail and inclusion scenarios (2015–2021). Research: McDaniel, Morgeson, Finnegan, Campion & Braverman (2001), Journal of Applied Psychology 86(4); McDaniel, Hartman, Whetzel & Grubb (2007), Personnel Psychology 60; Christian, Edwards & Bradley (2010), Personnel Psychology 63; Mittelstädt, Maier, Goerke, Zinn & Hermes (2024), “Large language models can outperform humans in social situational judgments”, Scientific Reports 14, 27449. Situational judgement tests are part of FindingPotential.
The clearest thirty minutes in your week.
Book a walk-through with a real person, or start with a pilot or a 30-minute conversation.