Miraai Judgment

Scenes from your jobs, scored on what a candidate would do.

We build short scenes from situations your own people described, such as the difficult customer, the shift that goes wrong, or the moment a decision lands on someone. The candidate watches from the inside and then rates the responses, so they learn what the work is really like before they accept, and you see how they would handle it before you hire.

See how candidates respond Book a demo

Any role, any industry

Set where your people work.

Scenes can be live action or animation, in any setting and any language, and the decision, the responses and the scoring stay the same however a scene is made. These are examples from retail, restaurants, contact centers, theme parks, airlines and leadership.

**Resorts and theme parks**Front-of-house host

Every character is described in writing and approved before production, and that description drives every frame they appear in, so the people a candidate meets in the first scene are the people they meet in the last.

How a candidate responds

Watch the scene, then rate the responses.

A server on a Friday night has just been told about a child’s tree-nut allergy, and the cook wants the plate out. After the scene the candidate rates each possible response in turn. A candidate’s score comes only from how they rate the responses to each scene. Answering out loud is not part of the test. The demonstration below is silent, with subtitles.

Restaurants · ServerThe pass

Silent, with subtitles

Before you step in Who you are, and what you know

You are Maya, a server at a large casual-dining restaurant on a Friday night. As you set down the plates at table 9, the mother told you that her eight-year-old daughter has a tree-nut allergy. It was not mentioned when she ordered. The daughter's chicken comes with a small cup of pesto on the side. You have carried the plate back to the kitchen pass, where Luis is running the line with twelve tickets up.

Your turn

Luis has told you to take the cup off and run the plate. The table is waiting and the line is twelve tickets deep. What should Maya do next, and why?

Rate the responses

How effective is this response?

1Highly ineffective2Ineffective3In between4Effective5Highly effective

That is the whole candidate experience.

Your version is built from your own situations, in your brand, and delivered on Podium365.

The evidence

Scored the way the research says a judgment test should be.

How a score is produced, what our keying study found, and what the published research says about tests built this way.

The method paper on the Open Science Framework →

How answers are scored

Rated responses are scored against the key your experts set, weighted by how much they agree. A candidate’s score comes only from how they rate the responses to each scene. Answering out loud is not part of the test. No AI model scores, ranks or rejects a candidate. Because the scene fixes the situation, a candidate’s background and experience are not what is being judged.

Candidates are compared with the role profile, and no one is rejected automatically. Where AI is used

How the key is set, tested against a panel of experts

Before the AI panel drafts any key, it was measured against a human one. A panel of experts and the AI panel rated the same 96 responses of a leadership judgment test, and 9,433 real completions were rescored under every key. The AI panel’s ratings correlate .84 with the experts’, it rates harder by 0.57 on a five-point scale and agrees with itself too much, and the key is calibrated for both. The panel-keyed test is as internally consistent as the expert-keyed one, and its items behave the same for men and women, older and younger candidates, and across ethnic groups.

Read the keying study How the whole test is built

What the published research says

A 2026 meta-analysis of 302 studies and 244,257 people (Lievens and colleagues, Journal of Applied Psychology) found that situational judgment tests predict job performance (ρ = .23), citizenship behavior (.26) and counterproductive behavior (.35), and add to cognitive ability and personality tests. Video scenes predicted as well as text and carried less cognitive load. Answers in the candidate’s own words predicted as well as multiple choice, and answers recorded on camera carried the lowest cognitive load of any format, which the authors note bodes well for smaller ethnic subgroup differences. Validity was higher among real applicants (.29) than in research samples, so coaching and faking do not seem to hurt these tests, and the same paper found a customized SJT costs about 24 times an off-the-shelf cognitive test. The video scenes and the rated responses on this page follow those findings.

Lievens, F., Zhou, Y., Ng, J. W. X., Xi, L., Tian, Y., Sackett, P. R., and Martin-Raugh, M. (2026). A meta-analytic examination of the criterion-related validity and cognitive loading associated with traditional and new situational judgment test paradigms. Journal of Applied Psychology. doi:10.1037/apl0001420

What changes for you

Fewer surprises after the offer, on both sides.

Fewer surprises after the offer

A candidate who has watched the real job from the inside knows what they are saying yes to, and the people who would have left in month two tend to step back during the process instead.

Hiring managers believe it

The scenes are their situations, described in their words, so a score on your own job starts a different conversation from a score on a generic one, and managers act on it.

Decisions you can defend

Your own experts set what a good answer is, and every scene keeps a record of where it came from, who approved it and what the fairness checks found.

Judgment, not test-taking

Answers are weighted by how much your experts agreed, so going against their consensus costs a candidate and sitting on the fence does not pay.

Built for your job

From your situations to a working test in weeks.

We start from material you already have, such as job descriptions, performance write-ups and conversations with the people doing the work. Real situations come out of it and are matched to what the job needs before anything is written. The SJT Studio is why a test built for one role costs what an off-the-shelf test does, where a customized judgment test has typically cost about 24 times more (Lievens et al., 2026).

Week 1

Blueprint

Your material, what the job needs and what we will measure, all agreed before anything is written.

Week 2

Scenes and responses

The scenes and the responses candidates rate, then a full review for reading level, bias and what each response measures.

Week 3

Scoring

Your own experts decide what a good answer is, weighted by how much they agree. No AI model scores, ranks or rejects a candidate.

Week 4

Produced and live

Production, setup and delivery, with the full documentation handed over.

Runs where you already work

Delivered on Podium365, inside your applicant tracking system and in your brand, with nothing to install and nothing for IT to maintain.

What we need from you

One subject-matter expert for a few conversations across the build, and two reviewers to read what comes back. Everything else is ours.

Full documentation, every time

Where each scene came from, who approved it, how the scoring was decided and what the fairness checks found, which most situational judgment tests never come with. The documentation also records how AI was used to build the test, and confirms that no AI model scores your candidates. The method behind the key is published in the study above.

Next step

Send us one job description.

We'll build the first scene for it and show you, so you can look at the real thing instead of reading about it.

Book a demo Ask about a custom build