Sit down, have your tea. You have just been told that the next hiring round will include a “psychometric assessment”, and you are the person who has to explain to the managers what it is. Good. Let us make sure you can do that without reaching for vendor brochures, because the brochures are written to sell, and the real picture is more useful and a little less glamorous.
Three different things hide under one word
“Psychometric” simply means measuring something psychological in a standardised way. At work, three families of tests get used, and they behave very differently.
The first is cognitive ability tests: numerical reasoning, verbal reasoning, logical patterns. They have right and wrong answers and a time limit. Of all the selection tools studied over decades, general cognitive ability has among the better track records for predicting job performance, especially in complex roles, though it is far from perfect and the size of the effect depends on the job.
The second is personality questionnaires. Here there are no right answers; you describe how you usually behave. The better-regarded ones are built around the Big Five traits (openness, conscientiousness, extraversion, agreeableness and emotional stability). Conscientiousness has the most consistent link to performance across jobs, with modest effect sizes. Be careful with popular type-based tools such as the MBTI. They are widely used in workshops and are fun, but they are generally considered a weak basis for selection decisions, partly because they sort people into categories when traits are really spread along a continuum.
The third is situational judgement tests: short work scenarios where you choose what you would do. These sit between the other two and can be built to look very like the actual job, which candidates find more acceptable.
Reliability and validity, without the jargon
A weighing scale that shows 62 kg in the morning and 66 kg an hour later is not reliable. A test that gives a candidate very different scores on two sittings has the same fault. Reliability is consistency.
Validity is the harder question: does the score actually relate to something that matters? A perfectly consistent scale can still be measuring the wrong thing. For hiring, the validity you care about is whether higher scores go with better performance in this kind of job. Ask any vendor for evidence: studies on roles like yours, sample sizes, and who ran them. If the answer is a slide with a logo wall, that is not evidence.
A score means nothing without a comparison group
Say a candidate answers 24 of 30 numerical questions correctly. Is that good? You cannot tell. It depends on the difficulty and on the norm group, the sample of people whose scores the result is compared against. A score at the 80th percentile against graduate engineers is a different statement from the 80th percentile against a general working population. Norms collected in another country or decades ago may not suit Indian campus candidates. Good providers state their norm groups; some allow you to build local norms once you have enough data of your own.
How to use them without doing damage
Consider an illustrative hiring drive: 200 applicants for 12 analyst posts at a logistics firm in Visakhapatnam. A cognitive test is given to all 200 as an early filter, and the top 60 move on to interviews. That is a sensible use: it is cheap, standardised and applied to everyone. But three cautions apply.
- Cut-offs should be set deliberately. If you fix the pass mark at the score that leaves exactly 60 candidates, you have a number, not a standard. Look at how successful current employees score and decide what the role genuinely needs.
- Check for adverse impact. If a test systematically screens out a group (by gender, region or language background) without a link to job performance, it is unfair and a legal and reputational risk. Compare pass rates across groups after each drive.
- Do not use personality scores as a verdict. They are easy to fake when candidates know what the role wants, and they describe tendencies, not capacity. A reasonable use is as a prompt for interview questions: “the profile suggests you prefer to work independently; tell me about a project that needed constant coordination”.
Also consider practicalities that people forget. Candidates should know why they are being tested and how the results will be used and stored. Under India’s data protection law, personal data collected for hiring needs a proper basis and handling; ask your legal team and the vendor how scores are retained and for how long. Provide accommodations for candidates with disabilities. And tests taken on a phone in a noisy hostel room tell you about the room.
What a test can never do
It cannot tell you whether someone will learn your particular system quickly, tolerate your particular manager or care about your customers. It captures a snapshot under artificial conditions. Combined with a structured interview and a work sample, it becomes one voice among several, and that is the right position for it. Used alone, as a gate, it hands a lot of power to a number that does not deserve it.
So when the managers ask you, “will this tell us who is good?”, you can say: it will tell you something, about specific things, to a limited degree, as long as we pick a test with evidence, compare scores sensibly and never let it be the only thing we look at.
Figures and names in the examples are illustrative.
Leave a Reply