Selection

Selection is the process of choosing between applicants once a pool exists, ending with an offer to one of them.

It is worth separating from recruitment because the two demand different things. Recruitment is a marketing activity: writing the role, sourcing, advertising and persuading people to apply. Selection is an assessment activity: screening, testing, interviewing and deciding.

An organisation can be good at one and poor at the other, and the symptoms differ. Weak recruitment produces too few candidates. Weak selection produces plenty of candidates and bad hires, or a process so slow that the good ones withdraw.

Six: application screening, a recruiter screen, an assessment, interviews, the reference and background check, and the offer. A conventional Indian corporate selection process runs roughly as follows, though the middle steps vary considerably by level and sector.

The notice period question in the second step deserves emphasis in the Indian context. A ninety-day notice is common and it changes the entire hiring plan. Establishing it early is the difference between a role filled in eight weeks and one filled in five months.

Structured interviews are the strongest single predictor of job performance, ahead of job knowledge tests, work sample tests and cognitive ability tests. The ranking is not what most hiring processes imply, and it changed in 2022, when Sackett, Zhang, Berry and Lievens showed in the Journal of Applied Psychology that the estimates the field had used since 1998 were inflated by a statistical correction applied wrongly. Their revision moved the structured interview above cognitive ability, which had held the top place for fifty years.

• Structured interviews, where every candidate is asked the same questions against defined criteria and scored, are the strongest predictor in the revised estimates.
• Job knowledge tests follow closely.
• Work sample tests, where the candidate does a representative piece of the actual job, come next and are underused outside technical roles.
• Cognitive ability tests predict less well than older guidance claimed, and they carry adverse impact risk that needs care in how they are applied.
• Unstructured interviews, the default in most organisations, are weak predictors. They mainly measure how comfortable the interviewer felt.
• Years of experience, taken alone, predicts far less than most job advertisements assume.

The practical implication is cheap to act on. Writing the questions in advance, asking every candidate the same ones and scoring against agreed criteria requires no budget and improves the decision more than adding another interview round.

Any round that does not inform a decision the previous rounds did not. Every additional stage costs candidates. In a market where a good candidate holds two or three processes at once, the employer that takes five rounds over six weeks loses to the one that takes three over two.

Where an interview exists because a senior person wants to meet every hire, that is a legitimate reason and should be scheduled in parallel rather than sequentially.

Two patterns cause most avoidable loss. The first is unexplained gaps between rounds, where the candidate hears nothing for ten days and assumes rejection. The second is discovering the compensation mismatch at offer stage, which wastes the whole process and is preventable in the first screening call.

Selection decisions should be documented at the point they are made, for two reasons.

The first is quality. A scorecard completed immediately after an interview records what the interviewer actually observed. A recollection assembled a week later records an impression.

The second is defensibility. Where a rejected candidate questions a decision, or where a pattern in hiring outcomes is examined internally, the available evidence is whatever was written down. Criteria defined in advance and scores recorded against them are a coherent answer; a set of unstructured notes is not.

Personal characteristics that cannot lawfully or reasonably bear on the decision should not be collected in the first place. Date of birth, marital status, religion and caste appear on Indian application forms by convention rather than necessity, and removing them from the form is simpler than training interviewers to disregard them.

The process does not end at the offer, and in India the gap between offer and joining is where a meaningful share of hires is lost.

Notice periods of sixty or ninety days are common, which means an accepted offer sits exposed for months while the candidate remains in the market and reachable by their current employer.

• Establish the notice period in the first screening call, not at offer stage. It determines the hiring plan and it is the cheapest thing to ask.
• Release the offer letter quickly after the verbal. A verbal offer with nothing in writing for a week reads as uncertainty.
• Stay in contact through the notice period. Contact during that window is the single most effective retention activity available and almost nobody resources it.
• Expect a counter-offer and ask about it directly. A candidate who has not thought about how they will respond to one is more likely to accept it.
• Complete background verification before the joining date rather than after, so a discrepancy is resolved while there is still time.

Where a candidate declines after accepting, record the reason honestly. Offer drop-off concentrated in one function or one compensation band is a structural signal, and it is visible months before it shows up as an unfilled position.

Campus and high-volume selection operate differently from lateral hiring, and applying the lateral process to a thousand applicants does not scale.

The usual shape is a filter sequence: an online aptitude or coding assessment first, since it is the only stage that handles volume without human time, then group or panel rounds, then a final interview. The assessment carries most of the load and its design decides the quality of everything downstream.

• Set the cut-off from a pilot rather than by intuition. A threshold chosen without data either passes almost everyone or almost nobody.
• Watch adverse impact. A test that systematically filters out one group needs examining, and volume hiring is where such a pattern is large enough to detect.
• Keep the candidate experience tolerable. Campus cohorts talk to each other and to the placement office, and a process that leaves people waiting for hours costs access the following year.
• Decide the offer-to-join ratio in advance. Campus offers made in one season and joining months later have predictable drop-off, and hiring to headcount rather than to expected joiners leaves a gap.

The measure that matters afterwards is not offers made but conversion and retention at twelve months, read against the assessment score. That is the only way to find out whether the filter was selecting for anything real.