Selection Quality Scorecard

Compare what you actually use to choose people with what the research says predicts job performance, then upgrade the cheapest thing first, usually the interview you already run.

Selection method validity is strong research (Schmidt & Hunter, 1998; revised estimates: Sackett et al., 2022); the exact coefficients are contested (Oh, Le & Roth, 2023). Structured interviews have strong support (Campion, Palmer & Campion, 1997; Levashina et al., 2014). Rater bias is long established (Thorndike, 1920). Culture fit judgments can slide into cultural matching (Rivera, 2012); "culture add" is a practitioner term without research of its own. The score bands on this page are rules of thumb.

Use it when:

• You have just finished a hiring round and want to know whether the method, not just the outcome, was sound.
• Hires keep looking better in interviews than in the job.
• You are about to redesign interviews or buy an assessment tool and want to spend effort where it pays.

Skip it when:

• The problem is that too few candidates reach the end: map the stages with 4.15 Recruitment Funnel Map.
• You are setting up hiring for a role for the first time: start with 4.7 Hiring & Onboarding Checklist.

Personnel selection is among the best researched decisions in management, and the message is uncomfortable: how you select predicts how hires perform, and the methods that feel most informative are often not the strongest. Schmidt and Hunter (1998) summarized most of a century of validity studies. A later reanalysis by Sackett et al. (2022) argued that the older figures were inflated, lowered most of them, and put structured interviews at the top of the ranking. Structure works by cutting noise: the same questions, behavioral prompts, a scoring guide and independent scores leave less room for impressions to decide (Campion, Palmer & Campion, 1997). The noise is old news. Thorndike (1920) described the halo effect, where one general impression colors every specific rating. And Rivera (2012) showed how hiring for fit at elite firms turned into choosing people who resembled the selectors.

So the scorecard asks what you use, how structured it is, where bias can enter, and whether your scores predicted anything. The limit: the evidence ranks methods on average across many jobs; your own look back is small, so treat it as a check, not a proof.

Pick one role and one recent round. Mixing roles hides what works for which job.

Strong answer: "Field service engineer, autumn round, two hired from 23 applicants; reviewing both this round and the four engineers hired over the last two years."

Role:

Hiring round reviewed:

Number hired, and past hires you can look back on:

Illustrative example: the roles, the company and all numbers are made up.

A regional energy company in Sweden reviews how it hires field service engineers after one of two recent hires left within a year.

Section 2: the decision rested on a long unstructured interview and two reference calls. No work sample was used.

Section 3: 2 of 6. Questions varied by interviewer, there was no scoring guide, and the panel talked before anyone wrote a score.

Section 4: the CV screen used written criteria; nothing at all at the panel stage, where the senior manager spoke first.

Section 5: four past hires; the two rated highest in interview were the two who struggled with fault-finding on site. Pattern: the process rewards confident talk.

Decision: add a 45-minute work sample on a real fault, scored with a three-level guide, and write five core questions with strong-answer notes. Drop the open "tell us about yourself" round. Each panelist writes scores before the debrief, and the most junior speaks first.