How to Evaluate Candidates Fairly: Scorecards and Bias Checks

Last updated: 21 September 2026 · By the Asuraa Team

What does it mean to evaluate candidates?

To evaluate candidates means judging how well each person's evidence matches what the job needs, using the same criteria for everyone. It comes after screening, which narrows a large pool, and before the offer.

Our guide to screening candidates covers the filtering stage. This guide covers the judging stage: scorecards, structured interviews, scoring and checking for bias.

How do you evaluate candidates step by step?

Use these eight steps in order. Steps 1 to 3 are done before you meet anyone.

1. Analyse the job. List the main tasks and the competencies each one needs.
2. Build a scorecard. Choose four to six criteria, define each and set a rating scale.
3. Pick methods. Combine, for example, a work sample and a structured interview.
4. Write the questions. Use the same questions in the same order for every candidate.
5. Take notes and score independently. Each interviewer records evidence and scores before any discussion.
6. Hold a calibration debrief. Compare scores against the rubric and resolve gaps with evidence.
7. Check for bias. Look at both the individual decision and the pattern across candidates.
8. Decide and document. Record the reason for each decision against the criteria.

Why start with a job analysis?

Start there because it tells you what to evaluate. OPM's assessment strategy guidance says an assessment strategy should begin by identifying critical competencies from job analysis, so that agencies assess what matters for performance.

OPM's guidance is written for US federal hiring, but the logic applies to any employer. A scorecard that comes from real tasks is easier to defend than a list of traits chosen at the last minute. Our job description guide shows how to turn tasks into skills.

How do you build a candidate scorecard?

A scorecard has a small set of criteria, each with a definition, a scale and example answers. Google re:Work describes rubrics that give illustrative examples of what poor, borderline, solid and outstanding answers look like for each attribute tested.

Four to six criteria is a practical range in our view. More than that, and interviewers stop scoring carefully.

How do structured interviews work?

A structured interview asks every candidate the same predetermined questions in the same order and rates answers on the same scale. That is OPM's description, and it adds that the method measures job-related competencies by asking about past behaviour or proposed behaviour in hypothetical situations.

Google re:Work's guide to structured interviewing describes four components: vetted role-relevant questions, detailed recorded feedback, standardised scoring rubrics and interviewer training with calibration. It separates behavioural questions ("tell me about a time...") from hypothetical ones ("imagine if..."), both with follow-up probes.

OPM also has a page on structured interviews with a guide, presentation and examples of questions with rating scales.

Should you use more than one assessment method?

Yes, because people respond differently to different methods. OPM's guidance says using multiple methods improves validity and reduces errors, since some people do better on written tests and others in interviews.

OPM also says technology can streamline initial screening but should not be used for final selection decisions. So if you use software to rank applicants, let people make the last call and record why.

How do you check for bias when you evaluate candidates?

Check both the design of the process and its outcomes. Google re:Work says research shows structured interviews increase predictive validity and decrease demographic group differences.

Google also reports that rejected candidates who had structured interviews were 35% happier than those who did not. That figure is Google's own reported result.

OPM says assessment procedures should minimise adverse impact across demographic groups and show job-relatedness where differences appear. The US Equal Employment Opportunity Commission says employers remain liable for discriminatory outcomes, including when a neutral-looking practice has an unjustifiable disparate impact, and gives the example of video interview software that scores an applicant low because of speech patterns linked to a disability. See the EEOC's explainer, and note it describes US law, so take advice on your own legal obligations in India.

In practice, try these checks, which are our suggestions:

• Score before discussing. This stops the most senior voice from anchoring everyone.
• Compare like with like. Ask whether candidates who gave similar evidence got similar scores.
• Review the rubric for proxies. Watch for criteria such as "culture fit" that can hide personal preference.
• Track outcomes by stage. If you collect demographic data, do it lawfully and with clear consent; otherwise review by source and stage.

What does a worked evaluation look like?

The example below is an illustration for a marketing analyst role, and the names, scores and criteria are invented.

Two interviewers scored Candidate A independently on four criteria, each from 1 to 4: analysis 3 and 4, communication 2 and 2, tool skills 4 and 3, ownership 3 and 3. Because the analysis scores differed, the debrief went back to notes: one interviewer recorded a correct funnel breakdown, and the other noted the candidate missed a data-quality issue.

They agreed on a 3 after checking the rubric. The decision then rested on the written evidence for each criterion, not on who spoke most confidently.

What do most guides on how to evaluate candidates get wrong?

Many guides list qualities to look for and stop. These are the gaps that let bias in.

• They start with traits, not tasks. Without a job analysis, criteria become opinions.
• They skip written anchors. A 4 from one interviewer may mean a 2 from another.
• They discuss before scoring. Group discussion first tends to pull everyone toward one view.
• They let a tool decide. OPM says technology should not be used for final selection decisions.
• They never check the process. Good intentions do not show whether outcomes differ across groups.

FAQ

How do I evaluate candidates fairly?

Build a scorecard from the job's real tasks, ask every candidate the same questions, and score each answer against written rubrics. Have interviewers score independently before they discuss, then compare evidence, not impressions. Record the reason for each decision against the criteria.

What is a candidate scorecard?

A scorecard is a short list of job-related criteria, each with a definition, a rating scale and example answers. Interviewers score every candidate on the same criteria and write evidence beside each score. Google re:Work describes rubrics with examples of poor, borderline, solid and outstanding answers.

What is a structured interview?

It is an interview where every candidate gets the same predetermined questions in the same order and is rated on the same scale, as OPM describes. Questions can ask about past behaviour or hypothetical situations, with follow-up probes. Interviewers should be trained and calibrated.

How many criteria should a scorecard have?

There is no fixed rule. Four to six is a practical range in our view because each criterion needs a clear definition and evidence. If you need more, the role may be too broad, so consider splitting it or deciding which skills you can teach after hiring.

How do I reduce bias when evaluating candidates?

Use structured interviews, written rubrics and independent scoring, and review outcomes for patterns. Google re:Work says structured interviews decrease demographic group differences, and OPM says procedures should minimise adverse impact. Do not let software make the final decision, and take legal advice on your obligations.

Can AI tools evaluate candidates?

They can help with early screening, but people should make the final decision. OPM says technology should not be used for final selection decisions, and the US EEOC says employers remain responsible for discriminatory outcomes. Test any tool for job-relatedness and review its results by group.

Final thoughts

Fair evaluation comes from process: a job analysis, a scorecard, the same questions for everyone and scores written before the debrief. None of this needs special software, only discipline.

After the decision, our guide on how to reduce time to hire shows how to keep a fair process fast, and our post on what skill-based hiring is explains why evidence beats credentials. You can also post and browse roles on the jobs page on asuraa.in.