SpacesOpen app
People — excited man in superhero
D · Faster, fairer hiring

Reducing Bias in Hiring: A Practical Playbook

What the evidence supports when you want fairer hiring: standardised criteria, structured interviews, scored shortlists and measured selection rates.

UnframedNovember 16, 20266 min
  • The hiring changes with real evidence behind them are changes to the process: agree the criteria before you advertise, score every applicant against those criteria, ask all candidates the same interview questions on the same rating scale, and keep a record you can audit afterwards. None of that promises an unbiased outcome. It narrows the range of things a decision can turn on, and it leaves a trail you can inspect when the numbers look wrong.
  • The training day has been studied heavily and does not do this.

Start with what the evidence does not support

  • The Behavioural Insights Team reviewed the evidence on unconscious bias training for the Government Equalities Office in December 2020. Its finding on behaviour was flat: there is currently no evidence that this training changes behaviour or improves workplace equality in terms of representation of women, ethnic minorities or other minority groups in leadership positions, or in reducing pay inequalities.
  • The review went further. Drawing on the Equality and Human Rights Commission's assessment of 18 papers, it reported that such training can raise awareness of implicit bias and can reduce implicit bias, with no evidence it eliminates it, and that most of the evidence reviewed did not use valid measures of behaviour change at all. It also recorded a risk of backfiring, including a study of 830 medium to large US companies over 30 years in which mandatory diversity training either did not change the number of women in management or reduced it.
  • Its recommendation to employers was to invest in initiatives focused around processes, for example recruitment and progression, that have better evidence of efficacy. The rest of this playbook is what those look like in a hiring pipeline.

Fix the criteria before you fix anything else

  • Bias has the most room to operate where the standard is unwritten. So the first change is deciding what the role requires, in writing, before the advert goes out.
  • UK government guidance on inclusive job descriptions is specific about the edits that matter. Cut long lists of must-haves. Replace character traits with observable behaviours, so that "you are reliable" becomes "you keep confidential information to yourself". Avoid terms tied to male stereotypes such as competitive and dominant. Remove phrasing that deters carers and people returning to work. We covered the rewrite in detail in [how to write a job description that attracts the right people](/spaces/write-job-description-that-attracts).
  • If you draft adverts with automated writing tools, the same guidance warns that models trained on historical data can repeat the gender bias in past recruitment practice, and that reviewing the output remains the employer's job under the Equality Act 2010.

The process changes, in the order they pay off

ChangeWhat it does
Score CVs against the written criteria, one criterion at a timeStops the first-pass reviewer from forming a whole-person impression and then justifying it
Ask every candidate the same questions in the same orderMakes two answers comparable, which is the only way a panel disagreement can be about evidence
Agree benchmark answers before the first interviewRemoves the moving standard that lets a weak answer count when the candidate is likeable
Run a panel of at least two people, mixed by sex and ethnicityUK government guidance recommends this; a second rater makes one person's habits visible
Drop questions about pay historyRequired across the EU under Article 5 of Directive 2023/970, and it stops one employer's underpayment following a person through their career
Publish the pay range in the advertThe same article gives applicants the right to the initial pay or its range before the interview, based on objective, gender-neutral criteria
Keep the notes, scores and guidesWithout a record there is nothing to audit, and nothing to defend
  • Structured interviewing is the single change with the most guidance behind it. UK government guidance says structured interviews are more likely to lead to fairer outcomes for applicants than unstructured methods, and can reduce the impact of an applicant's characteristics, such as their sex, on the interviewer's decision. The mechanics are in [how structured interviews improve hiring quality](/spaces/structured-interviews-hiring-quality).

Name the rating errors out loud

  • The OPM structured interview guide lists the errors that show up in scoring: giving higher ratings to a candidate who appears similar to the interviewer, letting a strong score on one competency lift the others, rating everyone in the middle to avoid a call, and rating a candidate against the person interviewed before them rather than against the scale.
  • Panel members correct each other on these once they have the vocabulary. Five minutes in the kickoff is enough to hand it over.

Measure the pipeline, or you are guessing

  • If you do not count what happens at each stage, you cannot say whether a change did anything. Count how many candidates enter and survive each stage, and where the law lets you collect the data, break those rates down by group.
  • The EEOC's guidance sets out why this matters in the US: a neutral test or selection procedure that disproportionately excludes people on a protected ground has to be shown to be job related and consistent with business necessity, and where a less discriminatory alternative would predict job performance as well, the employer is expected to use it. Employers also stay responsible for the procedures they run even when a vendor supplied them and the vendor's documentation says they are validated.
  • That last point decides how ranking software should sit in your pipeline. A ranking orders a pool against stated criteria and shows its reasons so a recruiter can act on them or overrule them. The decision to advance or decline a person is made by a person, on a record. No score should carry that decision on its own, and we do not build one that does. The reason to rank at all is that a keyword filter is already making that call invisibly, which is [how qualified people get sorted out before anyone reads them](/spaces/keyword-matching-vs-skills-ranking).

What this playbook honestly gets you

  • Structured process reduces variance in how candidates are compared. Two applicants with similar evidence end up with similar scores more often, and when they do not, you can see where the scores diverged and ask why.
  • It does not make your organisation fair, and nobody should sell it to a board on that basis. The criteria can still be wrong. The pool can still be narrow. Ranking [on proof of fit](/spaces/hiring-on-proof-of-fit) changes what gets measured, which is a better starting point than keyword overlap, and it still leaves the judgment with the people doing the hiring.

Frequently asked questions

How do you reduce bias in hiring?
Change the steps of the process. Training the beliefs of the people who run it has been studied at length and has not been shown to change what they do. Agree the criteria before you advertise, score CVs against those criteria, ask every candidate the same interview questions with the same rating scale, use a panel of at least two people, and record the reasons for each decision. UK government guidance recommends all of these as evidence-informed employer actions.
Does unconscious bias training work?
The Behavioural Insights Team's evidence review for the Government Equalities Office found no evidence that unconscious bias training changes behaviour or improves workplace equality in terms of representation or pay. The same review flagged a risk of backfiring effects, and recommended employers invest instead in process changes in recruitment and progression, which have better evidence behind them.
What is a diverse shortlist?
A shortlist whose candidates were selected by scoring everyone who applied against the same written criteria, rather than by a reviewer's sense of who looks right. Widening where you advertise changes who applies; scoring against criteria changes who survives the first cut. Both steps have to happen or the shortlist reverts to whoever resembles the last hire.
Can software remove bias from hiring?
No, and any vendor who says otherwise is selling a claim they cannot stand behind. Ranking software can order a pool against stated criteria and show its reasons, which supports the person making the decision. The EEOC's guidance is explicit that the employer stays responsible for a selection procedure even when relying on a vendor's documentation, so you monitor the effect of any screen you run.
Is it legal to ask candidates about their current salary?
It depends where you hire. Article 5 of EU Directive 2023/970 states that an employer shall not ask applicants about their pay history during their current or previous employment relationships, and Member States had to bring the directive into force in national law by 7 June 2026. Several other jurisdictions have their own pay history bans, so check the rule for each country you recruit in.

Sources

  • Unconscious bias and diversity training: what the evidence says (Behavioural Insights Team for the Government Equalities Office, December 2020)
  • Use fair and structured interview techniques (GOV.UK employer guidance)
  • Make job descriptions inclusive (GOV.UK employer guidance)
  • Employment Tests and Selection Procedures (US Equal Employment Opportunity Commission)
  • Directive (EU) 2023/970 on pay transparency, Article 5 (EUR-Lex)
  • Structured Interviews: A Practical Guide (US Office of Personnel Management, September 2008)
Next step

Try it on your next role

Unframed builds shortlists by ranking people on proof of fit.

Open the appBook a demo
Read next
  1. 01Recruiting Metrics Every Talent Team Should Track in 20265 min
  2. 02Proof of fit: what to collect before the first interview6 min
  3. 03Why You're Not Hearing Back From Job Applications5 min

Unframed.

Shortlists ranked on proof of fit.

SpacesCompaniesTalentContact
© 2026 UnframedHire on proof.