The job interview many managers still run, a friendly conversation that goes wherever it goes, is one of the weaker ways to predict who will do a job well. Structured interviews, in which every candidate answers the same job-related questions and is scored against a guide, came first for predicting job performance among the methods compared in a 2022 re-analysis of decades of selection research.¹

A good hiring process builds on that finding: a fixed sequence of steps that gathers the same evidence on every candidate, scores it against criteria set in advance and keeps discrimination law in view. It comes down to eight steps:

1. Define the job and what success looks like
2. Write an ad that describes the work
3. Screen every application against the same criteria
4. Use a structured interview
5. Add a work sample that mirrors the job
6. Score independently before anyone talks
7. Decide by the criteria, then discuss
8. Make the offer and plan the first weeks

Why a hiring process beats a good conversation

A planned hiring process beats an open conversation because conversations feel far more informative than they are. In a 2013 experiment led by psychologist Jason Dana, students predicted a fellow student’s grades less accurately after interviewing them than from background information, such as past grades, alone. The authors name two traps: sensemaking, where interviewers build a convincing story out of almost anything, even answers secretly given at random, and dilution, where vivid but irrelevant details crowd out the few facts that do predict. The participants were students, but the authors note that workplace reviews also find unstructured interviews poor predictors.²

Picture a candidate who studied at your old university and follows your football team. The conversation flows, and the one hard question, whether they have ever run a budget, gets less weight than it deserves: that is dilution at work. Hiring also rarely tells you quickly and clearly whether you were right, which is the setting where a gut feeling deserves less trust.

The first figure shows how far people trust interviews; the second, from a 2013 meta-analysis led by Nathan Kuncel, shows the payoff from combining evidence by a rule rather than by impression.³ In practice, write down the evidence you need before any interview, and let the conversation serve that list.

Which selection methods predict job performance best?

Methods built around the specific job predict best. In 2022, Paul Sackett and colleagues re-examined the meta-analyses behind Frank Schmidt and John Hunter’s widely cited 1998 summary and concluded that earlier statistical corrections had overstated how well many methods predict.¹

The study

Moderate evidence

Sackett and colleagues' 2022 re-ranking of hiring methods

After revising how earlier meta-analyses corrected for restricted samples, the authors estimated that structured interviews had the highest average validity for job performance, a correlation of 0.42, against 0.19 for unstructured interviews and 0.31 for cognitive ability tests. Most methods that ranked high before stayed near the top, with lower estimates.¹

Put plainly, the structured interview’s correlation with performance was roughly double the unstructured one’s, yet no method comes near certainty: a correlation of zero means no link and one means perfect prediction. The four strongest methods, structured interviews, job knowledge tests, scored biographical questionnaires and work samples, were all job-specific; years of experience barely predicted performance. Structured interviews’ validity also varied widely across studies, so in a given workplace it may fall short of the average.¹ A 2023 critique by In-Sue Oh, Huy Le and Philip Roth argued that the re-analysis underestimated how restricted samples of current employees were.⁴

What this means for you: spend assessment time on questions and tasks taken from the job, not on stand-ins for it. Hiring an accountant? A short reconciliation exercise and questions about month-end problems they have solved will tell you more than a chat about career goals.

Myth

Intelligence tests are the single best predictor of job performance.

Fact

A 2022 re-analysis ranked cognitive ability tests below structured interviews and other job-specific methods, though the size of the estimates is still debated.

Eight steps from job definition to first week

Each step collects better evidence or weighs it more fairly; deeper guides to each are planned in the hiring, onboarding and talent development category.

1. Define the job and what success looks like

Write the four or five things the new hire must achieve in the first year, then the skills each needs, and agree how much each counts before the ad goes out. For a customer support lead, that might be cutting reopened tickets (weighted most), coaching new agents and handling escalations calmly.

An analysis of the job comes first among the components of a structured interview listed in a 2014 review by Julia Levashina and colleagues.⁵ Every later question, task and score is checked against this list. Be wary of copying the last person’s profile: the best performer in one role is not always right for the next one, and the role should fit what the team needs to work well.

2. Write an ad that describes the work

Describe the work and its outcomes, not an imagined ideal person. Acas, Britain’s workplace advice service, gives examples of wording that can discriminate, such as “recent graduate” (age) or “handyman” (sex), and suggests advertising in at least two places.⁶

Masculine-coded words such as “competitive” made jobs less appealing to women in experiments with Canadian students, but gendered wording had negligible effects on who showed interest in a large 2023 study of real online postings.⁷⁸ In company records from nine large firms, referred workers were less likely to quit than other hires and similarly productive on most measures.⁹ Yet the US Equal Employment Opportunity Commission (EEOC), as of September 2026, warns that relying on word of mouth can break the law if nearly all new hires end up sharing one background.¹⁰

So use neutral wording because it costs nothing, not because it will widen your pool, and treat referrals as one channel among several; why acquaintances carry news your close contacts do not shows the value of reaching wider.

3. Screen every application against the same criteria

Turn each point of the job description into a checklist and have at least two people score every application against it, rating the skills and outcomes from step 1 rather than stand-ins such as years of experience.

Hiding names helps less reliably than many expect. A Swedish study of real recruitments found women and applicants of non-Western origin likelier to reach interview under anonymous applications, though minority applicants were no likelier to get offers.¹¹ In a French randomized trial, participating firms became less likely to interview minority candidates when CVs were anonymized. The authors traced this partly to which firms volunteered: their earlier working paper suggests that firms already favoring those candidates joined, then lost the ability to do so. Anonymity can only cancel a bias that is there.¹² If you try it, track who reaches interview before and after.

4. Use a structured interview

Ask every candidate the same questions, drawn from the step 1 criteria, and score each answer on a scale with examples of weak, good and strong answers. One line of the support lead’s guide might read: “Tell me about a time you calmed an angry customer.” A 1 blames the customer or the policy; a 3 resolves the complaint; a 5 resolves it and fixes the cause.

Levashina’s review lists 15 possible components of structure; those research uses most include job-based questions, the same questions for everyone, anchored rating scales and trained interviewers. Structure seemed to reduce the influence of demographic factors on ratings, the authors concluded, though few studies were available.⁵

The aim of every component is comparability: with the same questions and the same scale, a good story cannot stand in for evidence.

5. Add a work sample that mirrors the job

Ask candidates to do a small, real piece of the job, such as drafting a customer reply, debugging a function or planning a shift rota, and score it with the same rubric for everyone. Keep it short and clearly bounded, and ask HR whether it must be paid. Work samples ranked among the strongest predictors in the 2022 re-analysis, which also noted that they suit candidates who already have training or experience.¹

In Great Britain, Acas guidance as of September 2026 says an unpaid work trial should not usually last longer than a day, that the National Minimum Wage applies if it is not genuinely for recruitment, and that the Equality Act 2010 requires reasonable adjustments, such as extra time, for disabled candidates.⁶ For the US duty to accommodate applicants, see the law section below.

6. Score independently before anyone talks

Have each interviewer rate every answer alone, on the agreed scale, and submit the scores before hearing anyone else’s view. Levashina’s review counts not discussing candidates between interviews among the components of structure.⁵

Picture a debrief where the most senior person opens with “I loved her.” Everyone else now rates with that view in their ears, and four independent reads of the candidate can turn into one read repeated four times.

1. Score alone: each interviewer rates every answer on the agreed scale before talking to anyone
2. Combine by a rule: add the scores using weights set when the job was defined
3. Then talk: discussion checks the result and the evidence; it does not replace them

7. Decide by the criteria, then discuss

Add up the scores using the weights you set in step 1, and only then discuss. Kuncel’s 2013 meta-analysis of 25 samples from work and school found that combining scores by a formula predicted job performance better than holistic judgment, even when the experts judging knew the job and often had more information.³

8. Make the offer and plan the first weeks

Put the offer in writing, and plan the first weeks before day one: who explains the role, who introduces the team and what the first real task will be. Acas lists what an offer letter in Great Britain should include, such as the job title, pay, start date and any conditions like references.⁶ As of September 2026, GOV.UK says UK employers must check a job applicant’s right to work before employing them, and must not discriminate against anyone because of where they are from.¹³ In the US, employers complete Form I-9 to verify the identity and work authorization of every person they hire, citizens included.¹⁴ So check every new hire, not only those you assume come from abroad.

Hiring does not end at the signature: a 2007 meta-analysis of 70 samples led by Talya Bauer linked newcomers’ role clarity, confidence and sense of acceptance to their performance, satisfaction and intention to stay.¹⁵ A first-weeks plan is aimed at exactly those three.

WiserHours

The eight steps on one page

• Write first-year outcomes and weighted criteria
• Describe the work in the ad and advertise in more than one place
• Screen every application against the same checklist
• Ask every candidate the same job-related questions
• Use a short, bounded work sample with a rubric
• Collect every score before the debrief
• Total the scores, then discuss
• Send a written offer and plan the first weeks

Find this checklist online:
https://wiserhours.com/talent-development/how-to-design-a-hiring-process/#checklist-how-to-design-a-hiring-process-steps

Where the evidence is strong, and where it is mixed

The evidence is strongest for how to assess candidates and weakest for popular fixes to ads and CVs; “mixed” means good studies disagree.

Short of time? Start with the firmest rows, and treat the mixed ones as options to monitor, not fixes.

Hiring and discrimination law: US and UK basics

Discrimination law applies at every step, from ad to offer, and differs by country; this summary reflects official sources as of September 2026.

In the US, the EEOC says the federal laws it enforces make it illegal to discriminate against applicants because of race, color, religion, sex (including, in its guidance, pregnancy, sexual orientation and transgender status), national origin, age (40 or older), disability or genetic information. Neutral-looking rules that screen out one group more than others can be unlawful unless they are job-related and necessary (age has its own test). Employers may not ask about disability before an offer, should generally avoid questions that reveal race, sex, national origin, age or religion, and must provide an accommodation an applicant needs unless it causes significant difficulty or expense.¹⁰ Most of these federal laws cover only employers with at least 15 employees (20 for age), and state or local laws may also apply; ask HR or a lawyer which cover you.¹⁶

In England, Scotland and Wales, the Equality Act 2010 protects nine characteristics: age, disability, gender reassignment, marriage and civil partnership, pregnancy and maternity, race, religion or belief, sex and sexual orientation. Section 60 restricts questions about an applicant’s health before a job offer, and apart from a few sections the Act is not part of Northern Ireland law.¹⁷ In Northern Ireland and elsewhere, look for the local equality body or labor ministry.

So a well-meant “Any health conditions we should know about?” before an offer is one to drop in both countries, unless HR confirms an exception applies.

When to bring in HR or an employment lawyer

Bring in HR or an employment lawyer whenever a hiring decision turns on legal specifics you cannot check yourself. Start with official guidance: the EEOC and USCIS in the US, Acas and GOV.UK in Britain. As a rough guide:

• Now: a candidate complains of unfair treatment, or you are thinking of withdrawing an offer. In Great Britain, Acas warns that withdrawing an unconditional offer could be against the law, and runs a helpline.⁶ In the US, the EEOC notes that retaliating against someone who complains of discrimination is illegal.¹⁰
• Soon: before adding a test, an automated screening tool, a degree requirement or health checks, any of which could screen out one group more than others, or before hiring in a new country.
• Routine: once a year, review the ad wording, the channels and who passes each stage, keeping monitoring data away from hiring decisions. Treat hiring like any recurring process and improve it in small plan-do-study-act rounds.

The bottom line

The strongest hiring processes are settled before the first candidate arrives: written criteria, the same questions and tasks for everyone, and every score in before anyone talks. Structured interviews and work samples have the best evidence; gut feel from an open conversation has some of the weakest. When a choice turns on local law, ask HR or a lawyer.