Short answer: To run structured interviews at high volume:
- List a few job-related requirements.
- Ask one question per key requirement that no document can prove.
- Write the scorecard before the first candidate.
- Deliver every interview the same way.
- Budget reviewer hours honestly.
- Calibrate reviewers.
- Record outcomes.
A named person reviews the evidence and decides.
A structured interview asks every candidate the same job-related questions and rates each answer against a standard written in advance. Validity here is an estimated correlation between what a selection method says about candidates and their rated job performance, corrected for measurement error: 0 means no relationship, 1 a perfect one. On the revised 2022 estimates, structured interviews average 0.42, a moderate link to job performance, and the highest of the 25 methods reviewed. Unstructured interviews average 0.19. Years of job experience average 0.07 (Sackett and colleagues, 2022). These are averages from many studies, mostly outside the Netherlands, and they measure overall job performance, not retention.
Most recruitment teams do not need to be convinced that a consistent, criteria-based interview is better than a friendly chat.
The harder question comes on a Monday morning with 200 new applications in the queue. How do you ask every candidate the same questions, rate every answer the same way, and still make offers before good candidates accept a job somewhere else?
On this page: Why high volume puts pressure on structure · Steps 1 to 7 · An example question set for a warehouse role · How staffing agencies can use this approach · Where technology helps, and where a person decides · Checklist · FAQ · How Radius Hire runs structured interviews
Why high volume puts pressure on structure
The Dutch labour market is cooler than it was, but most occupations are still short of people. In the first quarter of 2026, UWV still rated 87 of 93 Dutch occupational groups as tight or very tight. In that market, speed is not a luxury.
The researchers behind the revised estimates note that, in their traditional in-person form, structured interviews are generally not a viable strategy for high-volume jobs, and that when technology takes over delivery, validity has to be checked again (Sackett and colleagues, 2022). They call for research on adapting such methods for high-volume use while keeping their validity (Sackett and colleagues, 2023).
The arithmetic shows why. In the worked example in The Radius Papers 01: A New Standard for Frontline Hiring (2026), 200 applications a week at 30 minutes per candidate, including rating, take 100 recruiter hours if everyone is interviewed, and still 20 if one in five is. Those are example inputs, not measurements from any employer. The full model is in Step 5.
Even at the narrowest funnel, structure is a standing weekly workload. It has to be delivered the same way by every interviewer, every week. Where recruiter hours are fixed, structure is usually the first thing to give way. So the goal is not only to design a good interview. It is to design one that survives your real volume.
Step 1: Define a short list of job-related requirements
Start with the requirements that decide whether a hire works out, written the way a supervisor would say them. Do not start from the full job advertisement.
For a warehouse role, that list might be:
- operates a reach truck safely
- keeps pace across a full shift
- arrives reliably for a 05:30 start
- holds a valid certificate
- is allowed to work in the Netherlands
- accepts the real shift pattern
Treat the list as a working hypothesis. It gets stronger when it is based on a careful look at the job. It gets stronger again when you later check it against how hires actually perform and how long they stay.
A short list is the first protection against volume. Every requirement you add is another question, another rating and more reviewer time per candidate.
Step 2: Build a consistent question set
How many questions should a structured interview have?
Research does not give a single correct number. A practical rule: one question for each requirement that carries weight in the decision and that no document can prove. Where a document can prove it, such as a certificate, verify the document instead. For many frontline roles that means a handful of questions, not twenty.
Good questions for volume hiring:
- Tie to exactly one requirement, so each answer can be rated on its own.
- Are short and in plain language. Many frontline candidates answer on a phone, sometimes in a second language.
- Ask about real behaviour or real conditions. For example: "This role starts at 05:30, including some Saturdays. How would you travel to the site for that start?"
- Include what a CV cannot carry. Start time, shift pattern and physical demands, described honestly, also help candidates decide for themselves whether the job suits them.
Keep the set fixed for the role. If a question needs to change, change it for everyone from a set date, and note the change.
Step 3: Write the scorecard before the first candidate
Rating every answer against a standard written in advance is half of what makes an interview structured. In the research behind the 0.42 average (Sackett and colleagues, 2022), "structured" means standardised questions and a standardised way of scoring the answers (Huffcutt and colleagues, 2014).
For each question, write a simple rating scale with three short examples: strong, acceptable and weak. That is your interview scorecard. Written standards do two things. They make two candidates comparable, and they create a record someone can review later. Whoever applies the criteria must apply them the same way every time.
A full example follows below.
Step 4: Keep delivery consistent
Consistency is lost in small ways:
- a recruiter skips a question because the candidate "clearly" meets it
- the wording changes between interviewers
- follow-up questions are added for some candidates and not for others
To keep it:
- Ask the same questions, in the same order, with the same wording.
- Keep the channel the same for everyone for that role, whether phone, voice interview or in person.
- Offer an alternative channel on request, for example for a disability or language need, and use the same questions and rating standard in it.
- Record each answer and its rating, not only the final outcome.
- The GDPR requires you to tell candidates what their data is used for, on what basis, how long it is kept, their rights, including access, and whether any decision is automated (Articles 13 and 14). Telling them what is asked and who decides goes further. It builds trust.
Step 5: Plan reviewer workload honestly
Before launch, decide three numbers for the role:
- the share of applicants who will be interviewed
- the minutes per candidate, including rating
- who reviews, and when
Then check the weekly hours against what the team actually has.
How much recruiter time do structured interviews take?
In the paper's worked example, at 30 minutes per candidate, including rating, 200 applications a week take 100 recruiter hours if you interview everyone, 40 hours if you interview 40 per cent and 20 hours at 20 per cent. At a 40-hour week, that is 2.5, 1.0 or 0.5 full-time equivalents. Use your own minutes and volumes.
| Share interviewed | Interviews a week | Recruiter hours a week | Full-time equivalents (40-hour week) |
|---|---|---|---|
| Everyone | 200 | 100 | 2.5 |
| 40 per cent | 80 | 40 | 1.0 |
| 20 per cent | 40 | 20 | 0.5 |
Illustration built for the paper, not client data.
The formula: applications a week × share interviewed × minutes per candidate ÷ 60 = recruiter hours a week. If your full-time week is 36 hours, the same hours come to 2.8, 1.1 and 0.6 full-time equivalents. The full workload model is in the paper.
Be honest about the trade-off. Better evidence can cost more time per candidate. The business case depends on whether that time is paid back through fewer poor matches, fewer late surprises and less repeated recruitment. No study verified for the paper measures that return for Dutch frontline hiring. Your own figures are the only reliable answer.
Step 6: Calibrate the people who evaluate
The research spread for structured interviews is wide. The authors link it partly to procedures that are not built or run to the same quality (Sackett and colleagues, 2023).
Calibration keeps your ratings from drifting.
- Train every reviewer on the written standard before they rate.
- At set moments, have two reviewers rate the same set of recorded answers, compare the ratings, and discuss where they differ.
- When disagreements repeat, sharpen the written examples rather than leaving it to individual judgement.
- Keep the rating standard stable between measurement periods, or your numbers will not be comparable.
Step 7: Record outcomes and check them against the screen
Record a few outcomes for the people you hire:
- attendance in the first months
- early leaving within a period you define, such as the first 90 days
- a short supervisor rating at a fixed point, against the written requirements
Review these over several hiring rounds, not after ten hires.
This is the local evidence the researchers recommend (Sackett and colleagues, 2023). It is the only way to see where your own process sits within the research range.
If you compare outcomes between groups of candidates, take legal advice first. That can involve special categories of personal data, which the GDPR protects strictly.
An example question set for a warehouse role
This is an illustration written for this article. It is not a validated assessment framework. Adapt the questions and standards to your own role, test them, and check them against your own outcomes.
Role: reach-truck operator in a distribution centre, rotating early and late shifts, some Saturdays.
Question 1: Reliable arrival for a 05:30 start
"This role starts at 05:30, including some Saturdays. How would you travel to the site for that start?"
- Strong: names a specific, realistic way to arrive on time, including on weekends.
- Acceptable: names a plausible option but has not checked it for weekends.
- Weak: describes no realistic way to arrive at that hour.
Question 2: Match with the real shift pattern
"This role works rotating early and late shifts, including two Saturdays a month. Which parts of this pattern would be difficult for you, and how would you handle them?"
- Strong: confirms the full pattern and describes concrete arrangements for the difficult parts.
- Acceptable: accepts the pattern but has not thought through the difficult parts.
- Weak: cannot accept a core part of the pattern, or the answer is unclear.
Question 3: Reach-truck experience
"Describe the last time you operated a reach truck. What kind of warehouse was it, what heights did you work at, and how did you check the truck before use?"
- Strong: a specific, recent example, with a clear description of pre-use checks.
- Acceptable: a specific example with limited detail on checks.
- Weak: vague, or no practical example.
Note: this answer is still the candidate's own account. Dutch law requires operators of forklifts and reach trucks to be properly trained (Arbobesluit, Article 7.17c). A certificate from a training provider is the usual proof, and many sites and clients require a current one. It is not a state licence. Verify the certificate, and have competence demonstrated where that is safe and proportionate.
Question 4: Keeping pace across a full shift
"This role involves moving pallets for a full eight-hour shift. Tell me about a time you did physically demanding work or activity for several hours, paid or unpaid. What was hard, and how did you keep going?"
- Strong: a specific example, with a clear account of how they managed the pace.
- Acceptable: a relevant example with less detail.
- Weak: cannot describe any sustained physical effort, or how they kept going.
Note: ask about the job's demands and the candidate's experience of similar work, never about health or medical conditions. Check your questions with legal counsel.
How staffing agencies can use this approach
For a staffing agency, the same evidence problem sits inside its own screening. When a placement does not work out, the cost shows up as lost margin, recruiter time and client trust.
The same steps apply, with a few adjustments:
- Build the requirement list with the client. Ask the client's site supervisor what decides whether a new person works out, not only what is in the order.
- Keep one question set and one rating standard per client role, shared by every consultant who screens for it.
- Send the client structured evidence, not only a CV. Answers and ratings per requirement show the client what the shortlist rests on.
- Agree who decides. Record whether the agency, the client or both make the final selection, and on what evidence.
- Agree the data arrangements. Sharing candidate answers and recordings with a client involves personal data, so agree roles and responsibilities under the GDPR with the client and take legal advice where needed.
- Close the loop. Ask clients for attendance and early-leaving outcomes on placed candidates, so you can check whether your screen predicts them.
Where technology helps, and where a person decides
Technology can take on the work that makes structure hard at volume:
- delivering a fixed question set at any hour, on a phone, in several languages
- showing each answer next to the requirement it relates to
- keeping recordings, transcripts and summaries for review
Some things still need a person:
- writing good, job-related questions and the scorecard behind them
- checking that the criteria are applied consistently and fairly, because no tool removes bias on its own
- reviewing the evidence and making the decision: a named person, a record of what they saw and a real ability to disagree
Under Article 22 of the GDPR, a decision about a candidate that has significant effects may not be based solely on automated processing, unless a narrow exception applies, such as necessity for entering into a contract. AI used to recruit or select people is high-risk under Annex III of the EU AI Act. After the Digital Omnibus on AI (Regulation (EU) 2026/1744), the main high-risk duties apply from 2 December 2027. The full argument is in AI in Hiring: What Technology Should Evaluate, and What People Must Decide, and in The Tool Should Never Be the One Saying No.
Checklist: structured interviews at high volume
Before launch
- Requirements for the role are written down in a supervisor's words, and the list is short.
- There is one question per requirement that carries weight and that no document can prove. Documents, such as certificates, are verified instead.
- Questions are short, plain language and answerable on a phone.
- Start time, shift pattern and physical demands are covered with every candidate, without questions about health or medical conditions.
- A written scorecard, with strong, acceptable and weak examples, exists for each question.
- Candidate information explains what is asked, how answers are used and who decides.
- The share interviewed, the minutes per candidate and the reviewer hours are agreed and realistic.
While running
- Same questions, same order, same wording and same channel for everyone, with an alternative channel on request that uses the same questions and scorecard.
- Each answer and rating is recorded, not only the outcome.
- Reviewers are trained, and double rating is checked at set moments.
- A named person makes and records each decision, and can overrule any tool output.
After hiring
- Attendance, early leaving and a supervisor rating are recorded for new hires.
- Results are reviewed over several rounds, with the scorecard kept stable.
- Any comparison between groups is done lawfully, with legal advice.
FAQ
How do you run structured interviews at high volume?
Keep the question set short and tied to written requirements, write a scorecard before the first candidate, and deliver every interview the same way. Budget reviewer hours, calibrate reviewers and record outcomes. Technology can help with delivery and record-keeping; a named person decides.
How many questions should a structured interview have?
There is no research-based universal number. A practical rule is one question for each requirement that carries weight and that no document can prove; verify documents, such as certificates, instead. For many frontline roles, that is a handful.
Can asynchronous interviews be structured?
Yes, if every candidate gets the same questions in the same order and every answer is rated against criteria written in advance. The format does not make an interview structured; the fixed questions and the written standard do.
Is there evidence that AI interviews predict performance?
Too little to rely on yet. The researchers behind the main estimates say a body of validity evidence on asynchronous video interviews and AI-based scoring has not yet built up (Sackett and colleagues, 2023). Check your own outcomes.
Who should apply the rating criteria, a person or a tool?
A trained person should own the rating and the decision. A tool can deliver the questions and organise the answers.
Can staffing agencies use structured interviews?
Yes. Build the requirement list with the client, use one question set and scorecard per client role, and send structured evidence with the shortlist. Agree who decides and how candidate data is shared.
How Radius Hire runs structured interviews
Radius Hire asks every applicant the same structured questions, built on the requirements the employer sets, and shows the recruiter the evidence behind each requirement, labelled Proven, Claimed or Missing. It does not rank candidates or reject anyone on its own. A recruiter decides.
Applicants answer by voice in their own language: Dutch, English, Polish, Romanian, Ukrainian, Russian or French. They are told at the start that the interviewer is automated and that a person reviews the result.
See a structured voice interview in seven languages, with the evidence per requirement. Book a demo.
Sources
- Sackett, P. R., Zhang, C., Berry, C. M., and Lievens, F. (2022). Revisiting meta-analytic estimates of validity in personnel selection. Journal of Applied Psychology, 107(11), 2040 to 2068. DOI 10.1037/apl0000994
- Sackett, P. R., Zhang, C., Berry, C. M., and Lievens, F. (2023). Revisiting the design of selection systems in light of new findings regarding the validity of widely used predictors. Industrial and Organizational Psychology, 16(3), 283 to 300. Open access. DOI 10.1017/iop.2023.24
- Huffcutt, A. I., Culbertson, S. S., and Weyhrauch, W. S. (2014). Moving forward indirectly: Reanalyzing the validity of employment interviews with indirect range restriction methodology. International Journal of Selection and Assessment. DOI 10.1111/ijsa.12078
- UWV (2026). Spanningsindicator, first quarter of 2026. UWV spanningsindicator
- Regulation (EU) 2016/679 (GDPR), Articles 13, 14 and 22. EUR-Lex
- Regulation (EU) 2024/1689 (AI Act), as amended by Regulation (EU) 2026/1744. EUR-Lex
- Arboportaal. Gebruik van arbeidsmiddelen (Arbobesluit, Article 7.17c). Arboportaal
- Radius Hire (2026). The Radius Papers 01: A New Standard for Frontline Hiring. Sections 1.1, 3.7, 4, 5.3 and 7.2 to 7.7.