Short answer: To screen candidates for frontline roles, write down what the job needs, then collect, for each requirement, the kind of evidence that can answer it. Shift fit needs the same structured question for everyone, a certificate needs a check with the issuer, safe driving needs a supervised task, and staying power shows only after hire.
Key takeaways
- A CV mostly carries claims. It is a good source of questions and a weak basis for a decision.
- Candidate evidence comes in five kinds: claimed, stated, demonstrated, verified and performed. Each one answers a different question.
- Higher is not always better. A verified certificate proves a fact, not a skill.
- Some requirements, such as staying beyond 13 weeks, cannot be answered before hire. Record them after the start.
- One role can be mapped in a 45 minute session with its supervisor.
On this page: Why does screening at volume fall back on the CV? · What are the five kinds of evidence a recruiter can collect? · Which evidence does each requirement need? · Why is higher on the ladder not always better? · How does this relate to the Signal Gap? · What rests on research, and what is our framework? · How do you map one role in 45 minutes? · So how should you screen candidates for frontline roles? · How Radius helps · FAQ
Why does screening at volume fall back on the CV?
Because the CV is the fastest thing to read when time runs out.
Illustrative example (a composite scene, not a documented case). It is Monday, 08:10, at a staffing branch in Venlo. Over the weekend, 212 people applied for 30 order picker jobs, and the client wants the first group on site next Monday. Three recruiters have five days. If each of them gives first screening 10 hours that week, they have about 8.5 minutes per application for reading, calling and notes together.
Monday, 08:10, at a staffing branch in Venlo
At that pace, what gets read is a job title, an employer and a date range. Those are claims: what the candidate chose to write about themselves. They rarely say anything about what decides whether a frontline hire works out, such as the real shift pattern, the journey to a 05:30 start, the physical pace of the work and whether a certificate is still valid.
Date ranges then turn into "years of experience", which tells you surprisingly little. In the revised research averages, mostly from studies outside the Netherlands, years of job experience predict later job performance at an estimated 0.07, where 0 means no relationship and 1 a perfect one; structured interviews reach 0.42 (Sackett et al., 2022). Our article on structured vs unstructured interviews explains what those averages do and do not show.
How well each predicts later job performance
Revised research averages, mostly from studies outside the Netherlands
The recruiters are not careless. The screen collects what is quickest to read, not what each requirement needs, and the fix is to decide, requirement by requirement, which evidence to collect in the minutes available.
What are the five kinds of evidence a recruiter can collect?
Claimed, stated, demonstrated, verified and performed evidence, each answering a different question.
The Proof Ladder
Five kinds of evidence, ordered by where the evidence comes from
- ClaimedWhat candidates say about themselves on their own terms
- StatedWhat they tell you when everyone gets the same question
- DemonstratedWhat they show you
- VerifiedA fact checked against an independent source
- PerformedWhat they did in the job after hire
Higher is not always better.
Radius Hire calls this way of sorting candidate evidence the Proof Ladder: five kinds of evidence, ordered by where the evidence comes from. It is our framework for designing and checking a screening process, not a scientifically tested scale. A later section shows which parts rest on research.
In plain words: claimed evidence is what candidates say about themselves on their own terms; stated evidence is what they tell you when everyone gets the same question, recorded against a requirement written in advance; demonstrated evidence is what they show you; verified evidence is a fact checked against an independent source; and performed evidence is what they did in the job after hire. Stated evidence follows the standard research description of a structured interview: questions based on the job, the same questions for everyone, and answers rated against criteria agreed before the first candidate (Levashina et al., 2014).
| Kind of evidence | The question it answers | Typical evidence | What it can establish | What it cannot establish |
|---|---|---|---|---|
| Claimed | What does the candidate say about themselves? | CV, application form, informal chat | Work history as presented; what to ask next | Whether it is true or done well |
| Stated | What does the candidate confirm when asked the same question as everyone? | Structured questions on shifts, travel, experience | Comparable answers on what a CV leaves out | Truth or skill: it is still their word |
| Demonstrated | Can the candidate show the knowledge or skill? | Knowledge question with a model answer; short supervised task | That they know what to do, or can do it, in that setting | Performance over months of real shifts |
| Verified | Is a specific fact true? | Certificate confirmed with the issuer; licence; identity and right to work | That the fact is true | Whether the person does the job well |
| Performed | What happened in the job, and did the screen predict it? | Attendance, early leaving, supervisor rating | Whether hires worked out and the screen helped | Anything about a new candidate before hire |
A certificate copy handed over by the candidate is still the candidate's evidence until someone confirms it with the issuer. And performed evidence is noisy: across published studies, two supervisors rating the same person's overall performance agreed only moderately, with an average reliability of 0.52 (Viswesvaran, Ones and Schmidt, 1996).
See how the AI Interviewer collects stated evidence from every applicant for a role.
Which evidence does each requirement need?
The evidence that can answer the question behind the requirement.
Start from the requirement, not the method, and ask what question it really raises.
Illustrative example: a reach truck driver for a warehouse client. The CV says three years on reach trucks: claimed. Every candidate gets the same questions about the start time, the shifts, and the trucks and lift heights they know: stated. A question about what to check before lifting at height shows what they know: demonstrated, although safe driving still needs a practical check. Dutch rules say that self propelled mobile work equipment is operated by employees with specific expertise for it (Arbobesluit, article 7.17c(1)). The article names no certificate, but a training certificate is a common way to show that expertise, and it can be confirmed with the training provider: verified. Thirteen weeks in, attendance and the supervisor's view show whether the hire worked out: performed. Our blog on what three years on reach trucks really tells you looks closer at that first claim.
| Requirement | The question behind it | Evidence that can answer it | When to collect it |
|---|---|---|---|
| Can start at 05:30 and work the rotating shift pattern | Does the real schedule work for this person? | Stated: describe the actual shifts and ask everyone the same question, such as "How would you travel to the site for a 05:30 start, and how long would it take?" | First screening step |
| Knows basic site safety rules | Do they know what to do? | Demonstrated (knowledge): a scenario question rated against a model answer written in advance | First screening step |
| Holds a valid reach truck certificate | Is it genuine and still valid? | Verified: confirm with the training provider where possible and note the expiry date | Before the first shift, and earlier if your process allows |
| Operates a reach truck safely | Can they actually do it? | Demonstrated (practical), alongside the verified certificate: a short supervised check or a supervised first shift | Before or during the first shift, only where it is safe |
| Allowed to work in the Netherlands | Is it legally true? | Verified: an original identity document, and for workers from outside the EEA and Switzerland, whether they may work in the Netherlands | Before work starts |
| Stays beyond the first 13 weeks | Will they stay? | None before hire; performed evidence shows it | Record after hire and review at 13 weeks |
The right to work row is a legal duty, not a design choice: Dutch government guidance says an employer must check a new worker's identity against the original document before they start work, keep a copy, and check whether workers from outside the EEA and Switzerland may work in the Netherlands (business.gov.nl). Agencies and their clients share these duties differently; see which checks can happen before the offer.
Read the last row twice. In a review of 81 samples, prehire work experience barely predicted job performance (a corrected correlation of 0.06) and did not predict turnover at all (0.00, in the 32 samples that measured it) (Van Iddekinge et al., 2019). Before hire, ask every candidate about the conditions that make this job hard to keep, such as the journey and the shifts, and tell them about the hard parts early, as our blog on the 05:30 start argues. After hire, record what happens.
Our view for staffing agencies. For a returning flex worker, what they did on earlier placements is performed evidence, and it can tell you more than any new CV. Use it only within data protection rules, for a purpose that fits why it was recorded, and with the worker able to see and respond to it; check the details with your data protection adviser.
Why is higher on the ladder not always better?
Because each kind of evidence answers a different question.
It is tempting to read the five kinds as a scale from weak to strong and push every requirement as high as it will go. That wastes time, and it can make a screen worse.
- A verified fact is not a skill. A valid certificate shows that training was passed, not that the person drives safely today, which is why the table pairs it with a practical check.
- For some requirements, the candidate's own word is the best evidence there is. Only the candidate knows whether a 05:30 start fits their childcare and their bus. A structured question with the real start time in it is the right evidence; no document can answer it.
- Demonstration costs time and needs a safe setting. It makes sense for skills needed on day one. Testing skills the site teaches in the first week screens out people who would have learned.
- Demonstration is not automatically fairer. In US job applicant samples, work sample tests showed much larger differences between Black and White applicants than the figure usually cited: a standardised difference of 0.73, against 0.38 (Roth et al., 2008). Every method needs its own fairness check.
- Checking more than the role needs has a cost. Each extra check adds a step, adds personal data you must look after, and gives candidates one more reason to drop out.
The rule is simple: match each requirement to the kind of evidence that can answer it, and stop there. And treat a missing answer or a failed upload as a question for a person, not as a failed requirement.
How does this relate to the Signal Gap?
They are two lenses on one problem, used at different moments.
In What Is the Signal Gap? we described the distance between what a hiring decision rests on and what predicts the outcome, and recorded the evidence held for each requirement at the moment of decision at five levels: verified, demonstrated, structured, claimed and not covered. Those levels audit what a process knew when it said yes. The five kinds of evidence in this article help design the screen before the next candidate applies, and they differ in two ways. What the Signal Gap calls structured is called stated here, because a structured answer is still the candidate's own word until it is shown or checked. And performed is added: evidence that exists only after hire and shows whether the screen worked. A requirement with no evidence at all is still not covered.
What rests on research, and what is our framework?
The building blocks rest on research; the framework as a whole is untested.
| Supported by research | Our interpretation | Not yet tested |
|---|---|---|
| Structured interviews predict job performance better on average than unstructured interviews (Sackett et al., 2022) | Sorting candidate evidence into five kinds, and the names we give them | Whether using the framework as a whole improves hiring outcomes |
| Prehire work experience predicts performance weakly and turnover not at all (Van Iddekinge et al., 2019) | That stated evidence captures requirements a CV leaves out | Whether short structured confirmations predict as well as full structured interviews |
| How well a method predicts varies widely between settings, so local outcome data matters (Sackett et al., 2023) | That each requirement should be matched to the evidence that can answer it, with checks as early as is lawful and proportionate | How any of this performs in Dutch frontline hiring specifically |
The research mostly measures overall job performance, not attendance or staying, so treat the averages as a guide to design and check your own results.
How do you map one role in 45 minutes?
Bring one vacancy and its supervisor, and work through five steps.
Two people are enough: the recruiter or recruitment lead who runs the screen, and the supervisor who knows the work. Bring the job advertisement, the current screening questions if there are any, and the last ten hiring decisions for the role.
- 0 to 10 minutes. List the requirements in the supervisor's words and mark each one as needed on day one or trainable.
- 10 to 20 minutes. For each requirement, note what the current screen collects and where it comes from.
- 20 to 30 minutes. Place each requirement on the ladder: claimed, stated, demonstrated, verified, performed or not covered. Count how many sit at each level.
- 30 to 40 minutes. Pick the most important requirement that is only claimed or not covered. Write one structured question for it and one written criterion for a good answer.
- 40 to 45 minutes. Agree what you will record after hire for this role, and when you will repeat the session.
You leave knowing which requirements your screen proves and which it only assumes, with one question to use from tomorrow. The agenda is our design and has not been tested across organisations. To look back at past decisions instead, our guide to auditing your hiring process for one role in one hour applies the Signal Gap levels to your last ten hires.
Agencies can run it with the client's site supervisor and agree in writing which checks happen before the first shift; employers can run it with the shift supervisor and HR. If a change touches hiring policy or how personal data is handled, involve your works council early.
So how should you screen candidates for frontline roles?
Screen for the question each requirement raises, not for the paperwork.
Back in Venlo, nothing here gives the three recruiters more hours; it changes what their 8.5 minutes per application collect. Read the CV as a list of questions, ask every candidate the same questions about what it leaves out, check the facts the role needs at the moment the law allows, and use a practical task only for skills needed on day one. Record what happens after hire, and keep a person responsible for every decision, with the reason on record.
How Radius helps
Radius Hire is the AI hiring platform for frontline roles, built for teams that have too little time to collect structured evidence from every applicant.
- Stated evidence at volume. The AI Interviewer runs a short, structured voice interview that candidates can take on their phone, whenever it suits them, in the language set for the role. Every applicant gets the same questions, read against the role's requirements. It gathers answers; it does not check documents or observe practical skill.
- Evidence per requirement. For each requirement, Match shows whether it is backed by a document or a check, rests only on what the candidate said, or is still missing, with the moment in the interview that answered it.
- Checks before the offer. Where the employer uses Radius Passport, it checks a candidate's identity and right to work with a trusted verification partner and records certificates with their expiry dates, so a recruiter can review the evidence before an offer. It does not replace the employer's own legal duty to check identity.
- A person decides. Radius never rejects a candidate on its own. A recruiter reviews the evidence, can overrule anything Radius presents and makes every hiring decision.
Radius is new, so this article shows how the product works rather than claiming results. See how it fits warehouse hiring.
FAQ
What is the Proof Ladder?
The Proof Ladder is Radius Hire's framework of five kinds of candidate evidence: claimed, stated, demonstrated, verified and performed, each answering a different question. It is a tool for designing and checking a screen, not a validated scientific scale.
How do you screen candidates without relying on CVs?
Use the CV as a source of questions. Turn each claim that matters into a structured question with a written criterion, check the facts the role needs with whoever issued them, and add a practical check only for skills needed on day one. On the Proof Ladder, that moves important requirements from claimed to stated, demonstrated or verified.
Which checks does a frontline role need before hire?
Only the checks its requirements call for. For most warehouse and logistics roles that means identity and right to work before work starts, plus any certificate or licence the work needs, ideally confirmed with the issuer. Reference checks or a certificate of conduct (VOG) are worth adding only where the role justifies them.
Can you screen for whether a candidate will stay?
Not reliably before hire. In the research, prehire experience does not predict turnover, and no kind of evidence on the Proof Ladder answers "will they stay?" in advance. Ask everyone about the conditions that make the job hard to keep, then record early leaving and review it at 13 weeks.
Do you need software to use the Proof Ladder?
No. A spreadsheet and one working session are enough to start; software matters at volume, when asking every applicant the same questions takes more recruiter hours than the team has. Whatever the tool, a person makes the decision.
General information, not legal advice. Position as of 21 October 2026.
Sources
- Sackett, P. R., Zhang, C., Berry, C. M., and Lievens, F. (2022). Revisiting meta-analytic estimates of validity in personnel selection: Addressing systematic overcorrection for restriction of range. Journal of Applied Psychology, 107(11), 2040 to 2068. https://doi.org/10.1037/apl0000994 (peer reviewed; checked 21 October 2026)
- Van Iddekinge, C. H., Arnold, J. D., Frieder, R. E., and Roth, P. L. (2019). A meta-analysis of the criterion-related validity of prehire work experience. Personnel Psychology, 72(4), 571 to 598. https://doi.org/10.1111/peps.12335 (peer reviewed; checked 21 October 2026)
- Roth, P. L., Bobko, P., McFarland, L. A., and Buster, M. (2008). Work sample tests in personnel selection: A meta-analysis of Black-White differences in overall and exercise scores. Personnel Psychology, 61(3), 637 to 661. https://doi.org/10.1111/j.1744-6570.2008.00125.x (peer reviewed, US samples; checked 21 October 2026)
- Viswesvaran, C., Ones, D. S., and Schmidt, F. L. (1996). Comparative analysis of the reliability of job performance ratings. Journal of Applied Psychology, 81(5), 557 to 574. https://doi.org/10.1037/0021-9010.81.5.557 (peer reviewed; checked 21 October 2026)
- Levashina, J., Hartwell, C. J., Morgeson, F. P., and Campion, M. A. (2014). The structured employment interview: Narrative and quantitative review of the research literature. Personnel Psychology, 67(1), 241 to 293. https://doi.org/10.1111/peps.12052 (peer reviewed; checked 21 October 2026)
- 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. https://doi.org/10.1017/iop.2023.24 (peer reviewed, open access; checked 21 October 2026)
- Arbeidsomstandighedenbesluit (Dutch Working Conditions Decree), article 7.17c, "Gebruik mobiele arbeidsmiddelen". wetten.overheid.nl, version in force from 1 August 2026. https://wetten.overheid.nl/BWBR0008498/ (official; checked 21 October 2026)
- Netherlands Enterprise Agency (RVO), business.gov.nl. "Verifying or filing identity documents" (no publication date shown). https://business.gov.nl/regulations/verifying-identity-documents/ (official guidance; checked 21 October 2026)