Reference checks that actually say something
The reference call is the last serious decision in hiring, and it is usually two minutes of “would you rehire them?” RefScore asks the people who worked with a candidate about behaviour they actually saw, and turns it into a report you can defend.
Referees answered 3.6 out of 5 on average, across 6 references — two clients, two colleagues and two managers. Past this point the curve is flat: a further reference would narrow the estimate slightly and would not add a side to the panel.
Seen from three sides
| Two managers 15 work styles they were positioned to judge · they broadly agreed with each other | A- · Very often | 3.7 |
| Two colleagues 15 work styles they were positioned to judge · they did not agree with each other | B+ · Often | 3.6 |
| Two clients 14 work styles they were positioned to judge · they broadly agreed with each other | B+ · Often | 3.3 |
Where the sides differ
More than a step apart on the frequency scale, and further apart than these sides are in general — so this is not one group simply marking harder than another. People behave differently with different audiences, and people in different roles notice different things. Worth asking about at interview — not a finding on its own.
| Adaptability Both clients said always; both managers said seldom. | A+ vs C+ |
| Cooperation Both managers said always; both clients said seldom. | A+ vs C+ |
All 15 work styles graded, and the 6 deliberately not assessed
What the references show
| Work style | Grade | Seen | Obs. |
|---|---|---|---|
| Attention to Detail Critical for this role A tendency to be detail-oriented, organized, and thorough in completing work. | A+ · Always | 4.8 | 18 |
| Intellectual Curiosity Critical for this role A tendency to seek out and acquire new work-related knowledge and obtain a deep understanding of work-related subjects. | B+ · Often | 3.5 | 10 |
| Cautiousness Critical for this role A tendency to be careful, deliberate, and risk-avoidant when making work-related decisions or doing work. | A- · Very often | 3.8 | 6 |
| Cooperation Critical for this role A tendency to be pleasant, helpful, and willing to assist others at work. | A- · Very often | 3.8 | 6 |
| Perseverance Critical for this role A tendency to exhibit determination and resolve to perform or complete tasks in the face of difficult circumstances or obstacles at work. Both clients could not speak to this. | A- · Very often | 3.8 | 4 |
| Initiative Critical for this role A tendency to be proactive and take on extra responsibilities and tasks that may fall outside of one's required work role. | B- · Occasionally | 2.3 | 6 |
| Dependability Critical for this role A tendency to be reliable, responsible, and consistent in meeting work-related obligations. | A · Almost always | 4.5 | 6 |
| Humility Important for this role A tendency to be modest and humble when interacting with others at work. One of the two clients could not speak to this. | B+ · Often | 3.4 | 5 |
| Integrity Important for this role A tendency to be honest and ethical at work. | A · Almost always | 4.5 | 6 |
| Sincerity Important for this role A tendency to be genuine and sincere in interactions with others at work, without concern for personal gain or self-interest. | A- · Very often | 3.8 | 6 |
| Optimism Important for this role A tendency to exhibit a positive attitude and positive emotions at work, even under difficult circumstances. | B · Sometimes | 3.0 | 6 |
| Empathy Important for this role A tendency to show concern for others and be sensitive to others' needs and feelings at work. | B- · Occasionally | 2.3 | 6 |
| Innovation Important for this role A tendency to be inventive, to be imaginative, and to adopt new perspectives on ways to accomplish work. | B+ · Often | 3.3 | 6 |
| Adaptability Important for this role A tendency to be open to and comfortable with change, new experiences, or ideas at work. | B+ · Often | 3.3 | 6 |
| Tolerance for Ambiguity Important for this role A tendency to be comfortable with ambiguity and uncertainty at work. | B · Sometimes | 3.0 | 6 |
Not assessed
The questionnaire is tailored to Software developers, so it asks about the work styles that matter most for it. These were not covered, and the report says nothing about them either way.
Achievement Orientation · Leadership Orientation · Self-Confidence · Self-Control · Social Orientation · Stress Tolerance
How to read this. Each referee says how often they saw a behaviour, from never to always. A grade is a band of that answer — A+ at 4.6 and above, down to C- at 1.0 — and the overall grade weighs each work style by how much this occupation needs it. There is no failing grade: a reference check reports observed behaviour, not a verdict. Work style definitions from O*NET Web Services, US Department of Labor (O*NET-SOC 2019). Algorithm 3.0.0, T-score 53.0.
The problem
A reference call gives you an impression from one person, shaped by how well they remember and how much they liked the candidate. It is unstructured, unweighted and unrecorded. Two candidates get asked different questions, and nobody can say afterwards why one came out ahead.
How it works
You start a check
Pick the occupation you are hiring for. One price and one panel: six references for €79. You pay for it, not the candidate.
The candidate agrees and names their referees
Nothing is sent to anyone until they do. They name two people from each of three kinds of working relationship — two managers, two colleagues and two clients, say — and can withdraw at any point.
Referees answer, and you get a report
No accounts, no phone tag. Each referee gets a questionnaire tailored to the role and to how they worked with the candidate.
Why the score holds up
- Questions come from O*NET, the occupational taxonomy published by the US Department of Labor — up to 21 work styles, chosen for the role you are recruiting. 423 occupations are covered, from accountants and software developers to shop supervisors, electricians and welders.
- Referees rate observed behaviour, never opinion. There is no free-text box for “anything else we should know”, which is where reference checks usually go wrong.
- Each referee is scored on their own answers before those are combined, because a manager, a colleague and a client stood in different places and saw different things. None of them counts for more than the others. How long they worked with the candidate, and how recently, does still count.
- Referees come in pairs. One manager is one person’s opinion; two who saw the same work from the same place can be reported as a side of the story — and where two sides disagree, the report can tell you it was the vantage point that differed rather than the temperament. That is what the 2×3 panel is for, and why it is fixed rather than chosen.
- Nothing is scored on thin evidence. A work style is graded only where at least two referees could speak to it, and no score appears at all until at least three referees have answered. A referee who was not in a position to see something says so, and it is listed as not assessed rather than guessed at.
- Results are letter grades, banded by how often referees said they saw each behaviour — from always down to never. No pass mark and no failing grade: a reference check reports observed behaviour, not a verdict.
What it is not
It is not a background check, a credit check, or an identity check. It does not follow a candidate between applications, and there is no profile that other employers can look up. Each check is bought by one employer, for one role, and the candidate has to agree to it.
Common questions
What is a structured reference check?
A structured reference check asks every referee the same predetermined questions about behaviour they personally observed, then scores the answers against a consistent scale. It replaces the open-ended reference call, where two candidates get asked different questions and nobody can say afterwards why one came out ahead.
How is RefScore different from calling a referee?
A phone call produces one person’s impression, shaped by how well they remember and how much they liked the candidate. RefScore asks each referee a questionnaire chosen for the occupation and for what that referee was positioned to observe, scores each of them on their own answers, and then shows you where a manager, a colleague and a client described the same person differently — with the evidence behind every grade.
Does the candidate have to agree?
Yes. Nothing is sent to any referee until the candidate has agreed and named the referees themselves. They can withdraw at any point, which stops collection immediately. Consent is the lawful basis for contacting a named person’s former colleagues, and it is enforced in the software rather than promised in a policy.
Who chooses the referees?
The candidate does, within a fixed shape. Every check is six references: two people from each of three kinds of working relationship — two managers, two colleagues and two clients, say, or two people who reported to them if that fits the work better. The candidate picks the people; the shape is the same on every check. They can replace a referee who declines or goes quiet, up to three times, but not one who has already answered, and the replacement stands in the same position.
Why do referees have to come in pairs?
Because one referee of a kind cannot be told apart from their own opinion. With a single manager, a difference between what the managers said and what the clients said might be the vantage point or might be that particular manager, and no amount of presentation can separate the two. With two of each, the report can say which. It is the difference between “worth asking about” and a number on a page.
How many references are needed for a score?
A check is six references, and three is the floor below which no score is produced at all — there is not enough independent observation to be defensible. Between the two the check still delivers, at whatever sides of the panel actually completed, and the rest is refunded automatically. Below three the employer is refunded in full, because there is no product to have sold.
Where do the questions come from?
From O*NET, the occupational taxonomy published by the US Department of Labor. Each occupation carries work-style importance ratings, and the questionnaire is built from the work styles that matter most for the role being recruited. 423 occupations are covered, from accountants and software developers to shop supervisors, electricians and welders.
Is RefScore a background check?
No. It is not a background check, a credit check, or an identity check. It does not follow a candidate between applications and there is no profile other employers can look up. Each check is bought by one employer, for one role, and expires with it.
What does a reference check cost?
€79 for one candidate — six references, in three pairs. One price and no subscription, and you pay for what is delivered: if a pair arrives one person short, that side of the panel is refunded automatically. Below three completed references there is no score and the whole amount goes back.
Join the closed alpha
RefScore is being tested with a small number of employers, across whatever roles they are actually hiring for. It is hands-on and we want to hear where it fails. If that sounds useful, tell us about your hiring and we will say whether it is a fit.
Open HR SASU · RCS Paris 994 927 242 · 60 rue François Ier, 75008 Paris. Work style definitions from O*NET, US Department of Labor.