Reference checks are one of the most familiar parts of recruitment. They usually happen near the end of the process, after the interviews have finished and everyone is starting to feel confident about the preferred candidate and then someone calls the referees.
If the feedback is glowing, everyone relaxes a little more. The candidate must be good because their previous manager just told us so.
The problem is that reference checks are not nearly as reliable as they feel.
One of the most influential reviews of personnel selection research, conducted by Frank
Schmidt and John Hunter in 1998, reported a validity coefficient of approximately .26 for reference checks, substantially below several other selection methods. Their broader research helped establish an important principle in recruitment: different sources of candidate information vary considerably in their ability to predict future job performance.
More recent research by Sackett and colleagues has revised a number of the older validity estimates after identifying problems with the way range restriction had historically been corrected. The precise numbers have changed, which is exactly why old validity coefficients should not be treated as eternal truths. The broader lesson, however, remains: we need to think carefully about the quality of each piece of evidence in a selection decision rather than assuming all information deserves equal weight.
Reference checks have some obvious weaknesses. The candidate chooses the referees, so unsurprisingly they tend to choose people they expect will speak positively about them. The referee may also be busy, distracted or reluctant to say anything that could harm someone’s employment prospects.
Even when the referee is engaged and trying to be completely honest, another problem remains: context.
Someone can be an excellent employee in one workplace and struggle badly in another. Their relationship with one manager may have been outstanding, while their relationship with a very different manager may be far less successful.
This analogous to a movie recommendation. I watched Once Upon a Time in Hollywood with my two adult sons. Same movie, same actors and exactly the same information in front of us. I thought it was boring, but they loved it, yet we were both right: we simply responded differently to the same thing.
Work relationships can operate in much the same way. A referee is telling you how the candidate performed within their role, workplace culture and working relationship. That information can be useful, but it cannot simply be transferred wholesale into another organisation.
This does not means we should abandon reference checks, but instead the better approach is a,
how do we use them properly?
A reference check should not be used to discover the candidate from scratch. By this stage of a good recruitment process, you should already have information from the CV, screening conversations, psychometric testing, structured interviews and incidental behaviours observed throughout the process. Therefore, the reference check should then be used to
confirm, challenge or add context to that existing picture.
If testing suggests someone is highly conscientious, ask the referee about reliability, organisation and follow through. If the interview raised a question about conflict management, ask how the person handled disagreement. If the candidate describes themselves as highly autonomous, ask whether they demonstrated initiative or generally preferred clear systems and direction. That is much more useful than simply asking, “What are their strengths and weaknesses?”
Structure matters as well. Recent
research published in 2026 found that structured online reference checks have some predictive validity but also demonstrated significant leniency bias in referee ratings. Interestingly, the researchers found that asking referees to take the perspective of the hiring manager rather than the candidate reduced some of that leniency.
That is a useful reminder that reference checks are not inherently useless. Poorly designed reference checks are weak. Better designed reference checks improve the evidence, but they do not eliminate error, which brings us back to one of the central principles of good recruitment: every source of candidate information contains some error.
The aim is not to find one perfect source of information because it does not exist. The aim is to gather multiple relevant sources of evidence and look for a “convergence of data”. If the CV, screening process, testing, interview and referee feedback are all pointing broadly in the same direction, confidence in the hiring decision increases.
If the referee says something that conflicts with the rest of the evidence, do not automatically panic or automatically dismiss it, but investigate it. Reference checks should be part of the recruitment process, but they should rarely determine the decision.
The real skill lies in understanding
how much weight each piece of evidence deserves.
And sometimes the information that feels the most reassuring is not necessarily the information we should trust the most.
Summary
Reference checks remain a useful part of recruitment, but they are affected by candidate nominated referees, context, relationship quality and leniency bias. Rather than treating a glowing reference as confirmation that the hiring decision is correct, organisations should use structured, job relevant questions to confirm, challenge or add context to information already gathered through screening, psychometric testing and interviews. The goal is not perfect evidence, but convergence across multiple sources.
Better recruitment decisions come from better evidence, not greater confidence. Fermion helps organisations combine psychometric testing with structured recruitment processes so hiring decisions are based on multiple sources of relevant information.
The focus of Fermion is
"Psychometric Testing for Recruitment" and
“Recruitment to Retention: How to Select Good Staff & Keep Them”. If you would like to learn how to select good staff and keep them, please
contact us at Fermion.
“Learn from the mistakes of others. You can’t live long enough to make them all yourself.”
Eleanor Roosevelt.
NB: I use ChatGPT as a writing and editing tool, but the ideas, judgement and final views in this article are my own.
