Policy

Off-List Reference Calls Produce Evidence the Candidate Cannot Answer

Calling a reference the candidate did not list is mostly lawful in the United States, and that is the wrong question to be arguing about. What matters is what happens to what you hear: an off-list comment enters the decision as a judgment the candidate never sees and cannot answer, which is the structure that makes a hiring decision impossible to audit later, whether or not the comment was true. Set one rule instead. Nothing enters the decision the candidate has not had a chance to answer.

The takeThe ethics argument on both sides is a distraction, because the two camps are arguing about the phone call while the damage sits downstream of it. A comment you cannot repeat to the candidate is a comment you cannot act on, and that single test resolves nearly every case without anybody having to settle whether backchannelling is honorable. It has the further merit of being checkable by somebody else six months later, which no argument about discretion has ever been.

Where Olive fits

Open a role and see what the work shows

The candidate is granted the identical report the employer reads, which is the opposite arrangement to an off-list call. Each of Olive's six findings carries the timestamped excerpt it rests on, so a candidate can argue with a finding rather than with an impression nobody ever told them about.

Rank your shortlist

Is an off-list reference call allowed?

As of 2026, a hiring manager in the United States calling a mutual contact is generally lawful, and the Fair Credit Reporting Act, which governs vendor background checks, does not reach a call you place yourself: its rules attach to reports furnished by an agency that regularly assembles information on people for that purpose 1. Legality is also the least interesting thing about the practice.

The two genres of advice online cancel each other out. Executive-search and private-equity talent blogs call it sensible diligence and spend the article on how to be discreet. Recruiting firms call it a breach of the candidate's trust and stop there. Both are arguing about whether to dial, and neither says what happens to the information afterwards, which is the part that determines whether anybody was harmed.

Start by splitting the term, because two unlike things share it. The first is passive: a mutual contact mentions something unprompted at dinner. You cannot un-hear that and pretending otherwise is theater. The second is active: you work the network to find people the candidate deliberately did not name. Only the second is a decision, and it is the one a policy has to cover.

The active version is tempting for a real reason. The listed references are chosen by the candidate, so the call is bounded by design. That limit is real, and the answer to it is question design rather than a wider net: asking every reference for a bounded episode gets more out of a friendly source than a general question does, and it does it on the record.

Which off-list calls turn into a regulated report?

The ones you hand to a firm that does this for a living. The Fair Credit Reporting Act defines an investigative consumer report as a consumer report, or part of one, in which information on a person's character, general reputation, personal characteristics or mode of living is obtained through personal interviews with neighbors, friends or associates 1. That describes a backchannel check, and when a firm like that prepares one for you, a separate set of duties attaches to it.

Those duties are specific. A person may not procure or cause to be prepared an investigative consumer report unless it is clearly and accurately disclosed to the individual, in writing mailed or otherwise delivered not later than three days after the report was first requested, that such a report may be made, together with a statement of the right to request more detail and of the consumer's statutory summary of rights 2. On a written request from the candidate, made within a reasonable time of that disclosure, the employer must then make a complete and accurate disclosure of the nature and scope of the investigation, within five days of the request or of the report first being requested, whichever is later 2.

Whether any particular engagement lands there depends on whether the firm doing it meets the statutory definition of a consumer reporting agency, which turns on what that firm regularly does rather than on what your contract calls it 1. That is a question for counsel, and it is a cheaper question to ask before a retained search begins than after a candidate's lawyer asks it for you.

A call you make yourself is unregulated and entirely on you. A call a firm makes for you may carry a written disclosure duty that removes the discretion the delegation was supposed to buy. Which of the two you are running is decided at the briefing, before anyone has picked up a phone.

What do you do with what you hear?

Put it to the candidate or discard it. Those are the only two honest endings, and choosing between them takes a minute: a comment specific enough to matter is specific enough to describe without naming its source, and a comment too vague to put to somebody was never evidence. What must not happen is the third ending, where it becomes a quiet reason nobody writes down.

The reason to be strict about this is not only fairness. An off-list comment arrives late, from a source of unknown reliability, and gets combined with everything else by judgment rather than by a stated rule, which is precisely the method the evidence treats as weaker. Kuncel and colleagues classify group consensus meetings and individual expert judgment together as holistic combination, defined by data being combined through judgment or intuition rather than an algorithm applied the same way each time, and calculate that the lower validity of holistic combination can mean a 25% reduction in correct hiring decisions at a selection ratio of .30 3. That figure is derived from the meta-analytic validities and moves with the selection ratio you assume; no employer measured it, and it is not an argument against holding a debrief.

There is a sharper illustration of the same shape. Highhouse reproduces Sarbin's 1943 result in which high school rank plus a college aptitude test correlated .45 with academic achievement, while the same two predictors plus counselors' intuitive judgment correlated .35 4. An admissions cohort from 1939 does not transfer to your loop. What it does is make the mechanism visible: a late human overlay on structured evidence can lower accuracy, and an off-list call arrives last, with the least structure behind it.

So the file has to say where every judgment came from. The same standard applies to a machine-written characterization of a candidate, and the reasoning does not change with the source: whether an unattributed impression belongs in the hiring file is one question, asked about two very different narrators.

Never call anyone at a current employer without written permission

Get written permission before contacting anybody at a company the candidate still works for, and put that in the process. One call can end the job somebody currently holds, at a moment when they have no offer from you. The discretion advice on the search-firm side treats that as a tradecraft problem. It is a harm the employer is choosing to risk on somebody else's behalf.

Deciding case by case fails in a predictable direction. The exception is never requested for the candidate everybody likes; it is requested for the one somebody already doubts, which means the extra scrutiny lands exactly where a written policy would have caught it. Same rule for every finalist on a requisition, or no rule at all.

There is also evidence that free-text commentary about a person is not neutral. Coding 624 authentic letters of recommendation, Madera and colleagues found 54% of letters written for women contained at least one doubt raiser against 51% for men, with the gap widening on letters carrying two or more, and a companion experiment found that hedging and negativity in a letter lowered ratings of research competence 5. Those are academic letters, the difference in the headline percentages is small, and male and female letter writers did not differ, so this is not a finding about biased individuals. What it shows is that free-text commentary carries systematic differences before any reader gets to it, and an off-list call comes with no request wording at all to shape it.

The hiring policy can carry this in a single line: whether off-list calls happen, who may make them, and what happens to what they produce. Write it before the search that raises it. And if a candidate is turned down partly on something you heard sideways, they are owed the same account as anyone else who asks why the decision went the way it did, which is another reason not to collect what you cannot repeat.

Read the evidence

Common questions

Is a backchannel reference check legal?

Yes, with limits. A call an employer places itself to a mutual contact is generally lawful in the United States, and the federal law that governs vendor background checks does not reach it. That changes when you engage a firm to gather opinions about a candidate, which can meet the statutory definition of an investigative consumer report and bring written disclosure duties with it. State law varies on what a reference giver can be held to for what they say, and that is a question worth putting to counsel before a retained search rather than after. Lawful is also a low bar for a practice this one-sided.

What if a mutual contact volunteers something without being asked?

You heard it, so treat it as heard and apply the rule you would apply to a call you placed yourself. If it is specific and material, describe the substance to the candidate without naming the person and give them a real chance to answer. If it is a vague impression, it was never evidence and does not belong in the debrief. The one outcome to refuse is letting it sit unspoken as the reason a decision goes the other way, because nobody can examine a reason that was never stated, including you six months later.

Should you tell the candidate you spoke to somebody off their list?

Tell them what you heard, at minimum, and in most cases that means telling them you spoke to someone. Naming the source is a separate judgment and often not yours to make, since people speak candidly on the understanding that they will not be quoted. The workable middle is to put the substance to the candidate plainly, say it came from someone who worked with them, and let them respond. If you would not be willing to say even that much, you have found the comment you should not be using.

Is this different for executive hires?

The practice is more common and the reasoning is not different. Senior searches involve smaller networks, higher stakes and a retained firm with its own relationships, which is exactly the combination that produces a lot of unattributed judgment about a small number of people. It is also where an engaged firm is most likely to be gathering opinions on your behalf, which is the version that can carry disclosure duties. The seniority argues for writing the rule down, not for suspending it.

Does an off-list call belong in the hiring record?

If it influenced the decision, yes, in the same file as everything else that did. A judgment that shaped an outcome and appears nowhere in the record is the hardest kind of decision to defend later, because there is nothing to point at and nothing to correct. Record what was said, what you put to the candidate and how they answered. If your instinct is that writing it down would be awkward, that instinct is telling you the comment should have been discarded rather than kept off the page.

References

  1. 1. 15 U.S. Code Sec. 1681a - Definitions; rules of construction Cornell Law School, Legal Information Institute (United States Code), 2018. law.cornell.edu Supports the definition of an investigative consumer report at (e), gathered through personal interviews with neighbors, friends or associates, and the definition of a consumer reporting agency at (f), which is why a call an employer places itself is outside these rules.
  2. 2. 15 U.S. Code Sec. 1681d - Disclosure of investigative consumer reports Cornell Law School, Legal Information Institute (United States Code), 2018. law.cornell.edu Supports the written disclosure duty within three days of the report first being requested, and the duty on written request to disclose the nature and scope of the investigation within five days.
  3. 3. Mechanical Versus Clinical Data Combination in Selection and Admissions Decisions: A Meta-Analysis Journal of Applied Psychology (American Psychological Association), 98(6), 1060-1072, 2013. gwern.net Supports the classification of judgment-based combination as holistic, and the derived 25% reduction in correct hiring decisions at a selection ratio of .30.
  4. 4. Stubborn Reliance on Intuition and Subjectivity in Employee Selection Industrial and Organizational Psychology, 1(3), 333-342, Table 1 (Scott Highhouse), 2008. edbatista.com Supports the Sarbin 1943 comparison in which adding counselors' intuitive judgment to two mechanical predictors moved the correlation with the criterion from .45 to .35.
  5. 5. Raising Doubt in Letters of Recommendation for Academia: Gender Differences and Their Impact Journal of Business and Psychology, 34, 287-303 (Madera, Hebl, Dial, Martin and Valian), 2019. s3.wp.wsu.edu Supports the claim that volunteered commentary carries systematic differences before any reader sees it: 54% of letters for women contained at least one doubt raiser against 51% for men across 624 letters, with hedging lowering competence ratings in the companion experiment.

5 sources, numbered by first appearance. How Olive sources claims

General guidance, not legal advice. Hiring rules differ by state and country and change often; check anything here against your own counsel before you act on it.

Olive assesses how a person works with AI. It does not detect AI-written documents, and it never produces a score, a ranking, or a match percentage for a person. Candidates read the same report the employer reads.

Back to answers

Open your first role Ten attempts a month against a live item bank, with a human-written report on every one.