Verify candidate credibility with background checks and references

A background check verifies facts: identity, criminal record, education, employment dates, licenses. A reference check collects opinion from people who watched the candidate work. The two answer different questions, and neither one tells you whether this person can do the job you are hiring for. That question needs its own evidence.

Plenty of teams run both because someone said they should, then treat the results as paperwork. That is the expensive way to do it. Verification only earns the days it adds to your process if you know which claim each check settles, who is allowed to run it, and what the law lets you do with the answer.

TL;DR

• A background check confirms facts that a third party keeps records of. A reference check collects judgment from people who worked with the candidate. Do not ask either one to do the other's job.
• If an outside company runs the check for you, federal credit-reporting rules apply: stand-alone written notice, written permission, and a two-step adverse action process before and after you turn anyone down.
• Those rules catch reference checks too. A firm that interviews people about a candidate's character is producing an investigative consumer report, which carries a disclosure your own phone call does not.
• References are poor predictors of performance. A 2022 re-analysis of selection research still carries reference checks at 0.26 and could not even re-estimate them, against 0.42 for a structured interview.
• Identity has become part of verification rather than a formality. Federal investigators have been warning about stolen identities and deepfaked interviews since 2022.
• Settle claims with records, judge capability with work. Mixing those two up is how a clean file turns into a bad hire.

What is a background check vs a reference check?

A background check is a records search. It confirms that a person is who they say they are and that the facts on their application match an official source. A reference check is a conversation with someone who worked with them, and it produces opinion, not proof. One protects you from a false claim. The other tells you what the claim felt like in practice.

Here is the split, side by side.

Notice the last row. Both failures come from the same habit: asking a check for a verdict instead of for information.

Why does credibility matter in recruitment?

Because the cheapest lie in hiring is the one nobody checks. A candidate who inflates a job title costs you a mismatch, which is the ordinary end of employment fraud in hiring. A candidate who is not the person on the screen costs you access to your systems, your customer data, and your payroll.

That second case stopped being hypothetical years ago. The FBI has warned employers since 2022 that people apply for remote IT roles using stolen personal information and deepfaked video interviews, and its advisory tells hiring teams exactly what to watch: lip movement that does not quite track the audio, a cough or a sneeze the picture does not match. A records check comes back clean in that situation, because the records are real. They just belong to somebody else.

For a 40-person company that hires eight or nine times a year, the practical read is simple. Identity verification is now the first check, not the last, and it belongs with the people who already handle screening rather than in a security review nobody scheduled. Get it wrong and the other checks verify a stranger's record very thoroughly.

What does a candidate verifier actually check?

A candidate verifier, whether that is a person on your team or a screening vendor, works claim by claim. Every line on an application, and every claim on the resume behind it, is either something a record can settle, something only a human can describe, or something you have to watch the candidate do. Sorting them is most of the job.

Read the right-hand column and the gap jumps out. Five of six claims are settled by somebody else's records or memory. Only the last one is settled by anything you control, which is why teams that verify hard and assess casually still hire the wrong person.

How do you run a background check legally?

If a third party produces the report, you are using a consumer report and the Fair Credit Reporting Act sets the sequence. The federal rules are specific about the order, and skipping a step is the most common way a well-meaning process turns into a claim.

1. Give written notice in a stand-alone document. Not a paragraph buried in the application form. Federal guidance for employers is explicit that the notice must be in writing and in a stand-alone format.
2. Get written permission. Ask once, in plain language, and keep the record.
3. Say so if you are ordering an investigative report. Where the report involves personal interviews about character or reputation, you must also tell the candidate about their right to a description of the nature and scope of the investigation.
4. Certify to the screening company that you gave notice, met the requirements, and will not use the information to discriminate.
5. Run the pre-adverse action step. Before you reject anyone on the strength of a report, send them a copy of the report and a copy of the summary of rights under the Fair Credit Reporting Act.
6. Send the adverse action notice after you decide, naming the screening company and telling the candidate they can dispute the report and get another free copy within 60 days.

Then there is the harder part, which is what you do with a record once you have one. Equal employment guidance sets out three factors for weighing criminal conduct: the nature and gravity of the offence, the time that has passed since it or since the sentence ended, and the nature of the job. The same guidance is blunt that an exclusion based on an arrest alone is not job related and consistent with business necessity, and it asks employers to give the person a chance to explain their circumstances before the door closes.

In practice that means writing down, before you order anything, which records would actually matter for this role and why. A cash-handling role and a warehouse role are not the same risk, and a screening package sold as one size is not a policy.

How do you run a reference check that works?

A reference call fails in a predictable way: you ask whether the candidate was good, the referee says yes, and you have learned nothing. Ask for descriptions instead of verdicts and the same 20 minutes turns into evidence. The same discipline runs through how to structure the questions themselves.

1. Ask for the right people, not the available ones. The manager who signed off their work beats a friendly peer. Say so when you request references.
2. Verify the referee before you verify the candidate. A company email address, a real phone line, a profile that matches the role they claim. Made-up referees prefer email and avoid live conversations.
3. Anchor every question to something specific. "Walk me through the last project they owned end to end" gets you a story. "Were they a team player" gets you a syllable.
4. Ask the same core questions for every candidate in the role. Consistency is what makes the answers comparable, and it is also what makes your process defensible later.
5. Ask about fit with the work, not fit with the people. "What kind of environment did they do their best work in" is useful. "Would they fit our culture" invites bias and tells you nothing you can act on.
6. Write it down the same day. Short notes, quotes where you have them, filed with the candidate's other evidence.

Pro Tip: ask every referee one closing question, "who else saw this work closely?" The names you were not given are usually the ones worth calling, and a candidate who has stacked the list will hesitate right there.

One caveat, because this gets oversold. References are a slow, low-yield check. Two calls is often the honest limit for a role below director level, and a third adds less than the hour it costs. Spend the hour on the work sample instead.

Should you use a third party reference check company?

Sometimes, and the decision is less about quality than about volume and law. A reference-checking company sends a structured questionnaire, chases the referee so your recruiter does not have to, and gives you comparable answers across candidates. That is worth real money at 50 hires a quarter. At five, it is usually not.

The part that surprises people is legal. When an outside firm gathers information about a candidate's character, general reputation or personal traits by interviewing people who know them, that report is an investigative consumer report under federal credit-reporting rules. Everything in the sequence above then applies to your reference checks, including the stand-alone notice, the consent, and both adverse action steps. Make the calls yourself and those particular obligations do not attach.

So the real question is not "are vendor references better". It is whether the consistency you buy is worth the compliance you take on. Small teams hiring in single digits per quarter usually get more from a hiring manager making two well-prepared calls. Teams hiring at volume, or in regulated work, usually get more from a vendor with a documented process and a record of who said what.

What checks cannot tell you about a candidate?

Whether the person can do the job. That is not a dig at verification, it is what the research shows. When selection-science estimates were re-examined in a 2022 paper in the Journal of Applied Psychology, most methods dropped by 0.10 to 0.20 once range-restriction corrections were fixed. Structured interviews came out top at 0.42, job knowledge tests at 0.40, work samples at 0.33 and cognitive ability tests at 0.31. Reference checks were not even re-estimable from the available data and still carry the older 0.26.

Read that as a job description, not a verdict. A reference check is a verification instrument. It is excellent at catching a claim that does not survive contact with the person who was there, and mediocre at forecasting performance. Asking it to rank your shortlist is asking the wrong tool.

This is where the Testlify Multi-Signal Talent Evaluation Model earns its keep. It combines several role-relevant signals, assessments, interviews, simulations, references and reviewer feedback, rather than leaning on one. Three different questions, three different sources: records answer "is this true", references answer "what was it like to work with them", and a scored assessment answers "can they do this". A candidate should move forward when those point the same way.

Where Testlify fits, stated plainly: it is not a background-check provider and does not search court or credit records. It integrates with the tools that do. What it does run is the evidence layer around them. References are collected as a linked reference assessment rather than a free-text form, triggered immediately, on completion, at a score threshold or at a pipeline stage, with minimum and maximum reference counts and automatic invites, so the answers arrive structured and comparable. Identity sits in the same place: photo ID verification with a face match against the ID, face detection during the session, and face-verification data deleted after 30 days with consent the candidate can withdraw. Proctoring flags are treated as evidence for a human to review, not an automatic rejection.

Hire on evidence you can check

Verification tells you the person is real and the history is true. An assessment tells you what they can do on Tuesday. Run both, in the order a sane vetting sequence implies, and the shortlist you hand your hiring manager is one you can defend line by line. Book a demo to see how reference assessments, identity checks and role-based scoring sit in one workflow, or start free and build the assessment for your next open role.

Key takeaways

• Match the check to the claim. Records settle facts, references describe behaviour, assessments show capability. When one check is asked to cover all three, the weakest evidence quietly decides the hire, and nobody notices until month three.
• The vendor decision is a legal decision. The moment an outside firm produces the report, the stand-alone notice, written consent and both adverse action steps apply, including for reference checks gathered by interview. Decide who runs the check before you decide what to check.
• Write your criminal-records policy before you order a report. Nature of the offence, time elapsed, nature of the job, applied the same way for every candidate in the role. A blanket rule is the version that gets challenged.
• References are verification, not prediction. At 0.26 they sit well below a structured interview at 0.42, so use them to test claims and catch contradictions, not to rank a shortlist.
• Verify the referee, not just the reference. Company email, live conversation, a role that matches what the candidate described. Fake referees fail all three tests and prefer to stay in writing.
• Identity is the first step now, not the last. Federal advisories describe candidates applying with stolen personal data and altered video, which is the one failure mode a records check cannot catch. Confirm the person in the interview is the person on the application before you spend money verifying anything else.
• Two calls, then stop. A third reference rarely changes a decision. Spend that hour on a work sample, which is the one signal you control and the one that actually predicts the work.

FAQs