Credibility Assessment in Workplace Investigations: How to Weigh Conflicting Accounts

What a Credibility Assessment Is, and What It Is Not

A credibility assessment is a reasoned judgment about which account of a disputed fact is more likely accurate, based on evidence you can point to. It is account by account and fact by fact. A witness can be reliable about what happened in a March meeting and unreliable about a date in January.

It is not a judgment of character, and it is not lie detection. You are not deciding who is a good person or who seemed nervous. Investigators are not polygraphs, and nothing in the published factors below asks you to read eye contact, posture, or fidgeting. People who are telling the truth are often anxious in an investigation interview; people who are not can be perfectly calm.

If the terminology is new, our Glossary of Workplace Investigation Terms defines the terms used here.

The Standard of Proof: Preponderance of the Evidence

Workplace investigators generally apply the preponderance of the evidence standard: is it more likely than not that the alleged conduct occurred? That is the standard used in civil employment litigation, and it is the one most employer policies and investigator training programs adopt. It is far lower than the criminal standard of proof beyond a reasonable doubt.

Two practical consequences follow:

• You do not need a confession, a witness, or a document for every finding. If one account is more believable than the other for reasons you can articulate, that can carry the finding.
• "Inconclusive" is a legitimate result, but it must be earned. It is the right answer only after you have run down the available evidence and the accounts remain in genuine balance. It is not a default for uncomfortable cases.

No federal statute dictates the standard a private employer must use in its own internal investigation. Check your policy, any collective bargaining agreement, and state law, because some of them set a different or more specific standard.

The EEOC Credibility Factors

The most widely used list comes from the Equal Employment Opportunity Commission. In its 1999 enforcement guidance on employer liability for harassment by supervisors, the EEOC said that when there are conflicting versions of relevant events, the employer will have to weigh each party's credibility, and it listed five factors:

The same guidance added an instruction that matters as much as the list: none of the factors is determinative. It gave two examples. The fact that there are no eyewitnesses does not defeat the complainant's credibility, because harassment often happens behind closed doors. And the fact that the accused engaged in similar behavior in the past does not by itself prove the current allegation.

Where that guidance stands today

Be accurate about the source when you cite it. The 1999 document is marked on the EEOC's website as superseded on April 29, 2024 by the Enforcement Guidance on Harassment in the Workplace. The Commission then voted 2-1 on January 22, 2026 to rescind the 2024 guidance in its entirety. The EEOC's announcement states that federal employment laws against discrimination, harassment, and retaliation, and the Supreme Court precedent interpreting them, remain in place. Verify whether the EEOC has issued replacement harassment guidance since January 2026 by checking eeoc.gov/laws/guidance.

So the five factors are not binding law, and they were never a regulation. They remain a sound, recognizable framework, and they track the way judges and juries evaluate testimony. A separate EEOC document, Promising Practices for Preventing Harassment, still advises that those who handle complaints "use guidelines to weigh the credibility of all relevant parties" and prepare a written report. If your report says you applied the factors the EEOC first published in 1999, describe them that way rather than as current enforcement guidance.

Using Each Factor Without Overreaching

Inherent plausibility

Test each account against facts that are not in dispute. If the accused says he was never alone with the complainant, and the schedule shows the two of them closed the store together eleven times in the period, his account has a problem. If the complainant describes a conversation in an office that the accused did not occupy until two months later, hers does.

Plausibility is not the same as what you would have done. "A person who was really harassed would have reported sooner" is an assumption, not a finding. Delay has many explanations, and you should ask about it rather than infer from it.

Demeanor

The EEOC listed demeanor, so you will see it in every template. Treat it as the weakest factor. What you can legitimately record is observable and specific: the witness declined to answer a direct question three times, changed the answer after being shown a document, or gave a detailed narrative on some points and "I don't recall" on every point that mattered. Those are facts about the interview.

What you should not record is an impression dressed up as evidence: "seemed evasive," "wouldn't make eye contact," "appeared rehearsed." Cultural background, disability, language, trauma, and ordinary fear all affect how people present. A finding that rests on demeanor alone is the one most likely to be challenged, and the hardest to defend.

Motive to falsify

Everyone in an investigation has an interest. The accused has an obvious one; that does not make a denial false. A complainant who was recently disciplined has one too; that does not make the complaint false, and treating it as presumptively retaliatory is a common and costly error. Ask what each person gains or loses, then ask whether the evidence supports the idea that the motive actually shaped the account. Give more weight to witnesses with nothing at stake, and to statements against the speaker's own interest.

Corroboration

Corroboration is usually where "he said, she said" cases are decided. Look beyond eyewitnesses:

• Messages, emails, and chat logs sent before anyone knew there would be an investigation
• Contemporaneous disclosure: who the complainant told, when, and whether their recollection matches
• System records such as badge swipes, schedules, call logs, and expense reports
• Changes in behavior that others noticed at the time, such as a transfer request or avoiding a shift

Preserve this material early. Our guide to Evidence Collection and Storage in Investigations covers handling and chain of custody.

Past record

Prior similar complaints against the accused are relevant, including ones that were not substantiated, but weigh them for what they are. Three unrelated employees describing the same specific behavior over four years is meaningful. One vague prior complaint is not. Apply the factor to everyone: a witness with a documented record of dishonesty in a previous matter carries that into this one.

Other Indicators Experienced Investigators Use

These are not on the EEOC's list. They are common in investigator practice and are consistent with it.

• Internal consistency. Does the account hold together from start to finish?
• Consistency over time. Compare the written complaint, the first interview, and any follow-up. Minor variation is normal. A changed core fact needs an explanation.
• Consistency with other evidence. An account that fits the documents is stronger than one that requires three other people to be mistaken.
• Level of detail. Specific, sensory, sequenced detail tends to support reliability, with allowance for the passage of time.
• Opportunity to observe. Was the witness actually positioned to see or hear what they describe, or are they repeating what someone told them?

Good interviews produce the raw material for all of these. See Workplace Investigation Interviews: Questions and Technique and the advice on capturing the original allegation in Documenting the Statement of Complaint.

A Worked Example

The names and facts below are fictional and for illustration only.

Allegation. Dana, a shift lead, reports that her manager, Mr. Reyes, told her during a closing shift on the 14th that her schedule "would get a lot better" if she went out with him. Mr. Reyes denies it and says Dana is angry about a written warning he issued on the 20th.

Evidence gathered. The schedule shows both worked the closing shift on the 14th. A coworker, Priya, says Dana called her that night, upset, and described the comment; Priya's phone log shows a nine-minute call at 10:42 p.m. on the 14th. Dana's schedule changed to fewer hours starting the 17th. A former employee states in a prior exit interview that Mr. Reyes made "comments about dating" to her. Mr. Reyes first told the investigator he did not work on the 14th, then said he had forgotten.

Analysis. Plausibility: the opportunity existed. Corroboration: a contemporaneous disclosure, supported by a phone record, six days before the warning. Motive: the warning could give Dana a reason to retaliate, but the disclosure to Priya predates it, which undercuts that theory. Past record: one similar prior report. Consistency: Mr. Reyes changed his account on an undisputed fact when shown the schedule.

Finding. It is more likely than not that the comment was made as alleged. Note what the finding does not rely on: anyone's facial expression.

How to Write the Credibility Finding

A credibility determination that lives only in your head does not exist. In the report:

1. State each disputed fact separately.
2. Summarize each person's account of that fact.
3. Identify the evidence bearing on it.
4. Explain, factor by factor, why you credited one account over the other.
5. State the finding and the standard: "more likely than not."

Write "I found Ms. A's account of the March 3 meeting more credible because..." and not "Mr. B was lying." Keep legal conclusions out; you are finding facts and policy violations, not deciding whether the law was broken. For structure and sample language, see our investigation report template and worked sample, and for where this step sits in the overall sequence, the HR investigation process, step by step.

Credibility work is a skill that improves with supervised practice. The Internal Investigations Certificate Program lists evaluating witness credibility among its learning outcomes and uses interactive exercises built on workplace scenarios.

When the Evidence Really Is in Balance

Sometimes it is. If so, say that the allegation was neither substantiated nor disproved, and explain what you did to try to resolve it. An inconclusive finding is not a finding that the complainant was untruthful, and it must never be treated as one.

The 1999 EEOC guidance advised that when no determination can be made, the employer should still take further preventive measures such as training and monitoring. In practice that means reminding both parties of the policy, checking in with the complainant, watching for retaliation, and keeping the file so that a later, similar report is assessed with full context. Our article on Workplace Retaliation: Prevention and Documentation covers the follow-up.

Some cases add difficulty that the basic factors do not solve, such as a complainant who will not be named or an accused who outranks the investigator. Those are covered in our guide to anonymous complaints, accused executives and reluctant witnesses.

Frequently Asked Questions

The Bottom Line

Conflicting accounts are normal. What separates a defensible investigation from a weak one is whether the investigator worked through the conflict with stated factors and a stated standard, and wrote the reasoning down. Lean hardest on corroboration and consistency, treat demeanor with caution, and be precise about where the EEOC factors come from now that the agency's harassment guidance has changed.

To see where credibility fits in the full process, start with the Workplace Investigations Guide and keep the Workplace Investigations Checklist at hand. To practice the skill with an instructor and case scenarios, review the Internal Investigations Certificate Program.