An investigation is only as good as its interviews. Documents establish what was recorded; interviews establish what happened, why, and what people knew. And unlike documents, interviews can be irreparably damaged by how they are conducted — a leading question contaminates a witness's memory permanently, and an interviewee who feels ambushed stops volunteering anything.
The goal of an investigation interview is not confession or confrontation. It is a complete, accurate, contemporaneous account from each person, obtained in a way that will hold up when someone else reads the file two years later.
The standard sequence is:
Interviewing the accused too early is the most common sequencing error. Without the full picture you cannot ask specific questions, and you give the accused an early opportunity to shape their account and potentially influence witnesses.
Prepare a question outline, not a script. Know the facts you need to establish. Do not read questions in order — the best material comes from following what the person actually says.
Review the documents first. Emails, messages, schedules, access logs, prior complaints, personnel records. Never learn a fact from a document during an interview that you could have known beforehand.
Choose the setting. Private, neutral, and not the accused's or complainant's own office. Allow enough time that nobody is watching the clock. Have a second person present as a note-taker where possible — it frees the interviewer to listen and provides a corroborating witness to what was said.
The first three minutes set the tone for everything after. Cover, in order:
For the accused, add that they will have a full opportunity to respond to each allegation.
Start broad, then narrow. Broad questions produce narrative; narrative produces detail you did not know to ask for.
"Tell me about your working relationship with [person]."
↓
"You mentioned things changed in March. Tell me about that."
↓
"What specifically did he say?"
↓
"Were those his exact words, or your summary?"
The final step matters more than it looks. The difference between a paraphrase and a quotation is often the difference between a substantiated and unsubstantiated finding.
Open: What happened? Tell me about that meeting. What did you observe? What did you do next?
Closed (use sparingly, to pin down specifics): What date was that? Who else was in the room? Did you report it?
Never leading: "He made you uncomfortable, didn't he?" — this contaminates the answer and taints the record.
Distinguish firsthand knowledge from hearsay explicitly and record which is which. "How do you know that?" is the most useful five-word question in an investigation.
After an answer, wait. Most interviewers fill the pause and lose the elaboration that was coming. The information volunteered in the second half of an answer is consistently more useful than the first.
"Is there anything else I should know?"
"Is there anyone else I should talk to?"
"Is there any document, message, or record I should look at?"
"Do you have any questions for me?"
These four questions surface more new evidence than any other part of the interview.
Get the complete account before narrowing. Let them tell it their way first, then go back through each incident systematically.
Establish for each allegation: what happened, when, who was present, whether anyone was told at the time, what documentation exists, and what effect it had. Ask what outcome they are seeking — not because it determines the finding, but because it informs remedy and sometimes reveals context.
Take the account seriously and remain neutral. Neither expressing sympathy nor expressing skepticism belongs here; both compromise your position as a neutral fact-finder. Believe that the interview should be thorough; do not pre-judge whether the allegation is true.
Interview witnesses separately and promptly. Memory degrades quickly and conversation between witnesses contaminates accounts.
Ask about what they observed, not what they concluded. "Did you see any harassment?" invites a legal judgment; "Describe what you saw and heard in that meeting" produces evidence.
Include witnesses who may support the accused's account. An investigation that interviews only the complainant's witnesses is not neutral, and it reads that way in a deposition.
Ask each witness who else was present or might know something. Witness lists should grow during an investigation.
This interview requires the most care and is most often mishandled.
Provide enough detail to respond meaningfully. You need not disclose the complainant's identity in every case, but the accused must know the substance of what is alleged — the conduct, the approximate timeframe, the context. An accused who cannot understand the allegation cannot respond to it, and a process that denies a meaningful opportunity to respond is procedurally indefensible.
Ask for their account first, before presenting specifics. Let them describe the working relationship and relevant events. Then go allegation by allegation.
Present each allegation neutrally and ask for a response: "It has been reported that on approximately [date], you said [X]. What can you tell me about that?"
Ask for evidence and witnesses supporting their account, and follow up on what they provide. Failing to pursue exculpatory leads is a serious defect.
Watch the transition from denial to justification. "That never happened" followed later by "and anyway it was a joke" is a meaningful shift and should be noted precisely.
Do not promise outcomes, and do not indicate that you have reached a conclusion.
Union employees have Weingarten rights — the right to request a union representative at an investigatory interview the employee reasonably believes may result in discipline. The right must be requested by the employee; the employer need not offer it. If requested, the employer may grant it, discontinue the interview, or offer the employee the choice of proceeding without representation. Denying a properly invoked request and continuing the interview is an unfair labor practice.
Non-union employees generally have no equivalent right to a representative, though some employers permit a support person as a matter of policy.
When accounts conflict, you must still reach a conclusion on a preponderance standard. Base credibility findings on articulable factors, not intuition:
Two cautions. Demeanor is a poor indicator. Nervousness, poor eye contact, and emotional flatness correlate weakly with deception and strongly with anxiety, trauma, and cultural difference. Do not build a credibility finding on how someone seemed.
And document the reasoning. "I found the complainant more credible" is a conclusion. "The complainant's account was corroborated by two witnesses and by messages sent that evening, while the respondent's account changed between the first and second interviews on the material point of whether the meeting occurred" is a finding.
Take notes during the interview, not afterward. Record what was said as closely as possible, distinguishing quotations from summary. Note the date, time, location, and who was present.
Write the interview summary within 24 hours, while recall is accurate.
Notes are discoverable. Write facts and observations, not opinions, speculation, or legal characterizations. Avoid editorializing about people's personalities.
Recording. Whether to record is partly a legal question — several states require all-party consent — and partly a practical one. Recordings are accurate but tend to reduce candor. Whatever you decide, apply it consistently across the investigation, and never record covertly.
Do not have interviewees sign statements you drafted unless they have genuinely reviewed and adopted them. A signed statement the person did not write in their own words is weak evidence and looks coercive.
Complainant first, then witnesses, then the accused, then follow-ups. Interviewing the accused before you understand the allegations prevents you from asking specific questions and risks witness influence.
You may request confidentiality tied to the integrity of the investigation, but blanket confidentiality directives can unlawfully restrict employees' rights to discuss workplace conditions. Frame it as a request with a stated reason rather than a blanket prohibition.
Not always, but the accused must know the substance of the allegations well enough to respond meaningfully. Withholding so much detail that a response is impossible undermines the process.
Union employees may request a union representative at an investigatory interview they reasonably believe may lead to discipline, under Weingarten. The employee must request it. Non-union employees generally have no equivalent right, though employer policy may allow a support person.
Apply articulable credibility factors — plausibility, consistency, corroboration, motive, detail, contemporaneous conduct — and document the reasoning. Do not rely on demeanor, which is an unreliable indicator.
It depends on state consent laws and on your judgment about candor. Recordings are accurate but tend to make people guarded. Whatever you choose, apply it consistently and never record covertly.
Plan the order, prepare from the documents, open with a clear and non-threatening frame, funnel from broad to specific, and separate what people saw from what they concluded. Then document reasoning rather than impressions — the credibility finding is the part of the file that gets read hardest.
For structured instruction, explore our Training for Workplace Investigations and Internal Investigations Training, or review the Workplace Investigation FAQs.
Recommended In-Person Seminars
1-770-410-1219
support@HRTrainingCenter.com
770-410-1219
