A few years ago I was helping my cousin get ready for interviews with the big tech companies. It was not a serious coaching engagement, more a family favor over coffee: we talked about how interviews work, tried a few classic questions, compared notes between rounds. One of the questions we tried was about proactivity and taking responsibility, the "tell me about a time you owned something beyond your role" family. His answer was golden. Specific situation, clear personal actions, a measurable result, a humble lesson at the end. I was impressed, and because this was practice and not an interview, I asked more about the situation out of plain curiosity. What happened after that? He shrugged. "Oh, I don't know. This didn't really happen."
He got the job, by the way. At a company everybody knows, with an interview process people write books about.
What stays with me is how little it took. The story had survived every question it was designed for and collapsed at the first one it was not. Nobody in the real process, apparently, ever asked it. That question, planned in advance and attached to the question it serves, is called a follow-up probe, and it is the difference between scoring the answer a candidate prepared and scoring the experience they actually have. This article covers what probes are, what the research says about using them without wrecking your structure, and how to write a good set for any interview question in a few minutes.
What a follow-up probe is
A probe is a short follow-up question that fills out an incomplete answer or asks for clarifying detail. The research literature defines it almost exactly that way: a question "intended to augment an inadequate or incomplete response provided by the applicant, or to seek additional or clarifying information" (Levashina et al., 2014). The US federal government's structured interview guide uses the same definition and adds the practical part: decide your probes before the interview, per question, in writing.
A probe is not a new question. Your behavioral question asks for the story; the probes ask for the parts of the story your scoring needs. The previous article in this path put it as one instrument with three parts: question, probes, anchors, designed as a set. The question gets the candidate talking about a real incident. The anchors define what a strong answer to that question contains. The probes are the bridge between them, the questions that surface exactly the details the anchors ask you to score.
That is also why probes cannot be left to in-the-moment inspiration, and this is where things get more interesting than "just ask follow-ups".
The probing paradox
Here is the uncomfortable part. The classic taxonomy of interview structure (Campion, Palmer and Campion, 1997) lists fifteen components that make interviews predictive, and one of the content components is limiting ad-hoc prompting and follow-up. There is a reason for that: unplanned probing is a door through which the interviewer's biases walk back into the room. You probe the candidate you warmed to differently than the one you didn't, and older research found interviewers agreed with each other more when they were not allowed to probe at all.
So should you probe at all? The field's answer is a middle path with four levels behind it. Interview structure research distinguishes prohibiting probes entirely, allowing only planned probes, encouraging unlimited probing, and giving no guidance at all (Levashina et al., 2014). Prohibition protects standardization but leaves you scoring whatever surface the candidate chose to present, rehearsed or real. Unlimited probing gets you detail at the cost of every candidate getting a different interview. Planned probing, the same short list of neutral follow-ups attached to each question and available for every candidate, is the version the researchers themselves recommend: neutral wording that does not cue the desired answer, applied equally, used when an answer is thin, vague, or off track.
Honesty requires a caveat here, and this site tries to be honest about evidence: probing is one of the least-studied components of interview structure, and the recommendation of planned probes is the field's considered judgment rather than a settled meta-analytic fact. But the two failure modes it steers between are both well documented, and one of them is worse than it looks.
Probes and the rehearsed candidate
The obvious theory of probing is that it catches fakers: press on the story and the invented ones crack. The research found something stranger. In the one substantial study on the question, follow-up probing did not reduce faking, it increased it (Levashina and Campion, 2007). Debriefed applicants explained why: aggressive probing signaled which answers mattered, so they embellished harder exactly there. And when interviewers probed candidates about experiences they had never had, some candidates invented experiences to be polite. The probe was read as a demand for a better story, so a better story was supplied.
Read my cousin's shrug again with that finding in mind. My follow-up worked precisely because it carried no pressure. It was not a trap or a challenge, just casual curiosity asking for more of the reality, and a story that happened has an endless supply of reality while a story that did not runs out immediately. In a real interview with stakes on the table, the same candidate under a hostile "oh really, prove it" probe might well have produced another fluent paragraph instead of the truth.
The design consequences are worth spelling out:
- Probes ask for facts, not better performances. "What exactly did you do next?" retrieves memory. "Are you sure that's the whole story?" demands theater, and rewards whoever performs best under suspicion.
- Probe toward the checkable. Questions about who was involved, what happened when, and what the outcome cost are hard to invent consistently and easy to take notes on. Invented stories are thin on exactly this kind of detail.
- Let absence be an answer. If a candidate has no incident to describe, a swap to a situational variant tells you more than pressing until they manufacture one.
The who, what, when pattern
The most useful probe kit in the public domain comes from the OPM guide, which attaches probes to each part of a behavioral answer. If you think in STAR terms (situation, task, action, result), the pattern gives every part its follow-ups:
| Answer part | Probes that fill it |
|---|---|
| Situation | What led up to this? What made it hard? When was this, and what was your role at the time? |
| Action | What exactly did you do? What did you consider and reject? Who else was involved? |
| Result | What was the outcome? How did you know it worked? What did it cost? What happened afterward? |
Notice the grammar these share: who, what, when. Every probe in the pattern asks the candidate to be more concrete about something that already happened. None of them suggests what a good answer would contain, which is the OPM's other rule worth stealing: a probe may be tailored to what the candidate just said, but its general meaning should stay the same from candidate to candidate. "What was the outcome?" can become "what happened with the customer after that?" without becoming a different question.
The pattern also names the single most common probe you will actually use. Most candidates answer behavioral questions with "we": we redesigned the process, we saved the account. The probe that separates the driver from the passenger is four words: what was your part?
Writing probes for your own questions
You need your chosen questions and their anchors. For each question, three steps.
Step 1: read your anchors and list what they need. Every line of "what good looks like" and every red flag is a claim you will have to judge, which means it is a detail the answer must surface. If your anchor says "checked back with the customer afterward", you need a probe that reaches the afterward. If a red flag says "the underlying problem never got fixed", you need a probe that asks what happened to the problem, not just the feelings. The anchors are a shopping list; the probes do the shopping.
Step 2: phrase each probe as neutral fact-finding. Who, what, when, what did it cost, what happened next. Strip anything that hints at the answer you hope for. "Did you escalate it to your manager?" teaches the candidate that escalation is on your rubric; "what did you do next?" finds out whether escalation is in their repertoire.
Step 3: cap the list at three or four. A planned probe list is a menu, not a script; you will rarely ask all of them. Three or four probes per question is enough to fill any honest answer's gaps, and few enough that the interview stays on schedule.
Here is what the finished product looks like in the pantry, on the same angry-customer card the rating-scales article scored:
Map the card's probes back to its anchors and the shopping-list logic is visible. The anchor "separates the person's anger from the actual problem" is served by the probe what were they actually angry about, underneath the first complaint? The anchor "names a concrete fix or an honest limit" is served by what did resolving it cost you or your team? The anchor "checks back afterward" is served by where did the relationship land afterward? Every probe exists because an anchor needs it, and every anchor can be scored because a probe feeds it.
Using probes in the room
The full craft of running the interview gets its own article, but three habits belong to the probes themselves.
Probe on deficiency, not on charm. The planned list is available for every candidate, and the trigger for using it is the state of the answer: thin, vague, off track, or all "we" and no "I". The failure mode to watch in yourself is probing the candidates you like, to help them shine, and letting the others stay vague. Equal access to probes is what keeps the interview one instrument instead of two.
Try silence first. The cheapest probe is a beat of quiet after the answer stops. Most people fill it, and what they fill it with was not rehearsed. It costs nothing and cues nothing, which makes it the purest probe there is.
Write down what the probe surfaced. A probed detail that never reaches your notes might as well not have been asked, because scoring happens against the notes, right after the answer, on the anchored scale.
One question away
I think about my cousin's interview process sometimes. Several rounds, trained interviewers, a company with an employer brand built on rigor, and a fabricated story sailed through because every question got its prepared answer and no question got a follow-up. The gap between what they measured and what I stumbled into over coffee was one casual, factual, entirely friendly probe.
Every question card in the pantry ships with its probes already attached, written against that card's anchors the way this article describes, so if you build your interview from the library the shopping is done. Sharpen the wording for your role if you like, put the probes on the interview guide next to their question, or let the planner assemble the whole plan for you, free, in about three minutes, no account needed.
Rehearsed answers are not lies, mostly. They are just surfaces. The probe is how you find out what is underneath, and it works best when you decide what to ask before anyone is performing for you.