Why Structured Interviews Drift Once the Conversation Starts

Ask a hiring team whether they use structured interviews and they'll almost always say yes. They have good reason to. For decades, meta-analyses have found that structured interviews predict job performance much better than unstructured ones, and every modern HR textbook recommends them. It's one of the few questions in hiring the evidence has settled.

So why do so many structured interviews look unstructured if you sit in on them?

On paper the method is strict. The questions are decided in advance, answers are scored against the same rubric, candidates are rated before they're compared, and the interviewer doesn't improvise. Now picture how it usually goes. A hiring manager has read the question guide once. He asks the first few questions as written, gets an interesting answer to the second one, and spends the next twenty minutes following that thread. Afterwards he fills in the rubric, scoring the candidate on questions he mostly didn't ask. What happened was an unstructured interview with some paperwork attached.

The drift isn't random, which makes it worse. When an interviewer takes to a candidate in the first few minutes, because of how they come across or a shared background or any number of cues he isn't aware of, the rubric softens and he starts building a case for them. When he doesn't, he stays strict. So the rubric ends up confirming the first impression instead of correcting it, which is the opposite of what it was for.

I don't blame the interviewers. Consider what real consistency asks of them. Every candidate has to hear the same questions in the same order, with the same follow-ups when an answer is thin. People are bad at that. Memory slips, the clock is running, and following an interesting remark is the natural thing to do in a conversation. Ask someone to suppress all that and deliver the same 45-minute interaction fifteen times a week, and the quality falls off fast.

Consistency also requires scoring each candidate before comparing them with anyone else. That depends on a kind of mental isolation high-volume recruiting doesn't allow. The same person is reviewing applications, running calls, and typing scores into ATS notes in batches, often in the same afternoon. By the time they rate the fifth candidate, the first four are already in their head.

So the method is sound and the people are normal. What fails is the assumption that humans can perform a demanding behavioral discipline reliably, at volume, week after week, without it wearing down. The gap between structured-interview policy and structured-interview practice has been known in organizational psychology for decades. It got accepted as good enough because there was no practical alternative.

I think there is one now, and it's a large part of why we built Asendia. It screens candidates by phone against a rubric the hiring team sets for the role. Each candidate gets the same structured questions, the same follow-up probes when an answer is shallow, and scoring against the same criteria, whether they applied first thing in the morning or late at night and whether they're the second call of the day or the two-hundred-and-second. Software doesn't get tired and doesn't take a liking to anyone, so consistency stops being something you ask people to keep up and becomes the default. What reaches whoever runs round two is a qualification summary and verbatim excerpts from each conversation in the ATS, and because every candidate answered the same core questions, those summaries can be laid side by side and compared fairly. There's more on why it matters that AI actually carries out pipeline steps, instead of just assisting, in the post on agentic recruiting.

For agencies the case is stronger. Imagine three recruiters running parallel campaigns for several clients, with 400 first-contact calls to get through in a week. Nobody could keep a rubric intact across that. Software can, and then the recruiters' judgment goes where it's worth the most: reading the shortlist closely, looking after clients, and running the later conversations that need a person. I've written separately about how the recruiting agency model is splitting in two, and this is part of the reason.

What I find interesting is that none of this needs new evidence. Structured interviews always promised a first-pass evaluation that reflects what a candidate can do, not how the interviewer felt about them. The research said the promise was real. The delivery was what kept failing, and delivery is the part a machine happens to be good at.

Ready to transform your hiring strategy? Schedule a Demo with our founders today!