Why Interviews Fail
Most interviews produce noise, not signal. Understanding why helps you avoid the same traps.
The problem with unstructured interviews
Research consistently shows that unstructured interviews — casual conversations without a consistent framework — are among the worst predictors of job performance. Yet they remain the most common interview format.
The top reasons interviews fail
| Failure Mode | What Happens | The Cost |
|---|---|---|
| No structure | Every interviewer asks different questions | Can't compare candidates fairly |
| Confirmation bias | Interviewer forms an opinion in the first 5 minutes, then seeks evidence to confirm it | Miss strong candidates, hire weak ones |
| Halo effect | One impressive answer colors the entire evaluation | Overweight a single data point |
| Leading questions | "You're good at handling pressure, right?" | Get the answer you want, not the truth |
| Recency bias | Remember the last candidate better than the first | Unfair comparison across the loop |
| Similarity bias | Favor candidates who remind you of yourself | Homogeneous teams, missed talent |
What good looks like
Great interviews share three properties:
- Structured — Same questions, same evaluation criteria, every time
- Behavioral — Focus on what the candidate actually did, not what they would do
- Calibrated — Interviewers align on what "good" looks like before the loop starts
Exercise: Spot the failure mode
Read each scenario and identify which failure mode is at play.
Scenario 1: An interviewer starts the interview by saying "I see you went to MIT — I went there too!" and spends 10 minutes reminiscing about campus life.
→ Similarity bias — the shared background creates an unfair positive impression before the interview even begins.
Scenario 2: A candidate gives a mediocre answer to the first question. The interviewer spends the remaining 45 minutes asking increasingly easy questions, already mentally writing a "no hire" assessment.
→ Confirmation bias — the interviewer decided early and stopped looking for signal.
Scenario 3: The interviewer asks: "We move really fast here. Can you keep up with a fast-paced environment?"
→ Leading question — the expected answer is baked into the question. No useful data can come from this.
Which interview approach is the strongest predictor of job performance?
An interviewer loves a candidate's answer to question 3 and rates all subsequent answers higher than they deserve. Which bias is this?
Why are leading questions harmful in interviews?