Loading
Most jobs are handed out after a conversation. Both sides treat it as the fair part of the process.
How well it actually predicts anything has been studied for decades.
One where everyone gets the same questions in the same order, scored against the same scale. And one where it is a chat that goes wherever it goes.
The chat. It feels human, and interviewers usually prefer it, and say they can read people.
Candidates often prefer it too.
Which predicts later job performance better?
In the largest review of selection research, structure raises the predictive power of an interview to about half again as much, and other reviews put the gap wider still.
Same people, same time, different format.
Because it measures the wrong thing well. An unstructured conversation reliably detects who is comfortable in conversation.
That is a real trait. It is just not the job.
With no fixed questions, an interviewer follows what interests them, which is often what resembles them.
Structure closes that door by removing the choice.
Which format does each belong to?
Tap an item, then a bucket
Interviewers are usually sure they are good at this, and that confidence does not track accuracy.
A strong impression is easy to form and hard to check.
Even the best single method leaves most of the variation in performance unexplained.
Combining a structured interview with another measure does better than either alone, and still not by much.
A rejection after an unstructured interview carries very little information about you.
The process was mostly measuring something else.
Writing the questions down in advance and scoring them the same way is the single cheapest improvement available.
It costs an hour, once.
The interview everyone enjoys is the one that predicts least. Asking everyone the same questions and scoring them the same way works substantially better, and almost nobody does it.