When candidates are interviewed for jobs or admissions, we can generally assume that they aren’t all equally qualified—there is a distribution. For example, it might be the case that half are above average and half are below average.
Because there is a distribution, we also know (at least intuitively) that there are probabilities involved. Using this example:
- If you randomly select one candidate to interview, the probability that they are above average is ½
- With 2 candidates, the probability that they are both above average is:
½ × ½ = 1/4
- With 3 candidates the probability that they are all above average is:
½ × ½ × ½ = 1/8
- With 4 candidates, the probability that they are all above average is:
½ × ½ × ½ × ½ = 1/16
- And so on.
It’s pretty unlikely that you would interview a bunch of great candidates all in a row. So if you already interviewed a few that seem like a good fit, you know the next one is probably going to be a dud. If you give them a favorable evaluation, you probably did something wrong.
…That’s our intuition, anyway. And it turns out that this intuition is often wrong. Our perceptions are often too narrow (focused on the interviews within a given day, rather than all of the interviews) or based on faulty assumptions (unlike the example above, we probably don’t know the actual distribution of candidates that have been selected to interview).
Researchers from University of Pennsylvania and Harvard analyzed 10 years of interviews. They found that when there are favorable interview outcomes earlier in a given day, the outcomes later in the day are much less favorable. This could make sense, given probabilities like what we described above. But they also found that this tendency goes way too far—strong candidates that are interviewed later in the day receive much less favorable evaluations than they should have.
Now that you know this, don’t hesitate to give a strong candidate a positive evaluation—even if they aren’t one of the first to interview.
We should also keep in mind that even very good interview strategies only predict about 9% of the variation in actual performance. If you want to do better than that, you need to include more objective tools.