[hibikir]

Quotas aren't necessarily the best ideas, but if 90% of your applicants are men, and men and women have the same pass rates, what you have to work on is resume sourcing. This is especially true in companies that mostly hire referrals: You are mostly hiring people that are like the people you have, so you'll lose diversity on average.

You also have to look at differences in the middle of the pipeline. Imagine you only give a phone screen to 5% of female applicants, but 10% of male applicants. You have to think very hard about why that happens. Maybe your ways of rating resumes have a built in bias.

For instance, imagine that I only interviewed new grads that had at least two internships in large tech companies. A rubric like that looks neutral, but you'll discover that the demographics of CS graduates vs those that have those two internships are very different (far fewer women, and a lot more people that will identify themselves as asian).

[aries1980]

> Quotas aren't necessarily the best ideas, but if 90% of your applicants are men, and men and women have the same pass rates, what you have to work on is resume sourcing.

I agree with your approach, but the gender is not the only diversity metric. If you push this toward ethnicity, religion, sexual orientation, introversion, geekness level, you might end up in a difficult situation and your company will look like Noah's ark. Instead, what I would do is to specify the tasks that needs to be carried out and define metric to measure it, then do a blind hiring. Attributes like graduation or work experience is irrelevant if the candidate passed the test. During the probationary period, if the team doesn't like the new members or the other way round, you can let them go.