Or simply that they have significant bias in their interview process for selecting candidates with strong interview skills. At that point it’s tautological—if you only select people who ace interviews then you’ll have a bunch of people who ace interviews.

Author here. We didn't publish a histogram showing how many users we have from each company, so I guess you'll have to trust me on this one. Dropbox has waaaay fewer engineers on interviewing.io per capita than some of the other companies... many of whom didn't make the top ten list.

In our experience, they're quite good at retaining talent.

Doesn't this exact comment undercut your results fairly heavily?

You're admitting that it's much more likely dropbox is a statistical anomaly because of fewer data points, rather than a robust data set.

it's possible. we did see statistical significance when comparing dropbox to others despite the relatively smaller sample size. but yes, that is possible.

Maybe add error bars to give it a little more color?

I remember seeing a job description that was something like

  must be self starter, enjoy high energy postive environment!  enjoy coding in RoR etc

To me this reads like legacy Rails codebase people are too scared to touch that's always causing fires. No thanks.

There has never been any rails code at Dropbox