As a Bright employee, I constantly get asked the question, “What does Bright do?” The unofficial company line goes, “We try to make hiring more efficient.” While that’s an entirely accurate statement that acts as the guiding light for all we do, it’s a bit of an oversimplification.
As a Bright Data Scientist, I think about Bright’s mission statement as a question: “How do we select the best candidate out of a stack of resumes?” I like this particular phrasing because, under a certain set of rules, it becomes the celebrated Secretary Problem with its strikingly elegant n/e stopping rule.
To understand the Secretary Problem, let’s consider the mathematically equivalent “Marriage Problem.” Our friend is in his early-to-mid twenties, and being the traditional type, thoughts of settling down with a nice lady have now crept into his conscious existence. He begins to reason that it takes about a year to get to know someone well enough to decide her marriageability, and once he knows, he will either irrevocably call it quits or propose. If he would like to get married no later than forty, how should our friend proceed?
He could simply start dating and propose to the first woman he likes. But having spent his adolescent and college years predominantly on the internets, our friend’s relationship knowledge comes primarily from episodes of How I Met Your Mother and he has no clue what he’s looking for. In other words, he should probably date a few women, without the sole intention of proposing, but how many? He can realistically have serious relationships with about a dozen (n = 12) women before he’s forty.
Using the mathematics of the Secretary Problem’s optimal stopping rule, this number is four (~12/e). Reasonably, after four serious relationships, our friend would have a pretty good idea of what he wants in a spouse, and can confidently propose to the next woman who is a better match than all previous relationships.
Going back to the issue of hiring: recruiters want to choose the best possible candidate to send to interview. Say the recruiter performs a search in a resume database and gets 300 results, or posts a job online and gets 300 applicants (n = 300). Applying the math behind the Secretary Problem, they’d be advised to review and reject 110 resumes (300/e), then select for interview the next candidate on the list who is better than the rejected bunch. This method maximizes the probability of selecting the best interviewee, but it’s awfully time-consuming.
It is here that Bright’s mission statement about efficient hiring comes back into play. Unlike other recruiting websites, Bright calculates Bright Scores for all candidate search results and applicants. With all candidates scored on a consistent scale, the recruiter can focus on an absolute rank rather than play the odds and choose based on relative rank between candidates, as in the Secretary Problem.
Our lonely friend might best off if he were able to date all the women in the world and then pick the best match, but neither time nor emotional baggage nor mutual willingness allow such a process. Similarly, no company has time or resources to interview every applicant or candidate search result to find the best match. Using the optimal stopping theory in the Secretary Problem would maximize the likelihood of picking a strong candidate, but at Bright, our mission to make hiring more efficient led us to develop another solution: The Bright Score.