Could data science someday disrupt HR recruiting?

Got a nice email from Simone a while ago and now finally getting my act together to share it with everyone (thanks for your patience Simone!). She passed along a very cool article in the NY Times about how data scientist Luca Bonmassar (co-founder and CTO of startup Gild) is using algorithms to evaluate software developer job candidates. hiring

It can be really, really hard to find awesome developers. This is especially the case with “big data” experts who truly understand data science. Sometimes these people are well known, but a lot of times they’re quietly doing magic in a cave somewhere relatively unnoticed by the rest of the world. That’s why Bonmassar is using data science to find talent and evaluate job applicants.

Bonmassar and his colleagues do this by scraping info. from the Internet and using algorithms to crunch thousands of variables about people such as the sites they hang out in, the words they use, how others describe them, projects they’ve worked on, and so forth. It uses this data to get a better sense of job applicants and for spotting diamonds in the rough. Gild isn’t alone; other firms like TalentBinRemarkableHire and Entelo are in the race too, albeit through slightly different approaches.

Could HR recruiters someday be disrupted by data science?  If so, would that be a good thing or bad?

Imagine a world where job applicants are found and evaluated by their merits and contributions, rather than by how well they sell themselves in an interview or who they buy enough beers for. Sometimes it seems like getting a developer job can be more about hanging with the “cool kids” as opposed to raw results and talent. In that sense, a more merit-centric world governed by algorithms can seem more productive and fair. On the other hand, could an algorithm detect if someone’s a jerk? Are they a good fit for a firm’s culture? Are they a meth-head? Are there things only a human HR recruiter can do?

I think there are pros and cons for using algorithms like these, but I must confess I’m optimistic. There are some things human brains are good at, but other things algorithms excel at. Ultimately the best hires will go to those with the right combination of human-mechanical processes in their search for top talent.  You can’t ignore the “human” element, but you’d also be remiss to ignore the science.

Click here to read the article itself.


This Post Has 11 Comments

  1. Erich

    Based on the article cited, data science is already disrupting the HR process. It is important to see, however, that the statistical analysis mentioned is not replacing the expertise of HR managers. Rather it provides the necessary evidence.

  2. Paul

    Indeed – for many looking for a job, the prospect of going through an interview triggers the same emotional (and physical!) responses as passing a school exam – sweaty palms, increased heartbeat, discomfort, etc. Sure, gathering some peripheral information that helps to identify a candidate can be beneficial in supporting him/her; however, “Big Data” can only go so far as to gather information that exists & is available – and remember many candidates “are quietly doing magic in a cave somewhere relatively unnoticed” (as referenced in the article). This often means the impossibility to divulge the nature of the work due to contract agreement or security issue.
    On the “Innovative” side of the conversation, we can also look at underused resources – such as what SAP plans to do by hiring autistic adults to work as developers and analysts ( ).

    If, however, the topic is to monitor already hired resources, trawling through company resources can be somewhat accurate – again, providing the contract agreement and security issues are respected!

  3. Peter

    Maybe yes and no, Tom. “No”, because the obvious reason that many large companies are in dire trouble is that they cookie cutter (aka computer) select the same old worn out crowd who don’t know how to get into good new businesses. We await with frothing mouths a computer program that will select crazies like Bill Gates, Bill Hewlett, Steve Jobs, and Dave Packard. The cruel and obvious truth that can be seen by a blind man in a hurry is that the hired help that suffices for these guys is not up to the job, whatever the size of their incentive pay. Because these kinds of guys are not interested in the pay that HR people claim is so critical. Neither of course was Jesus or Mother Theresa, and they seemed to do rather well. And the missing piece here is that this applies to most normal people, as HBS articles over decades have pointed out. So garbage in garbage out. There again “yes”, because there is always hope that computer programs, unlike successive versions of Windows and Office, will actually improve and bring benefit to this field. And perhaps the real problem is that it is impossible for the computer selected geniuses to get anywhere in a large corporation as they prove themselves by signing time cards and deciding what color of carpet to put in new Dilbert cubes. So their only route is to start their own company and then to feed off the customer base of the dying dinosaurs. Natural selection may be a better model for HR software, like nature is for most things in life. Great article Tom, thanks!

  4. Gregg Niemi

    It’s already happening. Behavioral Science, Big Data, Predictive Analytics, and Machine Learning are teaming up to maximize workforce profitability across the entire employee life-cycle! See, for instance:


    Their influence will only increase as companies realize the potential profits.

  5. Miles H. Vargas

    This is a dictionary of algorithms, algorithmic techniques, data structures, archetypical problems, and related definitions. Algorithms include common functions, such as Ackermann’s function . Problems include traveling salesman and Byzantine generals . Some entries have links to implementations and more information. Index pages list entries by area and by type . The two-level index has a total download 1/20 as big as this page.

  6. Richard

    Separating the wheat from the chaff is quite old conundrum. Being known for what you do (i.e. reputation) is best for technical skills as it’s not bragging if you can actually do it. Fitting into a team is an art. There used to be a no-foul, try-before-you-buy provisional employment (lasting from 6 weeks to 6 months). I haven’t seen this in quite some time. I’m no legal expert, but I think laws exist now that prevent this.

    Having run my own small company, interviews were techno-social events. We didn’t hire sight-unseen. All candidates had to supply samples of work (even contrived ones we’d set up a week before the day of interview) prior to the interview. It was the ability to work (this is the social part) with the team and drive toward a good solution; where good did not always mean the best. We found the “prima donnas,” the “my way or the highway” and the “whatever!” types very quickly. If the candidate would argue for the best solution, and that may not be theirs, we found a good candidate. Remember I did say a small company (20-30 persons). This can be done in large companies if and only if there is a cost center approach, but as Keith says, HR will be an impediment as they will want a “corporate uniform” process.

  7. Gordon Long

    The second point is that, as I’ve written a bunch of times before (see here and here , for example) data-driven decisions, predictions, and diagnoses are much better than those that come from human intuition and HiPPOs (the ‘Highest-Paid Person’s Opinions’). The research is overwhelming on this point. HiPPOs might be good for some things, but their crystal balls just don’t work very well. The soulless output of a data-driven, mechanistic algorithm is demonstrably and significantly better, in domain after domain. So HiPPOs should become an endangered species.

  8. Gustavo J. Chapman

    My copy of The Design and Analysis of Computer Algorithms has arrived today. In the first chapter, the author introduced Turing Machines. I have two other algorithms textbooks, Introduction to Algorithms and The Algorithm Design Manual , but none of them talks about Turing machines, even though they are famous on the subject of algorithms and data structures.

  9. Jerome G. Jenkins

    Depending on the data structures and processes involved in an application, it may become necessary to sort the data stored within it. Different data structures enforce certain constraints on applications. For example, if a program is using a data structure such as a list, but this data structure only allows new items to be added at one end, the resulting store will comprise data that is not ordered. Sorting algorithms allow programmers to either rearrange data structures, ordering them by value, or to copy the items in order, into a second data structure.

  10. Elena S. Rush

    This is a dictionary of algorithms, algorithmic techniques, data structures, archetypical problems, and related definitions. Algorithms include common functions, such as Ackermann’s function . Problems include traveling salesman and Byzantine generals . Some entries have links to implementations and more information. Index pages list entries by area and by type . The two-level index has a total download 1/20 as big as this page.

Leave a Reply