Busting Your Bracket: Why Recruiting Stats Are Crap

Do you have a March Madness bracket? I personally have a few running right now. My strategy to pick this time? Random.

My logic here is that every year before now, I’ve gone on ESPN and checked the predictions from the sportscasters. I have that little bit of hope that this could be the year I actually win a bracket contest. And do you know who I lose to? Usually it’s the person who knows nothing about sports - the one who picked favorite colors or what mascot would win in a cage fight and I’m left with a completely busted bracket after day 1. And usually, I'm a little pissed about it. I'm competitive, what can I say.

Hypothetically speaking, if the stats were always right, you should be able to pick the #1 seeds in each conference and win every time - they’re #1 for a reason, right? Put it all down on paper and pick the winner of the four #1 seeds and boom – you should win. But we can’t do that because we’re working with something completely unpredictable – groups of humans.

I bet you’re catching my drift on how this applies to recruiting. If not, let me break it down for you.

There’s nothing inherently wrong with recruiting statistics but I’m saying, they just might not be right or something you can necessarily use. Rankings and statistics mean shit when you lose – in March Madness or with a great candidate. The #1 candidate isn't always going to pick your company and you won't always have the best in class employer brand to even get those candidates to apply. Then, we take a pile of statistics from the analysts and try to hold our companies to those standards (and fail miserably, just like I am in my bracket).

That’s the thing about recruiting - we're dealing with humans; every person is different, every job is different. Yeah, it’s great to know that 50% of the time x-y-z strategy works according to some study, but bottom line, we’re really flipping a coin on what will work if we don't know the metrics that align with our business, not the big brand standard. We need more companies to be really honest and talk about their stats, not citing the blanket stats that are the average for everyone. That’s how we can really help each other be better at this hiring and selecting thing.

A great "big data" recruiting strategy should start small - with the baseline metrics for YOUR company, looking at annual reviews, staffing projections, workforce planning – not the baseline metrics on average from companies you don't work for. Why? Because you don’t know if that’s bull shit in the first place. 84% of all stats are made up. (Hell, I just made that one up.)

What do you think? How do you make the big data work for your small company?

(Have to give a shout-out to Steve Levy for inspiring this post with a late call on a Friday afternoon. You're the man, Levy)

Comment by Steve Levy on March 27, 2015 at 2:00pm: Note to anyone in recruiting: If you're going to use ANY data, will you please take a friggin' stats course (or update the one you hated when you first took it). If you plan on doing anything predictive, learn about multivariate stats which is much harder than univariate stats (NO - percentages aren't predictive).

Comment by Matt Charney on March 27, 2015 at 2:06pm: A recent poll by Bersin found only 12% were actually crap. Approximately 1 in 4 employers are far enough along on the analytics maturity model to figure this out, according to the Brandon Hall Group. 0 shits given by anyone.

Comment by Steve Levy on March 27, 2015 at 2:31pm: Charney, you had me for a moment - almost got me going. I'll get you back.

Comment by Nicholas Meyler on April 14, 2015 at 2:05am: Comment by Nicholas Meyler on March 28, 2015 at 3:13amDelete Comment

I liked Markovian statistics, when I learned about it by reading Iannis Xenakis' great book "Musiques Formelles" in the original French. I could read enough French to figure most of that out, and the math was actually incredibly complicated, but still simple, in some ways, since Xenakis was rooted in the classic Greek traditions of B.C. mathematics, as well.

In any case, with an engineering degree and an esoteric interest in the music of Xenakis, I was able to muster together enough knowledge of statistics to figure out some purported "big data" statistics about recruiters (like a href="http://www.recruiterspam.com>" target="_blank">www.recruiterspam.com>;;) were totally bogus. In that specific case, proof involved dividing the number of emails by the number of people claiming they had been 'spammed'. I realized that even if there were only 11 software engineers complaining (there were more like 500 to 2000), then they still were only receiving a recruiting email every other day, at most. So, the claim "we get job offer spam every day" is clearly false. In reality, it was more like they were receiving recruiting-related email about once every three months to a year.

Yes! Lots of statistics are quite fake, if you look into them. A lot of people are so intimidated with math that they will believe any statistic thrown at them, if the claim is that the stats were prepared by a 'scientist'. Recruiterspam.com is one website which has been permanently debunked as a very weak fraud based on statistics that a 4th-grader should be able to see through. I'm sure I'm preaching to the choir on this one, but I really expected a whole lot more out of 'computer scientists' who actually made it past 4th grade.

On the other hand, Markovian statistics (based on extreme discontinuities in infinite-dimensional data-sets) is really "big data" statistics, and worth looking into. I only read the book on how Xenakis used it to compose some of the greatest music of the 20th century, but it was very stimulating.

As a chemical engineer, I had to study the statistics of boiling (fractionating) different liquid mixtures to separate them via distillation. Xenakis (who studied Chemical Engineering and Architecture, as well as Mathematics and Music) wrote some very great musical compositions based on the incredibly irregular rhythms of liquids boiling up through stacked trays in distillation towers. I remember watching a demonstration of this kind of process of distillation of Pyrogallic Acid when I was a student.