Saturday 27 April 2013

WANT TO KNOW WHICH HUMAN TO HIRE? (DON’T) ASK THAT ROBOT!

It’s not the sort of issue many would lose sleep over. Yet one that would have the souls of many-a-technocrat go uneasy. What do technology- driven recruitment softwares have to do with hiring the right talent? Some would say that human resource managers still make more logical decisions, irrespective of how true Moore’s Law has proven to be. Others would side with algorithms. One school of thought would have you believe that there is no better program written than the human genome that can predict the nature of job-seeking applicants and read more into the manner in which an interviewee does his necktie. The other obviously claims that “big data” can spot talent – having learnt from reading millions of data points in the past – and therefore the recruitment program on your server is a safer option.

What do I believe? HR representatives err. And why would we doubt the potential of bug-free, programmed-to-perfection recruitment softwares like SAP’s SuccessFactors, Oracle’s Taleo, SilkRoad's OpenHire, and others? A quick look at some recent correlations revealed by researches however, makes me doubt whether such softwares even make the grade when it comes to helping companies hire competent workers. And what justifies my lack of conviction in big data? That these shocking (to-say-the-least!) empirical findings are used in the very construction of many-a-logical assumption made by programmers of these software packages. Let me throw five of them at you. [The findings shared are those arrived at by studying “over 2.5 million granular management and supervisory data points”, presented in an April 2013 joint report by Evolv Inc., a San Francisco-based workplace performance solutions company, the Center for Human Resources at the Wharton School of the University of Pennsylvania.]

Read them and you actually might feel better about yourself if you are a “cheating, insincere” employee!

Finding 1: Like wasting time at office on social networks or by downloading torrents? Big data says you’re a good hire!

The first one is hard to digest for a Steve Jobs-like efficiencydemanding boss. If you are an employee who spends time on anywhere up to four social networking websites during the course of a day at work or likes to keep himself engaged by downloading softwares (to experiment with web browsers that did not come pre-installed with his/her work computer), then you will serve your employer better and longer as compared to your colleagues! Big data says so. We “humans” don’t. So get on with tweeting, tagging, posting, reworking your biodata, updating your OS through the WiFi, and downloading pirated versions of films that will get screened across multiplexes a fortnight later…all while your customers wait in line for their cheese burgers and hotdogs. Who cares? Remember – if you spend good lengths of your manhours on up to four social sites and downloading third party content over the company’s Internet connection, you will be statistically tagged a more loyal employee. Ask a human recruitment officer, and he’d rather have such a candidate booted before he gets hired! Long live big data.

Finding 2: Changed ten jobs in the past three years? Big data says you’re a good hire!

This one is balderdash. Research has revealed that a candidate’s past job-hopping record should have no influence on his future performance at work and his tenure at the new employer. In short, give no preference to a person’s stable history at a company, and ignore how many short-term jobs a potential employee has had.

Finding 3: Lack experience? Big data says you’re a good hire!

The third finding goes against conventional wisdom. Human recruitment officers would naturally attach some importance to past experience – in some cases maximum. And most descriptions of job openings state clearly the required minimum number of years (and/or months) of experience. big data analysts report that previous experience is in no way related to either performance or tenure on the job. Conclusion – make fresh college passouts the CEOs of all Global Fortune 500 companies and soon we would have trillion-dollar corporations floating about a dime-a-dozen. Not wishful talk. Big data says so.

Finding 4: Spent time in prison? Don’t worry. Big data says you’re a good hire!

This one reads strange. It is strange. But big data says it is not. Data analysis shows that an employee’s criminal background has no influence on his/her output, sincerity or loyalty at work. Contrary to what proponents of old school might recommend, actually, hiring those with a criminal record means hiring employees who are better performers when it comes to “customer-support”-related work profiles. Hire a criminal, and get your customer satisfaction levels up to levels never seen before, says big data. Strange for an old-schooler. Not so for believers in big data.

Finding 5: Dishonest? Big data says you’re a good hire!

In a study conducted at Xerox Corporation, more than 48,700 employees were interviewed in a six-month-long process to find out the honesty quotient of the employees. It was discovered that those who fell in the “Dishonest” personality type, were better candidates (than the "Honest" lot) to be hired in the Sales & Marketing department! Strange (rather, shocking!) coincidence many would reckon, but that is what big data is.

Problem with Information Technology being used to understand and hire talent is that big data often fails to distinguish between ‘signal’ and ‘noise’ – and that it’s the ‘T’ which assumes more importance over the ‘I’. Two cases that Prof. Prof. Peter Cappelli of Wharton has written about, illustrate the big problem. First, a Philadelphia-based HR executive told Prof. Capelli that he had applied anonymously for a job in his own company to test whether the hiring software was errorfree. He didn't make it through the screening process! Second was an email that Prof. Capelli recevied. A company received 25,000 applicants for an engineering position. The recruitment software concluded that not one candidate was qualified. Reason: none of the applicants had a certain title in their previous jobs. Why? The title was unique to the prospective employer! This example underlines well the problem with big data. With keywords being critical to avoid being weeded out by the software, screening softwares reduce the meaning of 'Information' furnished by applicants to nothing! Parrot all the words mentioned in the job description skillfully, and wolla, you will find yourself in the next round.

Data Analytics is big business today. But for both those who create data and those willing to pay for it, it is critical to understand that data is only indicative. To expect softwares to choose the right job candidate (and therefore the right team) or the right product or the right market would be wishful thinking. Number crunching and providing indica

tive data tables and charts is fine. But irrespective of high the chance of an error with a human in charge, big data cannot and should not be used as an alternative to human expertise when it comes to final decision-making. Not today. Not in another 50 years.

We can safely recommend that even advocates of big data should have the patience of a saint when it comes to recommending the replacement of human recruitment ‘decisionmaking’ officers with big data servers and PCs. They should hold their peace till the very failure rate of big data projects fall (from the current 45%, as per a survey by business-software firm Infochimps Inc.). Let us look at an example. Catalyst IT Services, a Baltimore-based technology outsourcing company, is one firm that is trying hard to do away with the cumbersome, time-consuming approach to hiring. So what does it do to achieve this end? It asks candidates to fill out an online assessment form – something which a company like Google has replicated. Using this online form, Catalyst collects many bytes of information about each applicant. The secret is through understanding ‘how’ they answer questions, rather than ‘what’ their answers are – right or wrong. For example, when a complicated mathematical problem is put before the applicant, the program analyses the ‘problem-solving’ skills of that individual, rather than how correct the answer is. This, experts at Catalyst claim, helps assess whether some candidates are suitable for problem solving tasks. Sounds like a fair measurement of applicant attributes, but the interpretation is highly subjective. A candidate may spend hours behind a problem – but does that necessarily mean the dominant attribute in his work character is not ‘obstinate’ but ‘labourious’? And how can a software decide whether a talented, intelligent candidate who solves a problem in a matter of seconds is even a degree less hardworking than one who fails to arrive at the solution, despite having spent hours on that one problem? So we say, statistical analysis in this respect is only indicative. Technology can be used as a facilitator, not as a decision-maker.

My recommendation to big data advocates is – first allow big data to take care of the talent hunt for itself (demand for talent in this big data is expected to outstrip supply by 60% by 2018, as per The McKinsey Global Institute). Then we will bother HR specialists to pay more attention to big data robots. Technology isn't a substitute for human expertise.

The Moneyball approach does not work every time. Let robots do the calculations. Let humans decide on the results. Should you hire A or B? Leave that to big data and you might have a case where a job-hopping, dishonest and quiet neurotic with a criminal record is hired when a socially networked, expressive extrovert is what the job demanded!

No comments:

Post a Comment