The rise of Big Data as an HR topic should, I would hope, already have been noticed. It is one of the current ‘new’ topics that every journal, blog and conference is eager to cover (I would recommend, for example, that you read Jonathan Kettleborough’s excellent series of articles at Training Journal). That coverage seems to be divided into two main categories:
- This will change everything/is the future/will revolutionise the workplace – and (of course) get HR that place at the top table
- There are a few important things that HR needs to grasp or this could end in tears. Or court.
I can’t help but think that if HR is going to achieve the technically-supported nirvana implicit in the first of these, it might need to read as many articles as possible in the second category as a preliminary. Why? Perhaps the shortest answer to that question is ‘raw materials’.
Mastery of a discipline (accepting the use of the term for HR) requires many skills, but also a sure-footed understanding of the raw materials and how to handle them. The raw materials of a Human Resources Manager are primarily, as the job title alludes, people. (Whether ‘human resources’ is a devaluing or derogatory way of describing them – or should I say ‘us’? – is a separate but not entirely unrelated point.) The raw materials of Big Data, by comparison, are more explicitly ‘in the title’: data. Quantitative measures. Numbers.
Using Big Data – and the first of those three words tends to be unhelpfully omitted – is fundamentally about analysis and analytics: interpreting, questioning and validating data so that relationships, lessons and conclusions can be drawn. The promise of big data is the idea of unequivocal truth: that facts will be revealed using something as pure as numbers, and the unintended consequences of unconscious bias will be banished. Without that mastery requirement, however, this may turn out to be a case of ‘no so fast …’
On behalf of the comparatively unattractive everywhere, let me illustrate the problem of unattractive bias with a quote from an article in The Atlantic by Don Peck:
Examples of bias abound. Tall men get hired and promoted more frequently than short men, and make more money. Beautiful women get preferential treatment, too—unless their breasts are too large. According to a national survey by the Employment Law Alliance a few years ago, most American workers don’t believe attractive people in their firms are hired or promoted more frequently than unattractive people, but the evidence shows that they are, overwhelmingly so.”
Although I’m not aware of research to prove it, I would hope that most of us would prefer that it would be the C-Suite that was going to get uglier but, without diligence and understanding, the application of Big Data in unskilled hands could make many things uglier. As HR is increasingly called upon to provide evidence, to argue its case and to ‘step up’ strategically, HR will – understandably – seek ways to derive and formulate compelling narratives that will demonstrate both the value of its arguments and the value of itself. But the difference between a good story and a detailed analysis depends on just as big a difference in raw materials as that between Human Resources and Big Data.
Storytelling – whether it be fiction or non-fiction – depends on careful selection, exclusion and manipulation for its impact. Whatever doesn’t add to the story is either excluded or adjusted. To achieve its potential value, however, the approach required to Big Data is not to use it to prove a ‘story’ but to investigate the data to find out if it has a story to tell. It may not. Its failure to provide one may be disappointing, but that’s no excuse for imposing one. (We are supposed to be confronting and eradicating unconscious bias, not finding it the mental equivalent of a back door.)
Numbers are not destinations so much as signposts. As a number of authors on the topic have pointed out, a statistical correlation is not the same things as cause and effect: confusing the two leads to false logic. (As HR Data Compass has shown, a decline in the number of pirates correlates with a rise in global warming. Buzzfeed provides more examples, showing – amongst other things – that lazy thinking equates both ice cream and Internet Explorer with murder. Only true if someone comes between my Haagan Daaz and my browser of choice, but I do not constitute a statistically valid data set.) Data needs to be checked diligently for validity, and the relationships between different facets require further inquiry: just because both A and B appear to be happening doesn’t mean that A caused B. Coincidences are just as real as causes, but confusing them means that you are making myths rather than telling stories.
Let me provide a headline-grabbing example. In the early 1990s, I sat in a seminar about equality and diversity organised by a University HR Department. While the University proudly thought of itself as liberal and egalitarian, this was a time before anti-discrimination legislation covered sexual orientation. The University’s Personnel Officer (as was her title at the time) explained that there would be no voluntary extension of policy to cover this, as this would mean that gay men would turn up for work in women’s clothing. She was, I remember, rather surprised when a Professor of Mathematics took issue, explaining that the overwhelming majority of male cross-dressing occurs among heterosexual, married men. (Statistics courtesy of GoogleBooks for those who may be interested.)
Nor is this the only potential pitfall of applying Big Data that HR should beware. Even diligent analytics may reveal ‘problems’ that don’t truly merit auctioning: to introduce a medical analogy, there is a big difference between a bruised calf and broken leg. Identifying the latter requires a remedial response; identifying the former should prompt a (brief) review as to whether treating the ailment is a worthwhile use of resources. (To continue the analogy, there is also potentially the ‘Internet medical symptoms dictionary/self-diagnosis’ syndrome aspect: symptoms (eg low engagement, high turnover) can run in fashion cycles just as much as solutions (eg employee surveys, process reviews). Big Data is not about arriving at a conclusion and then using the data to prove it.
As CIPD’s 2013 Research Report, Talent analytics and big data – the challenge for HR (download as PDF), argues, there are many hurdles for HR to jump: a preference for ambiguity and complexity is just one example. But there is a skills issue for HR. Analytics is a discipline and even if age isn’t everything, one with a longer-established history. As Tom Calvard wrote in a piece for HRZone in January 2014:
More broadly, statistical, methodological, and evidence-based training for HR professionals will be just as relevant as ever. Even more broadly, there are implications for higher education programs, those looking at talent pipelines, future skills, and refined forms of professional development (e.g. Royal Statistical Society (RSS); CIPD; SHRM). Concepts like validity and reliability will need to be revisited, as well as statistical techniques, such as meta-analyses, factor, analyses, latent structural models, correlations, and regressions. How many working in HR can reliably and effectively master all of these in relation to big data sets?”
I can’t help but think that Big Data is an issue that will not go away: not just because it is partly driven by technology – one of the inescapable drivers of our times – but because it is also partly driven by a business requirement for evidence and a zest for competitive advantage. Big Data extends the promise that it can provide ‘proof’. While CIPD express concern that analytical skills will raise an issue about out-sourcing and ownership, sooner or later HR functions will need either to grasp the nettle or delegate the grasping to someone else. If it doesn’t want to be badly stung in doing so, an acceptance that skills in quantitative methodology and statistical interpretation are a vital foundation needs to be arrived at.
As Tom Calvard’s article points out:
One great book that still stands the test of time is ‘How to Lie with Statistics’ – a simple book with profound implications, on sampling, biases, graph formats, averages, and post-hoc justifications.”
The meeting of HR and Big Data should be an opportunity to prevent lying with statistics, not an opportunity for statistical misrepresentation to conquer a whole new arena.