“The net of delusion is being cast ever wider, as we are bombarded with more and more information masquerading as knowledge, more and more material for the calculus, which far outruns our ability to sift it into truth and falsehood.” (Robert Skidelsky, January 2014, on economics and information).

The problem

As with any significant technological development or innovation, big data and analytics appear to be riding through domains of business and HR with the usual hype, fanfare, and confusion.  ‘I’m not sure exactly what it is’, ‘there could be good money in it’, ‘I sort of know what it is but not how to do it’, ‘seems important’, and ‘seems dodgy’ are all opinions one might continue to hear from a range of practitioners and academics. Nevertheless, emerging wisdom and previous forays into the terrain do provide hints about the challenges and opportunities swirling around the idea.

Can analytics be applied to HR?

Certainly many big firms providing diversified sets of IT services, such as IBM and KPMG, are treating analytics as a legitimate new frontier and desirable arm of their business. The issue here is if and how these concepts and services can be applied to HR, and the notoriously difficult to pin down ‘people side’ of business.

Certainly, something is at stake politically and professionally for HR. KPMG itself has argued in 2013 that ‘people are the real numbers’, and that ‘workforce analytics’ is the key for HR to move away from offering unconvincing generic models and towards more insightfully tailored solutions that carry employee value (and HR practitioners) further into the boardroom, alongside finance and other executive functions. KPMG note the many labels (HR analytics, big data, talent analytics, strategic workforce analytics) but define the phenomena as “the synthesis of qualitative and quantitative data and information to bring predictive insight and decision making support to the management of people in organization”. This sounds great in principal, but problems abound and change is needed; in the form of better data management/integration, new skill sets, and being able to ask the right questions critical to a particular context.

Is ‘Big Data’ the term we should be using?

The surface continues to be scratched. Think about the language used. Does ‘big data’ imply the devaluation of smaller data, or a concerning emphasis on quantity over quality? (Probably – and this is a frequently noted concern, but will it be heeded?) The word ‘analytics’ implies patterns, processes, trends to be analysed – but how do we know what we should be looking for, and what will be most useful when we find it out? Analytics and smart decision making has more of an established track record around operations, logistics, market trading, and supply chains.

But how far can analytical principles – concerning transport, utilities, weather, nature, financial markets and so on – be reliably and validly applied to making generalised claims about people, social phenomena and human resource practices? These are all fascinating debates, and undeniably our access to large quantities of data is unprecedented, given the current levels of social media usage and internet connectivity.  What is important is that what we can find out and achieve with the help of lots of data doesn’t become something mythical but stays firmly rooted in philosophy of science and knowledge considerations. A lot depends on views and interpretations of truth and evidence – which can seem dry at times, yet become central in relation to large amounts of complex data. Knowledge management has been vaguely alive within HR for decades; thinking about how to turn raw data into organised information and then finally into actionable knowledge.

Classic technological dualisms can also help shape our thinking about big data. Determinism (technology shapes us) versus voluntarism (we shape technology). Also, fishing for answers in large amounts of data (bottom-up/inductive) versus testing our pre-ordained ideas and beliefs to see if the data conform to them (top-down/deductive). Whether we interpret data more objectively in terms of how things are (rationally) versus more subjectively/socially in terms of how things could or should be (normatively).

Five concerns and predictions on HR analytics

In the spirit of bridging the gap between concerns about the nature of knowledge and truth and the potential value of big data analytics, I put forward my five most prominent concerns, predictions, and suggestions on HR analytics below.

1) Grasping that big data analytics is not new, and has its own informative history. As I noted above, many businesses such as IBM or KPMG are already relatively familiar with the idea of delivering analytics as a service component of their businesses. True, some of the tools have changed and the ease of data acquisition has increased, but some of the ideas have been around since the birth of statistics and even knowledge. Within the field of HR, selection has long been a predictive analytical science, and benchmarking and scorecard/dashboard metrics have long been a concern of stakeholders such as PWC’s Saratoga Institute. Systematic evaluation tools, human capital/asset accounting (alongside boardroom reporting guidelines to communicate results), talent management, and strategic workforce planning (SWP) have all dabbled in aspects of analytics for many years. Hence it is important to learn deeply from this business and HR past before trying to move forward.

2) What matters just as much as the data is what we bring to the data and what we do subsequently. Most statistical technique is concerned with trying to test predictions and fit plausible models to data – it is therefore about narrowing down and eliminating alternative explanations for what we are seeing, and understanding cause, process, and effect as plausibly and confidently as we possibly can. I would argue that the theories, values, control variables, contextual knowledge, models, strategies, HR practices, communications, evaluation models, and other frameworks that we use to frame data are just as important (if not more so) than the sheer amount or accessibility of it. I acknowledge there is some debate to be had about the value of iterative cycles of data exploration and ‘how do I know what I’m looking for until I find it?’ But on balance, at the very least, HR analytics needs to be guided by reporting on the problems we want to solve – be it detecting the ‘black box’ of HR-performance links, diversity discrimination patterns, employee engagement, or tracking talent pipelines/populations. In these ways, analysing big data along topical lines may help us to put flesh on the bones of trickier HR concepts.

3) There need to be continuing efforts to capture the principles of best practice and avoid the pitfalls of bad practice. This point really applies to anything practised in HR or business, and is probably fairly easy to have frank discussions about, but less easy to embed with tangible change. Most blogs and articles on big data do tend to note the ‘dos and don’ts’ that might be most important. This includes advice such as: asking the right questions, thoroughly mining/refining/interpreting, using the best available software tools, choosing a careful set of metrics, not focusing excessively on quantity, integrating numbers with qualitative sources and stories, not getting duped by misleading averages, linking analyses to actions, and so on.

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.  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? Incentives will also need to be lined up accordingly, to avoid malpractice and encourage the right developmental aspirations.

4) Experimentation with it at different levels and in differently defined ways will prove important. Another thing existing commentaries often discuss is the level of application, the way in which big data can be used and to what degrees. Big data may in fact end up as small data, or at least divided up or summarised in a certain way.  In HR, we need to think about whether we are using the data for describing, predicting, or explaining. Characterising data by type, metrics, scope, and source is always good evidence-based advice. Alternative models and explanations need to be tested and ruled out, cross-validated, and replicated before strong claims can be made. The scope and form of data analytics can be defined in a common scheme by volume (quantity), velocity (how often/much it accumulates over time), and variety (number/type of measurement sources). As with any business strategy, well-specified experimentation and incremental learning will prove valuable. We shouldn’t forget the role played by big data in investment algorithms and financial model-building that in part contributed to the last global economic crisis (e.g. The Black-Scholes model). 

5) Refining networks, agendas, and collaborations to formalise ideas and learn from existing business applications is likely to be valuable to ensuring best practice. My final point is simply one of structure and how HR can tie all this together into a meaningful, accessible, developmental agenda of genuine change. Big data analytics has history and interdisciplinarity; it is already at work in business logistics and finance, and even in HR analytics, computer scientists and engineers are often those who end up starting their own businesses. As with any technology application, it will be a meeting of business knowledge and technical savvy. Another point is looking again at IBM and KMPG (amongst others) and how they structure analytics offerings (i.e. services/divisions), to see how this can map and translate to HR practices and innovations. For example, KMPG use analytics in areas of finance, customers, operations, regulatory remediation, and financial crime.

Interestingly, employees are not mentioned, so…what would the HR equivalent look like? There is also the issue recently flagged up by McKinsey about how much ‘open data’ there is, and where to find it in terms of governments, cities, private sources, and other stakeholders. Thus the open data is another agenda to be considered. At the bigger picture level, it is worth seeing big data as potentially related to a ‘cluster’ of technological developments, new and existing (i.e. mobile apps, social media, e-HRM, HRIS). Technologies tend to coalesce and evolve in interrelated clusters to some degree, so missing this means missing the bigger picture. The HRZone.com Technology Toolkit published on this website represents a good bringing together of commentaries on a cluster of technologies, for example.

Conclusion

In conclusion, I am excited about the prospects of HR analytics based on big data, but perhaps not by the same things as some. On the one hand, I am somewhat excited by a wealth of potential insights to long-standing, intractable HR issues. On the other hand, I am equally excited about how to solve the problems raised above by ensuring that big data analytics are utilised thoughtfully, rigorously, and critically. We must capture value by trying to avoid genuine problems of information overload, and information falling into the wrong hands or put to wrongly interpreted ends (e.g. a bottomless pit of efficiency, cost-cutting, or excessively invasive consumer marketing). Answers may not be easy as we would like to think.

But as Leon Trotsky said, “Tell me anyway – Maybe I can find the truth by comparing the lies”.