The Potential Perils of Big Data in HR


In my last blog I argued that big data offers potentially powerful tools in terms of creating insight, connectivity, and personalization. However, before unreservedly jumping on the big data bandwagon, I think it is important to consider the perils of big data in HR. Three in particular come to mind: the fallacy of generalizing from aggregate to individual, over-relying on data while ignoring relationships, and depersonalization of the workplace and workers.

As I teach my Masters in HR students in the HR Metrics class, I think that analyzing aggregated firm data can lead to better decision making than simply relying on gut instincts or "the way we've always done this." However, some are beginning to argue that using this data can help managers to identify turnover risks and suggest the strategy to reduce this risk. Some suggest an artificial intelligence-like system that points managers to the individuals at risk of leaving and tells them what to do to prevent departure. While a laudable goal, making links from the aggregate to the individual may be rather problematic.

The fact that data gathered within the organization for the past 5 years suggests that a pay increase reduces turnover does not necessarily mean that offering a pay increase to a top performer will reduce his/her likelihood of leaving. Such an approach assumes (a) the aggregate data accurately represents the individual, (b) the individual is currently at risk of turnover, and (c) pay is what the individual wants to stay (as opposed to promotion, development, recognition, etc.). Arguing that a pay increase across the organization (aggregate) will reduce turnover across the organization (aggregate), is clear. But going from organization (aggregate) to top performer (individual) is not nearly as easy.

Researchers recognize that data never results in perfect prediction (to the individual) but rather probabilities (the probability for the individual staying will be higher given a pay increase). But as my old marketing professor used to say to illustrate this fallacy "On average, every American has one male testicle and one female breast." Thus, using big data can help to identify those that "have a higher probability" of turning over and suggest strategies that "have a higher probability" of reducing the decision to leave, but neither prediction can approach perfection. This requires filtering the data through an additional lens: relationship.

While data can be turned into insight that leads to better decision-making, this should not suggest a robotic as opposed to relational process. Big data provides information to highlight problem areas, and suggests alternative strategies for dealing with them. However, the selection of the strategy has to be based on the personal relationship between a manager and his/her employee...a conversation to explore whether or not the problem exists, and what might be the best approach for dealing with it. Then the aggregate provides insight into "potential" problem areas and "potential" solutions, but the translation to the individual has to be done in the context of relationship. Managers who know their employees can be pointed to exploring their concerns, and developing solutions unique and effective for the individual rather than applying one-size-fits-all solutions to a predetermined problem.

Finally, I fear big data strategies can lead to depersonalization of the workplace and work relationships. Humans are not machines and their decisions are not completely determined by environmental contingencies, but stem from the interaction of their situation and their free will. Some advocates of big data seem to suggest that people can be managed most effectively by precise data determined decisions. They seem to ignore the inherent uniqueness and dignity of human beings. Years ago critical organization theorists decried the treatment of people like "a number." Let's hope that big data does not simply expand this to treating people like "an aggregate of numbers."