Big data guru Tom Davenport predicts we’ll learn to love the numbers
With the recent proliferation of metrics into mainstream media, it sometimes feels like you can’t get two businesspeople together without a self-professed big data expert butting in. But one man has been there longer than the rest, and can reasonably lay claim to the title of most respected and level-headed commentator on the topic.
Tom Davenport, professor of information technology and management at Babson College, has been tracking trends in the way businesses use metrics and broader technological tools for more than two decades, and his sober analysis of the rise of big data has helped plenty of businesses make sense of the hype. Over the course of two 2013 books – Keeping Up With the Quants and Big Data@Work – he provided plenty of inspiration for number-hungry HR professionals, as well as a stream of commentary via Harvard Business Review, where he is one of the most popular bloggers. People Management asked him to sort the truth from the hyperbole.
Does HR really ‘get’ metrics?
There has been slow and steady progress on metrics, in part because of the rise of HR information systems. They have the transactional abilities to capture all that data, which has encouraged an existing trend towards more measurement of performance. It’s also possible, as a result of these systems, to capture a lot of data around employee behaviours – how much vacation time do they take, how often are they absent, etc – that might be related to other performance variables.
But HR has been slow to do anything with those metrics other than report on them, which is the least valuable approach to analytics – it’s backward-looking and it doesn’t tell you anything about the implications of the numbers or why they might be going up or down. Metrics are a good thing in general, but you have to do something significant with them for them to be really useful.
Are we closer to being able to put a ‘value’ on people in organisations?
Yes, and people do it all the time in the sense that the majority do some sort of performance assessment. So you can look at how well that person is performing as judged by their supervisor, their peers or whoever is doing the assessing. That’s sort of a proxy, I guess, for value. In a way, we make an estimate of the value of people by how much we pay them, so maybe you should multiply compensation level by performance level and get a rough indicator of how much value people are providing to the organisation.
Does a data mining approach work?
In general, you’re much better off having some sort of business objective in mind. You can spend an awful lot of time trying to decide what sort of insights are useful, but it’s much more valuable to say ‘we think we have an attrition problem, so let’s identify who is leaving, and what were the predictors they were going to leave?’ Or if you know you’re going to hire a lot of people, that would be the time to understand the factors in a person’s background or during the hiring process that predict he or she will be a high-performing employee.
HR has been gathering employee satisfaction and engagement information for a long time. But it goes up, it goes down – they don’t really do anything about it. It’s possible now to understand if employee engagement in a retail store or a bank branch really drives performance.
Is this all just a rehash of the sort of ERP systems most HR directors regard with suspicion?
The basic transactional systems haven’t changed all that much. They’re easier to use than the traditional ERP systems. But that is transactional data – the real action and value comes when you take that data and start to relate it to other things, to predict the likelihood of certain employees’ attrition, to understand how length of service drives performance.
If you want to cancel your credit card or mobile phone policy and you’re a really valuable customer, a company will try to talk you out of it and offer you a special deal. Some companies do similar things for employees who say they want to leave: do a quick assessment of how valuable this person is, and if they are valuable they will make heroic efforts to keep them.
Is a data-driven approach less humanistic?
I suppose you could look at it in a couple of different ways. It does reduce a complex human personality and background to a series of numbers. But on the other hand, we now know humans don’t make very good decisions about people. They tend to seek out and hire those who are much like themselves, to be unfair in evaluation processes, they don’t make good judgements in interviews and so on. If we change our HR processes to be fairer, more objective and more accurate it seems like a good thing for humans.
What will data never do?
It will never replace the need for some level of human interaction before you hire someone. I have seen some companies that tried, for low-level employees, to do everything through data and analytics and not even meet or have a conversation with the person. It’s been quite disastrous. The fact that companies have done it means they probably didn’t take those employees very seriously in the first place. But I do think that augmenting human experience and intuition, which often isn’t very good, with something measured and analytical, is always going to be a good idea.