I have long believed that fact-based decision making is better, but ironically there has been little evidence supporting that belief. The Economist Intelligence Unit (EIU) conducted research on this very question in the fall of 2014, and concluded that “data-driven” businesses are more profitable, are better at knowledge-sharing, risk management and operational efficiency, and are more likely to have a culture of creativity and innovation. All pretty good things, regardless of what sector or industry you operate in. So how do you move an organization towards being data-driven? And why is it so hard to do this?
The EIU also conducted a webinar on how to become a data-driven business with a panel of experts, and I encourage any of you interested in this topic to watch it. There is a lot of rich content in the interesting discussion between Philip Evans (senior partner at BCG), David Trimm (CIO at Hertz), and Florian Zettelmeyer (faculty at the Northwestern University business school) moderated by Riva Richmond (senior tech editor at EIU). A few ideas really resonated with me that I want to explore.
Being “data-driven” is really about creating an analytics culture: testing ideas with data. The panelists all agreed that the biggest barrier to organizations moving in this direction was not technical, but cultural. Which means that it is fundamentally a managerial problem. Why? Because managers and leaders establish an organization’s culture by how they act. In order to change the culture, they have to be willing to lead the way by changing their own behavior, both because they are role models and because they have more leverage in the organization due to their positional authority.
Companies that have been successful at moving towards an analytics culture have some key elements that seem to support their success:
- They look at problems from an end to end view, in context of the bigger picture and not siloed – in other words, they adopt a systems thinking perspective, much like the value stream approach of Lean
- They use small, cross-functional teams to tackle problems through innovation and experimentation, with short cycle times for more effective learning (because there is less delay in the feedback loop about whether what they tried achieved the desired outcome)
- They used other mechanisms to diffuse the learning – taking it to scale by operationalizing it (a.k.a. standard work)
While there are still barriers to being data-driven on the technical side, namely with data itself – multiple versions, not being timely, being messy – these are technical problems that are solvable. And I would argue that solving them is neither necessary nor sufficient to moving towards an analytics culture. It is possible to have a data-driven culture without having first solved all the problems with data. Why?
Because, while “analytics culture” is new terminology, the idea of being data-driven isn’t new at all. It is really about having a learning culture, one in which members assess ideas against facts and data, and question their own and each other’s thinking. Earlier versions of this concept were introduced by Chris Argyris who coined the term “double-loop” learning as the ability to reflect critically on one’s own behavior. In other words, self-reflection on one’s own mental models is key. In his 1991 HBR article “Teaching Smart People how to Learn,” Argyris argues that one way we develop this ability is from experiencing failure, and learning how to learn from failure. Successful, high achieving people often have spent their lives going to school to get book smart and are very good at problem solving. But they often have not experienced very many failures. So they are not so good at questioning their own mental models, and in fact their very success reinforces the mental models that makes it hard for them to learn – their identity is based on their success, so even admitting a failure is difficult, let alone learning from it.
For real learning to happen, we don’t have to wait for good data. Instead, we have to be willing to recognize when our own behavior is not optimal for achieving our desired outcome, and adjusting it accordingly. This is much easier said than done, and requires self-reflection and being willing to admit to being incorrect in our thinking.