Analytics Hype: The Next Wave in Big Data Backlash
When considering big data technologies, don’t confuse data collection with smart, operational use of the collected data
While doing some morning reading recently, I came across an article that focused on how not to do big data in healthcare. The catalyst for the article was an ill-targeted campaign from a new health provider about proactively addressing a health problem the author didn’t have. The article then goes on to broadly attack big data technologies. Raising issues with what this health provider did was totally fair game; however, the criticism was way off the mark by simply characterizing the error as a big data problem. One should be careful not to confuse the success of big data infrastructure technologies with how the collected data is then applied to a given business problem. Getting the right data is not enough, though; it has to be applied in the right ways, and lumping both considerations into the same bucket is not helpful.
Revisiting the big data backlash
I’ve discussed the big data backlash previously,* and it seems as if we are now moving into Backlash Part Deux. The first backlash was really an infrastructure thought, largely driven by at-risk, traditional software vendors and consulting firms that felt threatened by big data. They were essentially complaining about the market and the evolving big data ecosystem moving away from their existing technologies and practices. The second backlash feels a bit different than the previous version. It is focused on analytics and the conclusions drawn from big data as opposed to the big data infrastructure that was the focus of the first one. As with the first part of the big data backlash, being skeptical of these technologies is a good thing. The analytics and business outcomes that are the subject of the second backlash can benefit from a skeptical eye as well. That said, being sloppy or ill considered is equally problematic in both parts of the backlash. Big data infrastructure technologies and the application of analytics need to be understood as related but discrete things. For example, in the healthcare-related use case discussed in the aforementioned article there is a broad amount of data related to an individual’s health status that can be challenging to capture. Collecting the data can include looking at a wide range of disease states and outcomes; undertaking an enhanced, deep, and highly accurate mining of a health record; and gaining an improved understanding of inputs and outcomes in terms of efficacy. It can also include looking at patient cohorts in ways that were not possible previously because the data was either too varied, too complex, or too processor intensive. Big data technologies deployed at an infrastructure level help address all these challenges. Keep in mind that the flexibility to use those data sets, rather than handling their sheer size, makes big data technologies attractive. Now imagine that a medical provider takes that rich and actionable data set and figures out what disease states can be effectively managed by early and aggressive interaction with its patient population. Fair enough. But the failure identified in the article was over what happened next—namely, how the information was used, not how it was captured, generated, or analyzed. This failure was not a big data problem; it was how the insight from the data and analytics was operationalized to the end consumer. Calling it a big data failure misses the point, and even worse, it draws attention away from the core issues at hand.
Shifting to business outcomes as the measure
As we progress into 2014 and 2015 the emphasis on big data technologies is expected to shift from being infrastructure-centric to being how they are measured in business outcomes. That shift is likely going to uncover gaps in the analytics maturity of many organizations and in how they operationalize these technologies. This gap, moving from data to outcome, may be harder than standing up the big data infrastructures, and it requires being able to focus on each step of the manufacturing process of driving outcomes. Not unexpectedly, the maturity and understanding of how to use the big data technologies to collect and process data is advancing much more rapidly than the ability to effectively operationalize what the big data technologies give us to work with. In other words, the analytics and new operational models enabled by big data technologies are maturing at a far slower rate than the big data infrastructure technologies themselves. The thinking around how to close the loop from information capture and generation to driving action to facilitate the right need deserves much more attention than it is currently receiving. The article about the health provider's campaign mentioned previously raised valid concerns about outcomes mattering, but the article's argument fell short in not being precise enough in explaining how to get to those outcomes. As 2014 unfolds, I foresee more of such proverbial perfectly good babies being thrown out with the bathwater, so look for more to come on this topic. As always, please don’t be shy about sharing your thoughts or questions in the comments. * “Backlash Against the Big Data Backlash,” podcast, The Big Data & Analytics Hub, January 2013.