Four Overlooked Components for Maximizing the Enterprise Value of Predictive Analytics
“Do you use ___?” (insert ‘machine learning,’ ‘deep learning,’ ‘cognitive computing’ or whatever else you fancy). I get asked that a lot. I normally reply “yes” just so we can move on from the buzzwords. There is a question behind the question though. I’m really being asked if my company is cutting edge, pushing boundaries, staying relevant and worth paying attention to. Predictive analytics has become the measuring stick of present technological maturity and remains the hope of future business transformation.
Bill Gates famously said, "We always overestimate the change that will occur in the next two years and underestimate the change that will occur in the next ten. Don't let yourself be lulled into inaction.” I would propose this is because humans generally don’t think exponentially. Predictive analytics will continue to transform how companies operate, but the changes will be exponential, incrementally building off their predecessors until one day our entire way of doing business looks nothing like it used to. The early stages of an exponential curve look pretty flat and uninspiring, but they foreshadow massive future gains. This phase is the “meat and potatoes” of predictive analytics maturity. It doesn’t sound exciting, but it is foundational. Here are four overlooked components required for reaping exponential benefits from predictive analytics in the enterprise.
Technology – Predictive Analytics Is Not Business Intelligence
Ready access to the data is the lifeblood of predictive analytics, but too often organizations view predictive analytics as a sub-domain of business intelligence (BI) rather than a different domain entirely. BI is primarily concerned with evaluating historical data trends and tracking performance while the aim of predictive analytics is to proactively use data to inform pressing decisions. The technical differences generally lie in the sequence of development and requirements for execution in a live setting.
Predictive models are developed on large stores of historical data and can be considered mathematical representations of a business’ prior experience. BI architectures support this phase of development well, but problems usually emerge when the model is tested in a live setting because this is the point where predictive analytics diverges from conventional BI development.
These models must be embedded directly into the live flow of data a business generates and pushed back for consumption, robust to any form of data that could emerge from source systems to function properly. Given the analytic is consumed at a defined decision point in a process, the speed at which this needs to happen is usually near real time. These requirements don’t exist in many BI applications, and a move to production generally just means copying whatever was completed in development into a new environment. It’s paramount the data engineering team understands the unique development sequence of predictive analytics and how the requirements for moving a model to production differ from conventional BI applications to work cohesively with data scientists.
Vision & Culture – A Spirit of Innovation
Most companies rightly identify the need for top-level executive sponsorship to support predictive analytics, but I believe success requires much more. Companies need to have a spirit of innovation and appetite for continuous improvement of even the most basic business processes. One of my favorite movies is “Remember the Titans.” The character Coach Boone echoes this idea on the practice field when he shouts to his football team, “We gonna change the way we run, we gonna change the way we eat, we gonna change the way we block, we gonna change the way we tackle, we gonna change the way we win!” Leadership should not only seek to empower front-line staff to bring innovative ideas to the table, but should establish it as an expectation. Calculated risk-taking needs to be encouraged, and the organization should objectively evaluate ideas according to merit rather than where the idea originated.
Process – Where Predictive Analytics Live
Once a culture of innovation is established, the systematic study of how business processes develop should become commonplace. Machine learning thrives in a decision-making or service-delivery process where there are decision points based on uncertainty. In this context the primary goal is to augment human intelligence, either directly governing the decision point or distilling large amounts of information into a form humans can consume and act on. Enterprise value from machine learning is maximized either when a few high-impact decision points are improved substantially, or smaller decisions made frequently throughout the organization can be improved marginally.
People – You Need a Bridge
Outside of the obvious roles in data engineering and data science, organizations should not overlook the role of the analytics connector when building predictive analytics capabilities. These people have established organizational rapport with managers and directors across business domains marked by a history of delivering quantitative results. They are adept at identifying business problems amenable to a predictive analytics solution. It is not a requirement that these individuals have an overly technical background as long as they can communicate business problems effectively to technical teams where the solution can be distilled into technical requirements by someone else. An analytics connector will develop into a conduit between the business and analytics team organically, ensuring resources are only allocated to the development of analytics solutions with a high likelihood of success.
To benefit from predictive analytics, avoid the mistake of focusing exclusively on the technical requirements when assessing organizational readiness for implementing them. Although they should remain the primary concern at the beginning, realizing high value from predictive analytics requires active maturation of data literacy within the culture and people of an organization.
Check out: Top Healthcare Analytics Companies