Evolving From Descriptive to Prescriptive Analytics (Part 1): Leadership Support
Review the essential first step of acquiring leadership support to shift focus from descriptive to prescriptive analytics.
Join the DZone community and get the full member experience.Join For Free
Over the last 12 months, we transitioned our business from mostly looking in the rearview mirror to seeing and responding to the road ahead. This is the first in a series of posts that describes our journey from descriptive to prescriptive analytics.
You might be in the same situation we were in a year ago: we didn't have experienced data scientists who knew our business, data was in disparate sources often without a common linkage, and data owners guarded their data and were reluctant to share. Those weren't exactly conditions for success.
On the bright side, we had a business intelligence team that knew where the data was and often had access to it or knew who did. They also understood how to turn data into insight. We had plenty of data and benefitted from being in a company that builds all the tools needed in the areas of hybrid data management, unified governance, and data science. These ingredients gave us confidence we could succeed. After years of enhancing existing analytic dashboards with new data and visualizations, it was time to evolve: we needed to not just surface what already occurred but what was likely to occur next. It was time to predict the future so that we could change the future.
Gaining Leadership Support
Our first step was getting leadership to agree on the value of focusing on predictive and prescriptive analytics, which is important for a couple of reasons:
- It means that you can turn your focus from pure knowledge and understanding toward decision-making and action. That does not mean we no longer provide descriptive analytics, but it comes as a byproduct of the journey to operationalizing machine learning for the organization.
- Leadership support lets you invest in skills and potentially tools, as well. Thankfully, IBM already has an incredible portfolio of the tools we'd need so our team's focus was skills. Whether by hiring these new skills or by evolving the skills of your current team, it takes investment.
To sell this vision, we identified a use case related to an important business metric with organizational focus. Our proposal explained how our team would move from explaining what happened and why to identifying the leading indicators of the outcome and using machine learning to suggest actions that could impact the outcome in advance.
Like many of you, we're focused on delivering a great client experience, which we measure through the Net Promoter Score (NPS). Typically, leadership would ask the team to create or enhance a dashboard in order to monitor NPS and provide descriptive analytics to understand what drives NPS. While helpful, both would be insights that came after the experience itself.
Our goal is to shift that mindset to predict which current client engagements are likely to result in a negative experience so that we can take action to change course and deliver a better experience. On the path to prediction, you'll need to acquire and analyze the data, which, if you have the right tools, will make it easy to provide monitoring and analytics. Ultimately, you want to deliver familiar monitoring and descriptive analytics but with the added breakthrough value of improving the outcome through predictive and prescriptive analytics. Doing so can be a game-changer for your team and company.
Once we had leadership support to make this transition, we needed to build a team that could accomplish our goals.
Building a Machine Learning Team
Ideally, you'll find data scientists who combine programming, statistics, and business knowledge, but those people are rare. More likely, you'll need to aggregate those skills across a wider team. That's why we like to say that machine learning is a team sport.
To start, we augmented our business intelligence team with early professionals with Master's degrees that covered machine learning. All had programming skills in at least Python. In fact, almost every resume had Python skills, which was our choice for a core language (more on that in an upcoming article). Other languages like R or Scala can also suffice.
We proactively set up a network of experienced senior data scientists. Look for these in your company, external networking sites, or perhaps from your preferred vendor. For example, IBM has launched a data science elite team to work directly with client teams to deliver data science outcomes; reach out to them if you're an IBM client.
Or maybe you have a team that's been in your company or industry for a while and they know the business. If so, you should still set up relationships with business partners in each part of the business. Our early professionals, in particular, benefitted from this. Business partners have been very helpful to our team understanding the business, identifying use cases, and co-creating solutions.
Within a couple of months, we had the foundation of a team and organizational commitment to unlock the power of machine learning. In our next article, we'll discuss building and improving machine learning skills so our team could take on the challenge of predictive analytics.
Published at DZone with permission of Chad Marston, DZone MVB. See the original article here.
Opinions expressed by DZone contributors are their own.