Why Aren’t You Getting Value Out of Your Data Science Team? Reason #1 is Below….
I’ve seen several articles on LinkedIn highlighting reasons that Data Science projects fail. I’ve also reviewed articles from HBR, McKinsey and Gartner citing staggeringly high rates of project failure.
Over the past 20 years, I have worked for top vendors in this industry including SPSS, SAS, KXEN (now SAP) and IBM. I’ve seen firsthand how and why organizations do not get the full benefit from their data science initiatives. Over the next few weeks, I plan to highlight some of my favorite reasons.
Reason #1
Lack of Direction – no clearly defined strategic mission or business objective tops my list.
Lack of Strategic Mission. I’ve worked with a few businesses that decide that they want a Data Science team before they have specific questions of their data. They believe that they can “Monetize their data,” but they don’t know what they are going to do to get value from it. Without a clear definition of the problems that they will solve (their Mission), these teams are often placed in IT where they struggle to have reach into the business to find relevant impactful projects.
Bottom line: If there is no Strategic Mission and no line of business can directly benefit (and "own" the team), the team shouldn't be launched.
Lack of a Well-Defined Business Problem. 20 years ago, I had the request “Tell me something that I don’t know.” This year, I was asked to “Do a deep dive on my customers.” Neither question will help monetize data.
Rather than “Find something interesting” or “Do a deep dive,” it would be more useful to begin framing a business problem such as “Who are my high value customers?” or “What customers are not coming back?” With these questions, I can begin defining a Business Objective to determine how, if we knew this information, we could impact change.
If you recall CRISP-DM (Phase 1 is highlighted in a circa 1999 slide deck screenshot above), the first step in any project is Business Understanding. While this may seem obvious, it is often not done, or not done well. Working through this phase of CRISP-DM, we can develop a roadmap for the execution of the project. I have found that, on occasion, just working through this phase identifies a number of functional issues that must be resolved before the project will be able to have an impact.
Towards Better Data Science. Scientists run experiments, collect data, and validate hypothesis. Data Scientists collect data, run experiments, and reject null hypothesis! Not every data science project will yield valuable insights. But we can do better by simply making sure that the team has a Strategic Mission and a well-defined business problem.
Stay tuned for next week’s edition where I will highlight the #2 Reason.