4 Ways to Democratize Data Science in Your Organization

Source: Harvard Business Review – Many organizations have begun their data science journeys by starting “centers of excellence,” hiring the best data scientists they can and focusing their efforts where there is lots of data. In some respects, this makes good sense — after all, they don’t want to be late to the artificial intelligence or machine learning party. Plus, data scientists want to show off their latest tools.

But is this the best way to deploy this rare resource? For most companies, we think it unlikely. Rather, we advise companies to see data science both more strategically and broadly.

Consider strategic data science. While organizations have relatively few strategic problems, they are of special importance to the company. Even though there may be relatively little data to analyze for strategic problems and “big swing” decisions, companies should bring everything they can to such issues. Data science provides much more value than just big data algorithms — from more clearly formulating the problem, to analyzing what “small data” is available, to experimenting, to creating great graphics. The potential to come up with better insights using data science is enormous. Further, since senior managers must ultimately lead the data science transformation, engaging them in the data helps them more clearly see the benefits and better understand what they must contribute to the transformation.

But data science also must be democratized broadly. If data science is to be truly transformational, everyone must get in on the fun. Restricting data science to only the experts is a limiting proposition. Data science programs that focus on professional data scientists ignore the vast majority of people and business opportunities. For instance, organizations are loaded with problems and data-driven decisions that can be solved and made by small teams of knowledge workers, middle managers, and partners using small amounts of data in two to three months. These individuals, being at the front lines of the organization, already understand the business and don’t need to be taught it as data scientists do. And vendors of various types are now offering a variety of new tools that ease or automate many aspects of data science, including massaging data, creating algorithms, and creating code to deploy a model into production.

While the idea of an organization-wide data science transformation sounds overwhelming, there are ways you can get started. Based on our consulting, conversations with senior leaders, and research, we recommend the following interrelated steps to make data science more strategic and democratic in your company.

Focus on problems with the highest level of strategic benefit.

As previously noted, most organizations focus their data science efforts where they have the most data — even if they don’t mean to. Companies should consider a full range of other criteria, two of which are most important.

First, they must think of the long-term strategic importance of the problem or opportunity. Consider two options at a mid-sized media company: Option 1 involves looking for insights that deepen the user experience using data generated by engagement with its apps; Option 2 involves using data to inform a bid for certain licensing rights, something that comes up every couple of years. There is plenty of data in support of Option 1 — it is certainly important. But even as there is relatively little data in support of Option 2, it is strategic. Bidding too low and losing can do immediate and long-term harm; bidding too high takes away from profit.

Second, they also must consider the probability of project success. By “success” we mean delivering business benefits of equal or greater value than its proponents promised. It takes a lot to meet this standard, from developing a new insight or algorithm to convincing people to act on or use it to building it into the company’s processes and IT systems. Indeed, developing the insight or algorithm is often the easy step, and many such models are never deployed. Sponsors of potential data science projects should make a stone-cold sober evaluation of these factors. While there are no set answers, we think that evaluating projects in this way will lead them to do more small data projects and more-carefully chosen “moonshots.” DBS, Southeast Asia’s largest bank, has largely given up on moonshots after an early failure but is pursuing other small data projects aggressively throughout the bank. Moderna Therapeutics, the creator of a Covid-19 vaccine, has also eschewed moonshots in favor of less ambitious AI and digital projects.

Democratize data science in the organization.

We sometimes ask companies, “Which would you rather have: a newly-minted PhD data scientist or 20 people who can conduct basic analyses in their current jobs?” Almost all opt for the latter. It leads to our second recommendation — namely, develop “citizen data scientists.” There are plenty of good business intelligence tools and, increasingly, automated machine learning tools make it possible for good business analysts to perform quite sophisticated analyses. Royal Bank of Canada, for example has had great success in this regard.

Some companies, such as Eli Lilly and Travelers, take this advice even further. They provide data and analytical literacy programs for all their employees — and much of the content is tailored to the employee’s level and business function. They view it as an essential capability of their employees to understand different types of data, what can be done with them, and how analytics and AI can enable competitive advantage with data. Finally, of course, companies should look for basic data science skills in all their new hires, for all positions.

…