How Platforms and Collaboration Help ‘Model-Driven’ Data Teams Excel

Few organizations are taking full advantage of data science-generated models, even though the models could significantly support business needs, according to a recent survey from Domino Data Lab. The resulting report, titled “Key Factors on Journey to Become Model-Driven,” distinguishes what it defines as “model-driven” companies—those that embed algorithmic-driven decision-making (or data science-created models) at the core of their business. In contrast, “laggard” organizations struggle to manage the few data science models they have, as well as quantify their business impact. More than 250 data science professionals took part in the research. This slide show features highlights of the survey, with charts provided courtesy of Domino Data Lab.

Only 14 percent of organizations are considered “model driven” in deploying data science. Two of five are categorized as “aspiring,” and 46 percent are “laggards.”

Just three of 10 companies have more than five data science models in production. Nearly one-quarter said they cannot accurately count their number of models, while 14 percent admit that they have “none.”

More than 80 percent of model-driven organizations are able to develop and deploy their data models within 90 days. Just 60 percent of laggard companies can make the same claim.

When asked whether their business owned a data science platform, 86 percent of respondents at model-driven companies said they did. In contrast, just 29 percent of respondents at laggards said their organization owned one.

Nearly three of five survey respondents expect their data science department to at least double in size in 2018. Only 12 percent said their department will not grow.

Overall, 91 percent of survey respondents said data science is contributing to their company’s business innovations. However, only 9 percent can quantify the business impact of all of their data science projects.

Only 37 percent of survey respondents said their business is driven by data and the work of data scientists. But one-half said they expect the impact of data science on their lines of business to grow.Collaboration Proves Essential

When asked about the top capabilities that contribute to data science success, 64 percent of survey respondents cited collaboration within the team and with business stakeholders. The ability to quantify and communicate the value of data science projects ranked second, as cited by 37 percent of respondents.

More than three of five survey respondents said static infrastructure presents a top challenge in becoming a model-driven data science organization. The presence of “iteration friction”—i.e., the lack of a continuous and repeatable data science project lifecycle—was the second biggest challenge, as cited by 55 percent of respondents.