Back

 Industry News Details

 
How Automation Is Changing Data Science Posted on : Jul 19 - 2019

Data Science: High Value, Even Higher Failures

Data science can provide a high return on investment across multiple industries and use cases. Whether predicting new target customers, measuring product demand or detecting high product failures - the use cases are nearly as infinite as the problems that face modern businesses. Although data science undoubtedly has significant potential to impact business decision-making, leaders across multiple industries have struggled with getting value from data science projects. In fact, according to research by the Gartner Group, nearly 85 percent of big data projects fail. Even more telling, a 2019 survey by Dimensional Research found that 96% of companies struggle with AI and Machine Learning. While there are several reasons for these failures, in large part, the disconnect between business users and the data science process is to blame.

Data science projects are complex, uncertain, and interdisciplinary in nature. Expectations of business users and data science teams are often misaligned, especially at the start. Iterations between business and data science teams are essential to bridge this gap, but data science teams are often too overwhelmed with projects and often struggle with how to leverage the vast quantities of data without having proper business context. As a result, in-depth and complex analysis and predictive models are neither tied to direct business value nor do they have enough diversity to provide the depth of insights for business users who are the ultimate users of the outcomes.

According to a 2016 Harvard Business Review article, there is a fundamental disconnect between data science teams and business users. Providing valuable impact to the business is, in fact, the single biggest hurdle that most data science teams face. Data science projects also take too long to deliver impactful results for the business. Data science is not merely machine learning but an interdisciplinary process, from raw business data through feature engineering to machine learning. View More