Ninety-seven percent of U.S. data executives say data science is crucial to maintaining profitability and boosting the bottom line. However, nearly as many say that flawed approaches to data science strategy, execution, and staffing make achieving that goal difficult.
Thatâ€™s according to a new study commissioned by Domino Data Lab and published by Wakefield Research, which surveyed 300 executives in data science roles at companies with annual revenue in excess of $1 billion. According to Domino cofounder and CEO Nick Elprin, the findings suggest that while executives have expectations for revenue growth from their efforts in data science, they arenâ€™t making investments in the right places.
One challenge is that expectations tend to outpace investment, with â€œsplashyâ€ short-term investments far outnumbering sustained commitments. While 71% of respondents to the Wakefield survey say their company leadership expects revenue growth from investments in data science, 48% say their company hasnâ€™t invested enough to meet those expectations. In fact, more than three-quarters (82%) believe their employers actually prefer investments that yield only short-term results.
Companies are also struggling to execute on the best-laid plans to scale data science. More than 2 in 3 data executives (68%) report itâ€™s at least somewhat difficult to get machine learning models into production to impact business decisions, and 37% say itâ€™s very or extremely difficult to do so. Moreover, nearly 2 in 5 executives (39%) say data scienceâ€™s impact is often hampered by the inconsistent standards and processes found throughout their organization.
Further complicating matters, companies face shortages of skilled employees and the tools they need. Forty-eight percent of data executives told Wakefield that their companies lack the necessary institutional knowledge or face barriers to hiring enough talent to scale data science in the first place (44%). More than 2 in 5 execs say their data science resources are too siloed off to build effective models (42%) and nearly as many (41%) say they havenâ€™t been given clear roles. Moreover, 37% of executives name outdated or inadequate tools to build and manage models as a key factor leading to reduced data science impact on the business, which might explain why a third of data executives (33%) say that not improving models can result in loss of productivity or rework.
Laments over the AI talent shortage in the U.S. have become a familiar refrain from private industry. According to a report by Chinese tech company Tencent, there are about 300,000 AI professionals worldwide but â€œmillionsâ€ of roles available. In 2018, Element AI estimated that of the 22,000 Ph.D.-educated researchers working on AI development and research globally, only 25% are â€œwell-versed enough in the technology to work with teams to take it from research to application.â€ And a 2019 Gartner survey found that 54% of chief information officers view this skills gap as the biggest challenge facing their organization.
AI technologies are becoming prevalent in enterprises around the world. While the adoption rate varies between businesses, a majority of them â€” 95% in a recent S&P Global report â€” consider AI to be important in their digital transformation efforts. Organizations were expected to invest more than $50 billion in AI systems globally in 2020, accordingÂ to IDC, up from $37.5 billion in 2019. And by 2024, investment is expected to reach $110 billion.
In light of this, the Wakefield report also explored â€œwhat keeps data science leaders up at night.â€ The results, Elprin says, deliver a stark warning for companies cutting corners with data science, particularly those lacking the proper vision.
Eighty-two percent of those polled say their company leadership should be concerned that bad or failing models could have severe consequences for the company, and 44% report that a quarter or more of their models are never updated. The consequences of model mismanagement, the execs say, could range from bad decisions that lose revenue to faulty internal KPIs for staffing or compensation decisions, security and compensation risks, and discrimination or bias in modeling.
Perhaps unsurprisingly, the Wakefield survey found that 82% of â€œhigh-maturityâ€ companies â€” i.e., those with strong organizational processes, analytical agility, and cohesion â€” report data science has had a â€œgreatâ€ or â€œfairâ€ amount of impact on sales or revenue. Thatâ€™s compared with just 14% of â€œlow-maturityâ€ companies. On average, 69% of data models at high-maturity companies impact business decisions versus 49% at low-maturity companies, the survey suggests. And executives at 65% of high-maturity companies say their companies treat data science as a first-class discipline, the same as finance or marketing.
â€œTo properly scale data science, companies need to invest in cohesive, sustainable processes,â€ Elprin continued. â€œ[They need to] develop, deploy, monitor, and manage [machine learning] models at scale.â€