Skip to content Skip to sidebar Skip to footer

Arsenii Palivoda/Getty Images

The role of data scientist — one who pulls stories and makes discoveries out of data — was famously declared the “sexiest job of the 21st century” in Harvard Business Review back in 2012. Just two years ago, the authors, Thomas H. Davenport and DJ Patil, updated their prognosis to observe that data scientists have become mainstream and absolutely vital to their businesses in the age of artificial intelligence and machine learning (ML).

The job role has evolved as well, partly for better, partly for worse. “It’s become better institutionalized, the scope of the job has been redefined, the technology it relies on has made huge strides, and the importance of non-technical expertise, such as ethics and change management, has grown,” Davenport and Patil observe.

Also: Business success and growth is dependent upon trust, data, and AI

At the same time, data scientists report that “they spend much of their time cleaning and wrangling data, and that is still the case despite a few advances in using AI itself for data management improvements.”

Even more significantly, “many organizations don’t have data-driven cultures and don’t take advantage of the insights provided by data scientists,” Davenport and Patil find. “Being hired and paid well doesn’t mean that data scientists will be able to make a difference for their employers. As a result, many are frustrated, leading to high turnover.”

People respect data scientists, but tend not to act on their recommendations or insights, a recent survey of 328 analytics professionals out of Rexer Analytics confirms. Only 22% of data scientists say their initiatives – models developed to enable a new process or capability – usually make it to deployment, observes survey co-author Eric Siegel, former professor at Columbia University and author of The AI Playbook, in a related post at KDNuggets. More than four in ten respondents, 43%, say that 80% or more of their new models fail to deploy.  

Even tweaking existing models doesn’t pass muster in many cases. “Across all kinds of ML projects – including refreshing models for existing deployments – only 32% say that their models usually deploy,” Siegel adds. 

What’s the problem? Interaction between the business and data science teams — or lack thereof — seems to be at the heart of many problems. Only 34% of data scientists say the objectives of data science projects “are usually well-defined before they get started,” the survey finds. 

Plus, less than half, 49%, can claim that the managers and decision-makers in their organizations who must approve model deployment “are generally knowledgeable enough to make such decisions in a well-informed manner.” 

Also: Is prompt engineer displacing data scientist as the ‘sexiest job of the 21st century’?

Overall, the top reasons cited for failure to deploy recommended machine-learning models consist of the following:

  1. Decision makers are unwilling to approve the change to existing operations.
  2. Lack of sufficient, proactive planning.
  3. Lack of understanding of the proper way to execute deployment.
  4. Problems with the availability of the data required for scoring the model.
  5. No assigned person to steward deployment.
  6. Staff unwilling or unable to work with model output effectively.
  7. Technical hurdles in calculating scores or implementing/integrating the model or its scores into existing systems.

The struggle to deploy stems from two main contributing factors, Seigel says: “Endemic under-planning and business stakeholders lacking concrete visibility. Many data professionals and business leaders haven’t come to recognize that ML’s intended operationalization must be planned in great detail and pursued aggressively from the inception of every ML project.” 

Business leaders or professionals need greater visibility “into precisely how ML will improve their operations and how much value the improvement is expected to deliver,” he adds. “They need this to ultimately greenlight a model’s deployment as well as to, before that, weigh in on the project’s execution throughout the pre-deployment stages.”

Significantly, the ML project’s performance often isn’t measured, he continues. Too often, the performance measurements are based on arcane technical metrics, versus business metrics, such as ROI. 

Also: Want to be a data scientist? Do these 4 things, according to business leaders

Still, data scientist is a great job to have, and keeps getting better, the Rexer survey suggests. In the previous survey in 2020, 23% of corporate data scientists reported having high levels of job satisfaction — a percentage that almost doubled to 41% in this most recent survey. Only 5 percent express dissatisfaction, down from 12% in 2020. 

The appetite for data science skills is still growing as well. Data scientists continue to be hard to find — 40% say they are concerned about shortages of talent within their enterprises. Half report their organizations have stepped up internal training to boost data science skills, while 39% are working with universities to promote interest in data science.

error: Content is protected !!