Remember dissecting a frog in middle school? The point of the project was to learn about the internal workings of the amphibian’s organs and get a better picture of ecology as a whole. However, the smell of formaldehyde often made it difficult to think about the purpose. Instead, the tendency was to finish the task at hand — one that usually involved a knife-happy partner paired with a squeamish one who documented results with eyes half-shut.
Jump to the corporate workplace and the sentiment of the science lab and the frog can be found today: forget the corporate ecosystem! But with no formaldehyde smell or squishy parts to handle, why is it that 50% or more of data science projects never leave the lab? What’s the deal?
Data Science Project Obstacles
To date, big data and data science have primarily benefited companies in two ways: cost-cutting with improved efficiencies and the creation of a data-driven culture.1 However, data science can do so much more. As an interdisciplinary field that squeezes a myriad of data sets into meaningful insights, data science applied across an enterprise can perform at its highest purpose: disruptive innovation that leads to unmatched value.
While some 64% of company executives see success in creating innovation and disruption utilizing data science, common obstacles remain, making projects hostage to the lab. Five of these obstacles include:
- Data Quality
- Data Scientists
Planning is an obstacle for companies that lack defined business objectives. This leads to a failure to identify questions that companies need to have answered. Personnel and tools are being put into action without planning, and this leads to the discovery of random insights. The problem is that individually or collectively, these ad hoc findings don’t map to a plan or a return on investment. For the company, it’s like the school classroom that conducts a lesson without first having the curriculum in place.
Gartner has noted that large companies executing data science projects are not allocating enough time, staff, and funding.1
By 2019, it’s projected that 90% of large companies will have a Chief Data Officer (CDO). In the next few years, the ethos of this role will shape outcomes with a mission-critical function no different than IT, HR, finance, and business operations.2
The exact scope of the CDO, however, varies greatly and is often not clear. Is the role meant to be the driver of strategies or coordinator of resources? The ambiguity may be one reason CDO and similar roles in the data science arena are walking out the door.3 These reluctant heroes are often missing clarity in the C-suite, and it’s why planning remains a primary obstacle to overcome.
Then, there are the CDOs who become protectors of the data, building a barrier for the rest of the organization. By isolating outcomes in the lab, opportunities to evangelize at the enterprise level evaporate.
Leadership without definition impacts other obstacles, too.
When it comes to big data, weak governance and maintenance are operational challenges that increase the risk of compliance and can even result in revenues losses. Additionally, these data issues question the quality of outcomes — no matter the stage of the project.
The inability to validate the data often keeps projects from moving forward.
Similar to the challenges of the CDO, the role of the data scientist has lacked definition in many corporate settings. The work of the data scientists remains a bit of a mystery to most corporate stakeholders. That can be an obstacle when it comes to collaboration with the business as a whole. As identified in an earlier post, data scientists are in high demand and short supply. Because of their expertise, data scientists may build models that rely upon their perspective and are not necessarily aligned with the business. This can make changes to scale beyond the lab more lengthy and reliant on this role.
The technology selection process within corporations relies on many considerations: budget, in-house and management consultant expertise, existing platforms, and time and resources. Implementing a new technology is an obstacle in itself. Add an advanced analytic solution that depends on the high-level skills of data scientists, and it’s easy to understand how getting out of the lab becomes an obstacle. It’s comparable to a frog experiment where one student gets a scalpel, and everyone else in the class receives a toothpick: results are lowered.
In summary, obstacles are ever present in the corporate environment. Could it be that overcoming one obstacle holds the key to breaking past the others? In Part 2, each is examined closer to see if this is the case.
1”NewVantage Partners Big Data Executive Survey 2017,” White Paper, January 2017
2”Six Pitfalls to Avoid When Executing Data Science and Machine-Learning Projects,” Gartner, February 5, 2018
3”3 Top Take-Aways From the Gartner Chief Data Officer Survey,” Gartner, January 29, 2018
4”Why Are Data Science Leaders Running for the Exit?,” Oracle+DataScience.com, January 18, 2018