Desktop version

Home arrow Computer Science

  • Increase font
  • Decrease font

<<   CONTENTS   >>

A Second View from the Ground

We asked one of our doctoral students in analytics and data science about their experiences working in a research lab sponsored by a healthcare company.

My formative project was the pattern mining of healthcare claims data. The data was the most abstract and complex dataset I had ever seen; the medical billing code system data may as well have been another language, and the data structures were unlike anything I had previously encountered. With the help of subject matter experts at the sponsoring company, I grew to appreciate the challenge and understand the necessity of the “way they do things” The project eventually became the foundation of my dissertation research.

I do not feel I could have conceived the idea from a literature review alone. The business question posed would not have emerged from just being exposed to theory and I would have likely not have known to pursue that line of research had I not been exposed to the business problem.

The biggest challenge for me was getting all of my work legally cleared by the company to publish. Many data science journals require the data set to be available to the scientific community for reproducibility (which is a good thing), and that is a non-starter for a healthcare company. As a result, one of the limitations for me was that I did not produce the number of published manuscripts) that I think I could have.

Overall, my advice for companies looking to engage data science doctoral students is three-fold:

First, select those projects that are academically interesting but that the company never seems to justify. Allow the projects to be open-ended, expect 66%—75% of them to perhaps be shelved or abandoned in favor of the projects that take off (the sponsor should have several projects). Reassess the projects that seem to be stalling to determine why those projects are stalling. Perhaps the project is no longer seen as useful by the assigned corporate team due to outside factors. It is very hard for a student to continue a project that they sense the sponsor has lost interest in due to no fault of their own.

Second, understand that the business questions asked may not result in the findings the sponsor wants, and that’s okay. For example, I found that the shortterm cost of a program actually increased initially but decreased over the long-term. This was obviously not what the company wanted to demonstrate, but the company could still use that information to improve or understand the phenomena better. These scenarios may result in the non-publication of studies, which are both unfortunate for the student as well as a negative aspect to the contribution of corporate sponsored research to literature (less-than-ideal results or null findings can still be of interest to peer review publications).

Finally, treat the doctoral student like student-consultants, as opposed to just a consultant or just a student. Plan to invest a lot of time in them upfront, and then taper that amount as the student earns their independence.

Our Summary Checklist for Research Partnerships with University Doctoral Programs

Working with doctoral students can facilitate innovation and research. However, analytics managers need to appreciate that doctoral students are not masters students or undergraduate students; they have longer program horizons, different objectives, and importantly have much deeper skills. Doctoral students are less likely to work on “one off” or “capstone” projects (unless it generates a publication), but rather in the context of research labs — either as part of a multi-party consortium or as a single sponsor/single university engagement. Engaging doctoral students will require deeper investments of time and funding than will engagements at the undergraduate or graduate levels but can generate valuable and highly tangible returns such as patentable research, published findings, intellectual property, and other forms of innovation.

Our suggested checklist for working with data science doctoral students:

/ Recognize that research relationships with universities are typically longterm engagements that require a significant investment of both time and funding.

/ The best way to get started is to approach the program director, the cen- ter/institute director or the office of research. This information should be available on the website. Research faculty may not immediately return your call.

/ Doctoral students will commonly have their own “page” on the university website. See what kind of research they are engaged in and how/if that research is funded. Doctoral students are always assigned to a faculty advisor and their research agenda is almost always an extension of the faculty member’s research agenda.

/ Be cognizant of the fact that doctoral programs are typically 4-5 years long and will enroll less than 50 students. These programs should not be considered as broad pipelines of talent like an undergraduate or master’s-level data science program. Given that an increasing number of graduates from PhD programs are entering the private sector, consider doctoral-level partnerships as long-term research engagements, which may generate a small number (but potentially transformational) of full-time hires.

/ Research publication is the “coin of the realm” for doctoral students (and faculty). Collaboration with PhD students must have a research agenda. Agreement on what can and cannot be published needs to be clearly documented and agreed upon very early in the discussion process. Note that the legal department of your company will almost assuredly start that conversation with “no”. But there is almost always an opportunity for mutually beneficial outcomes related to publication of findings.


  • 1. Find a PhD-Types of PhDs, Accessed July 2, 2020.
  • 2. The 7 Essential Transferable Skills All PhDs Have, career-advice/the-7-essential-transferable-skills-all-phds-have Accessed July 2, 2020.
  • 3. Linda Burtch (2018). The Burtchworks Data Science and Predictive Analytics Salary Report. Accessed May 24, 2020.
  • 4. United States Bureau of Labor Statistics (2018). cpsaatl8b.htm
  • 5. Jasper McChesney and Jacqueline Bichsel (2020). The Aging of Tenure-Track Faculty in Higher Education: Implications for Succession and Diversity (Research Report). Accessed June 9, 2020.
  • 6. National Center for Education Statistics (2018). dl8/tables/dtl8_315.20.asp Accessed June 22, 2020.
  • 7. Katie Langin (2019). In a first, U.S. private sector employs nearly as many Ph.D.s as schools do. employs-nearly-many-phds-schools-do# Accessed May 18, 2020.
  • 8. National Science Foundation (2017). Accessed May 27, 2020.
  • 9. Jennifer Lewis Priestley, and Robert J. McGrath. “The Evolution of Data Science: A New Mode of Knowledge Production”. IJKM 15.2 (2019): 97-109. Web 4 Aug. 2020. doi: 10.4018/IJKM.2019040106.
  • 10. Google Research. Accessed July 3, 2020.
  • 11. Google AI Research Division to Issue PhD Degrees (2019). halting-problem/google-ai-research-division-to-issue-phd-degrees-8a6954293047 Accessed June 8, 2020.
  • 12. Amazon Research Awards, Accessed May 18, 2020.
  • 13. National Science Foundation, id=505651 Accessed July 1, 2020.
  • 14. Kennesaw State University, Accessed May 1, 2020.
  • 15. Discover Data Science, phd/ Accessed May 18, 2020.
  • 16. National Science Foundation, Accessed July 14, 2020.
  • 17. National Science Foundation, time-analytics Accessed July 14, 2020.
  • 18. University of Cincinnati, html Accessed June 3, 2020.
<<   CONTENTS   >>

Related topics