Office of Institutional Research

November 9, 2017
EVP November 2017 Profile

The Office of Institutional Research (OIR) manages, analyzes and reports University data for a range of uses – from federal requirements to data-informed decision making and much more. OIR research analysts Grit Pientka, Jonathan Rossi and Albert Wang discuss how they work to understand clients’ questions, tackle new technologies to find the answers, and keep learning.

What do you do and what is your team's biggest contribution?

Jonathan Rossi: Our office is responsible for managing and reporting a vast amount of University data that spans all schools and units of the institution, as well as serving as a hub for analytical work that supports decision making at the school and University level. My work lately has focused on federal IPEDS data reporting, which is a mandatory step in receiving Title IV federal financial assistance, and the Tuition, Enrollment, and Financial Aid (TEFA) process, in which OIR helps facilitate the presentation of student data to the Corporation. I have also worked on a set of orientation exhibits for new leadership – including deans and members of the governing boards – on the University and its place in the higher education landscape.

Grit Pientka: I mostly work on projects that foster strategic decision-making based on data. Recently I have helped staff the Presidential Taskforce on Inclusion and Belonging by analyzing the first-generation college student experience, tracking the academic pipeline from undergraduate students to tenured faculty, and mapping changes in indicators of diversity over time. Often these analyses are not only concerned with Harvard but also consider peer institutions or the US overall.

The office is also invested in community building here, for example, by helping to facilitate the Analytic Staff Consortium, which connects about 250 analysts at various schools and units.

Albert Wang: I’d say that the biggest contribution of our team is the diverse skillset that we bring to data analysis questions; we have a mix of statistical and computational expertise that allows us to recognize the strengths and limitations of our data with respect to our client’s questions and to find ways to address them, ways that can often be quite innovative. Within the office I have done a lot to develop the technological infrastructure and toolsets. We often develop new techniques for addressing specific problems, and part of my work is to “generalize” these techniques so that they can be applied to future work. As an example, for a recent project, I approximated socio-economic variables associated with a population by geo-coding addresses and matching them up to publicly available census data. I have since been working to consolidate the code into a simple tool that can be easily used by other team members.

What don't people know about you and what your team does?

Grit: While we are part of the Provost’s office, we have helped with analytical questions from colleagues across the University. Also, we love to learn new tools and analytical techniques. For example, I learned a great deal of natural language processing for one of my projects and now use it on a regular basis, even advising other offices on methods for text analysis projects. Another example is that I saw an opportunity to facilitate the work of alumni affairs and development by deploying a web-based app that allows direct access to fundraising data, which had a huge impact on their productivity.

Albert: We use a lot of different tools and programming languages. I’m a strong proponent of R and we use it for most of our statistical work, but sometimes it’s just not the right tool for the job, and we adapt our toolset to the task at hand, which has made us conversant in quite a few different technologies. For instance, in making maps with tens of thousands of markers, we use JavaScript. For natural language processing, the R ecosystem is not as mature as Python, so Grit used Python for parts of that work. And for one project with customized simulations, R’s semantics made it too slow, so we wrote the performance-critical portion in C++. For more polished outputs, our team has also started to use Adobe Illustrator and InDesign.

Jon: People may not know that OIR helped guide Harvard’s decennial reaccreditation process, in collaboration with colleagues across the University. This was an enormous undertaking that required input in some way from every member of the team.

What is the most challenging thing about your work?

Albert: I think the most challenging thing about our work is consolidating data from a variety of different sources. Also a large part of the work is understanding clients’ needs and figuring out the best way to address those given the nature of the data. This challenge produces work that ends up quite rewarding; we usually approach the data and the client’s questions from a variety of perspectives, and we learn a lot from each other in the process.

Jon: Counting! Since Harvard is so decentralized, our definitions for what constitutes a particular unit of measurement (say, a faculty member or a field of study) can vary considerably. Context is key, and understanding how these definitions can shift is crucial to coming up with the “correct” numbers for a particular project.

Grit: For me the most challenging but also most exciting part of our work is to understand our clients’ needs and the contexts in which our analyses will be applied. This means constantly learning more about higher education in general, but also about processes in specific units across the University, which can be very diverse.

What are the professional backgrounds of your team members?

Jon: Our team is an incredibly diverse group, which is part of what makes OIR such an interesting place to work. You can learn from everyone. Team members have backgrounds in psychology, political science, economics, higher education, physics, statistics, math, and more, and many interests and skills outside of those domains.

Grit: As Jon said, our team members have a variety of backgrounds. Six hold Harvard degrees and five of us hold PhDs. Mine was in theoretical physics.

Albert: My PhD was in Economics, and I have a BA in Comparative Literature.

What does success/your best day look like?

Grit: There are three outcomes that I enjoy most: Delivering a useful or what clients sometimes describe as a “life-changing” analysis or app and witnessing that excitement; solving an intricate analytical or technical issue; and learning something new. So, my best days are those when any of this happens, which is fortunately pretty regularly.

Jon: A successful day can take on a range of flavors: the process for completing a reporting task gets automated or improved, a team member receives technical feedback on their analysis methods for a project, or a meeting takes place between OIR and another team at Harvard that will lead to OIR analyzing new Harvard data to answer an impactful question.

Albert: I really enjoy developing new solutions to problems. When we got our simulation code, which would have taken days to run in R, to run in less than 10 seconds with the C++ rewrite, I was happy. Data infrastructural work also scratches my perfectionist itch, so consolidating messy scripts into a neatly packaged library or program that others can use is great. But the most rewarding thing is just to see that my work is making an impact: if a client is excited about the new perspective we bring to the data, or if I’ve helped other members of the team, it’s a good day.

See also: Profile