Big Ideas – Degree Progress

Public institutions should be efficient. Universities should be run well to serve their students.

The structure for undergraduates at the UW is supposed to be simple:

  • Start
  • Spend 2 years ‘searching’, exploring a broad range of topics to identify the area you want to specialize in.
  • Spend 2 years in a chosen field of study (a major), specializing yourself for the career/research/endeavors you want to do.
  • Graduate (4 years later)

At the UW, 58% of undergraduate students finish their “4-year” degree in 4 years. We can do better. The UW can use data it already has to help students graduate faster (with less debt).


Once students are in a major, their path to graduation is simple. They have known requirements (typically a list of courses), set by the department that ‘owns’ the major. Despite that, many students need longer than the typical 2 years to finish and graduate from their major.

The job of helping these students graduate belongs to departmental advisors. Currently the tools that show progress towards are per student; an advisor with 100 students would need to run a report 100 times. That’s so impractical I bet it never happens.

What is needed is a way for a departmental advisor to see, at a glance, the state of all of the students in their major. The most common tool for this is one of the best: a dashboard. The UW already has UW Profiles, a set of visualizations that show information about academic, research, and finance across the entire university. However, information this specific isn’t available yet.

Luckily we have several useful data sets:

  • A complete list of students, including their current major
  • Student transcripts, with the grades students have earned in each course
  • Student registrations, with the courses a student is currently taking
  • The most common courses taken by students in each major. These are predominantly the courses required to complete earn a degree in that major.

From this we can build a visualization showing the current status of each student in a major, including their progress through common courses. I’ve been building this with the help of advisors in the Computer Science & Engineering department.

Example Screenshot

Here’s an example of a single major in the College of Arts of Sciences, in Seattle. Each row is a student. Each column is a popular course for that major.

Project Status: in Quality Assurance testing, soon to be released.

Time to Graduate

In my last post, I introduced myself and gave some background about data science in academia. This time, I want to look at one facet of higher education: the time to graduate.

It’s About Time

9908 - game of loansTime to Graduate: the number of years between when a student’s first college classes and when they leave with a degree.

Colleges don’t have fixed prices for degrees; they charge for each day/quarter/year you’re enrolled. Students adjust their finances each year to handle different combinations of loans, scholarships, grants, and tuition. State funding is similar: public institutions are often funded by giving a certain amount per student, per year.

Everyone would benefit if students graduate faster without sacrificing the quality of their education. Students would have less debt. Governments would spend less per student. Universities could have a greater enrollment capacity.

Since student debt contributes to income inequality, effectively lowering the cost per degree would help make society more equitable.

The Six-Year Metric

Universities are often judged by the percentage of 4-year students who graduate within six years. The UW graduates 82% of its 4-year student population within 6 years (per UW Fast Facts). That’s considered good.

what it means

Only 58% of students graduate in 4 years.

Wait a minute. Why is anyone measuring how many 4-year degrees take less than six years? Why are universities given 50% slack?

Let’s imagine other metrics with 50% slack:

  • 30-year mortgages paid off within 45 years
  • 30 mpg cars that get at least 20 miles a gallon
  • 4-hour flights that take less than 6 hours.
  • 40-hour workweeks that require less than 60 hours a week?

That’s a bit ridiculous, IMHO. Before we go further, let’s think about why it takes time to get a degree at all.

Hoops and Hurdles

What are the barriers, good and bad, that prevent students from graduating?

Some barriers to graduation are necessary. Learning takes time. Research and discovery take time. The ever-narrower specialties in modern industry take more training (read: time).

Some barriers to graduation are unnecessary. Required courses are full. Popular departments turn away qualified students. Students have to start over when searching for a specialty. Some students take a long time to find a specialty at all.

The Challenge of Choice

The University of Washington, like many universities, offers many different areas of specialization (majors), leading to different credentials (degrees). Currently the UW offers ~180 majors, and over a thousand courses.

How do students pick which courses to take? How do they choose which majors to pursue? How do they choose what research to learn about?

Humans don’t respond well when given too many choices . One reason is psychological; we can only keep ~7 things in memory at once.


“You literally ought to be asking yourself all the time what is the most important thing in the world I could be working on right now, and if you are not working on that why aren’t you?” – Aaron Swartz

I wonder if public universities can be made more efficient. Can a university reduce the time to graduate for its students? Can it reduce the total price of a degree by better allocating existing resources?

I can think of several areas where the time-to-graduate can be improved, because the existing processes haven’t been optimized for efficiency. Having spent 3 years working around the UW’s administrative systems and processes, I’ve seen how the sausage is made.

Help Students Pick Majors

  • Recommend courses and majors in an intuitive, transparent way.
  • Show ‘palettes’ of courses that have high information gain (entropy), which lead to recommendations about which majors are likely to be a good fit.
  • Expose the ‘overlap’ and ‘similarity’ between majors. For example, Physics and Astronomy are far closer, both in terms of prerequisites and subject matter, than Physics and Music.

Help Transfer Students

  • Make it clear which majors/degrees accept a lot of ‘transfer’ credits, and which ones don’t.
  • Be transparent about how well transfer students do with an AA or without.
  • Be transparent about which community college programs/classes lead to a fast time to graduate at a university, and which ones don’t

Make the Major Application Process Easier

  • Make it clear the typical course grades, background and GPA for students who enter every major, especially the popular ones.
  • Make it easy for students to find alternatives to popular majors that don’t require a lot of prerequisites to enter.

Allocate Resources Better

    • Predict course and major demand ahead of time, and adjust teaching resources (and building allocations) to compensate.
    • Be transparent about which courses are bottlenecks
    • Do a better job of allocating building resources. Many classes aren’t given a higher enrollment limit because there isn’t a bigger room available to that department, even though larger rooms exist elsewhere.

Final note: there are data, statistical analysis, and prediction models about all of these ideas. I’m sadly not permitted to disclose any of it.

The Bootstrap

This is the first post in a blog on data analysis, data-driven discovery, and decision making at the University of Washington.

My name’s Dev Nambi, and I am a data scientist in the UW’s Enterprise Data and Analytics team. I’ve worked at the UW since 2012. Before that I was a software developer and analyst at Microsoft’s Ads R&D group, an ETL developer at a startup, and more.

“The best minds of my generation are thinking about how to make people click ads…” – Jeff Hammerbacher

That was me. It the best job I could find that paid enough to let me work off my student loans. Now I am hoping to give back to the next generation of students at the university.

Data Science in Academia

There is quite a bit of excitement and activity on data science in academia. So far the emphasis has been, rightly, on data-driven discovery in scientific research. The UW’s emphasis there is its new Data Science Incubator and eScience Institute.

There are potentially far-reaching implications in fields as varied as astrophysics, oceanography, chemical engineering, genomics, and sociology.

I admire that, but I want to do something more pragmatic, more direct.

Data Science in Administration

A university can be made more efficient using data. There are so many ways to do this it’s mind-boggling, so I use a heuristic to pick areas to focus on: changes must directly benefit students.

My reason for starting with students is simple…

tuition too high

Tuition is very expensive compared to entry-level salaries, and that problem has been getting worse for decades. Student debt is now larger than credit card debt, and it’s practically impossible to get rid of.

My goal is to find ways to help students graduate with the same quality education they have now, but with less debt.