Microsoft XC Research

Microsoft XC Research

It’s time to get your UX researchers and data scientists collaborating

March 30, 2020

By Jodie Draper

Listen to this story:


Three people sit around a table, sharing data

Would you like to combine the power of high-volume quantitative data with a rigorous understanding of the why driving customer behaviors? If you’re like many I talk to, the answer is an emphatic yes, but you’re not sure how or when to set up collaborations between UX research and data science.

As a manager of a data science team at Microsoft that’s part of a UX research organization, I’ve seen the positive outcomes that customers experience when research and data science work together, along with some obstacles that keep our disciplines apart. I find the Crawl, Walk, Run model useful because it helps teams figure out the right collaboration strategy for a particular project or within a given org structure. And it allows them to get started immediately—no matter their level of experience.

Crawl mode of collaboration


Crawl collaborations are a great way to begin, especially if you’ve never worked across disciplines before. They involve a researcher or data scientist making a simple request to the other while completing their independent work, often toward the end of a project. Crawl collaborations are relatively easy to plan, usually framed around a data scientist wanting to know why or a researcher wanting to understand how much or how many.

One example of a Crawl collaboration was when a data scientist on my team approached a researcher to understand why customers were not taking a step that was essential to their successful use of an app. The data scientist’s retention analysis had already uncovered that a striking number of users abandoned the product rather than take this step. She could have presented these findings on their own, but thought it would enhance her results if she dug into the why. UX research revealed that users couldn’t find the button to take the needed step. Together, these two data points allowed the product team to pinpoint and address the problem.

If you’re thinking about a Crawl collaboration, your first question is likely to be, Who should we work with? Whether you’re on the research or data science side, look to your product team to help bridge the gap. Very likely, you already know an influencer, such as a product manager, who can help. Ask this person, “Is there someone you’re working with to get data?”

Walk mode of collaboration


Walk collaborations generally involve a researcher and data scientist working on the same problem in parallel while coordinating often. They require more planning, usually starting early in the development cycle, though the opportunity may come up a bit later.

For example, a researcher and data scientist on my team started collaborating midway through a project when they realized they’d landed on different recommendations regarding OneNote migration. At this point, they pivoted and built a new shared research plan. While they each worked separately, they met regularly to offer refinement toward the other’s work, resulting in a strong and unified message back to the product team. This led to a dramatic change of direction that benefited customers.

When considering possible Walk collaborations for your team, think about how you could sharpen insights with different data combinations—like augmenting telemetry data with a survey, diary study, or lab study. Be aware, however, that data sets will not always align the way you want. If your teams’ data seems to conflict, encourage them to seek out commonalities and build on those. Be open to new interpretations that arise.

Run mode of collaboration


Once your team develops an appetite for collaboration, it’s a good time to seek out those larger, more complex opportunities that generally result in the strongest insights and outcomes. In Run collaborations, data science and research work together against a tightly shared research plan. These collaborations require the most planning, and coordination must start at the beginning of the project. They enable a multidimensional understanding of user experience, such as analysis of people’s interactions with large systems.

For example, when we wanted to help our Microsoft Education partners understand the impact of technology in the classroom, we worked with a California school district that wanted to measure results of a tech-focused learning initiative. The three-year project involved ongoing collaboration between disciplines, with researchers performing ethnography and data scientists looking at software usage and student outcomes. Through this work, we found that the program was having a strong positive impact for students and teachers. This evidence supports the program’s continuation and, potentially, the initiation of similar programs moving forward.

If you’re considering a Run collaboration, you’ll need a level of resourcing that requires executive buy-in. Make sure you develop a clear plan for impact. And consider choosing a problem that your business is already focused on, especially if it’s the first time you’re pitching a Run collaboration. Before you pitch, comb through all existing knowledge on why the story you want to tell with this data matters from a business perspective. Your understanding will help executives connect the dots between your proposal and their stated priorities.

Four runners compete at a full sprint.

Now is the time to get crawling

As the saying goes, you have to crawl before you can walk, and walk before you can run. But even in Crawl mode, collaborations between research and data science help product teams understand customers from multiple viewpoints, driving the creation of more customer-centered products.

No matter where you find yourself today, the important thing is to begin. Take that first step to bring UX research and data science out of their silos. With the framework of Crawl, Walk, Run, you can bring your team into ever closer collaboration. Or you might choose to stop at Walking—and that’s okay. I’ve seen firsthand that small moves make large ripples over time.

How do research and data science work together at your organization? Tweet me @JPBDraper or my team @MicrosoftRI

Jodie Draper has been a data scientist at Microsoft since 2012, specializing in user insights derived from product usage telemetry, with a focus on Office and Education products. As the manager of the Data Science team for CIR, she leads a group dedicated to helping businesses and researchers use data to better understand what their users are doing, how their business is faring, and what investments may impact their products the most. Draper holds a BS in computer science from Purdue University, a master of computer science from the University of Illinois, and a certificate in data science from the University of Washington.