Ask data scientists what the UX team does, and a lot of them think of designers. Ask UX researchers what a data scientist does, and a lot of them just think of computer nerds. Now, the truth is data scientists and UX researchers have a lot in common. If they work together they can generate valuable insights for optimal product decisions.
Why Should UX and Data Science Teams Work Together?
- Data science knows the what, but UX finds the why.
- Users say one thing but do another.
- There are always data quality problems.
- Sometimes it's too late for UX.
- AI/ML models go into production more effectively.
What Do UX and Data Science Have in Common?
Data science and UX procedures have more in common than most of us think. Both functions help improve decision making and products by understanding users and their needs. Depending on the problem, this mission can range from interviewing a small but focused user group, surveying a larger group, mining a huge amount of data or building products that automatically make decisions and continuously improve based on user behavior.
Though UX researchers and data scientists can have overlapping responsibilities, the UX team usually takes care of user group studies and surveys while the data science team handles the rest.
Both teams follow the scientific method which includes the cycle of:
- Identifying the problem
- Hypothesizing and exploring
- Designing and testing
- Analyzing errors and feedback
- Iterating . . .
What Are the Differences Between UX and Data Science?
The main difference between the UX team and the data science team is how they collect data.
UX researchers are more interested in knowing user intentions; why they use a feature and how they feel about it. Diving deep into the user’s feelings is an important part of UX research but it’s a process that’s difficult to scale and automate. For this reason, UX-gathered data is both more focused and more static (i.e., costly to repeat).
On the other hand, data scientists are more interested to know what product features were used on a much larger scale and about users’ journeys while they were interacting with the product. Gathering this information involves mining a huge amount of data that cannot possibly be tracked manually, so the data science team looks into more dynamic data with a larger magnitude.
Just note that there are some overlap between the two teams’ responsibilities and the data they’re looking into, so the teams’ work isn’t mutually exclusive.
Why You Need to Bring UX and Data Science Together
Data Science Knows the What but UX Finds the Why
Data science teams can gather a lot of information about what’s happening in your product experience. However, the data science team can’t go further when it comes to the why because those questions require more focused research.
In the famous urban legend, Walmart supposedly used data mining techniques and found that sales of beer and diapers were highly correlated. Through their data mining, they learned that mostly men were buying both diapers and beer — usually on Friday evenings. The answer became clear: Men were picking up diapers after work at the request of their wives and rewarded themselves with a nice cold six pack. So what do you do? As the story goes, Walmart placed diapers near the beer and sales of both skyrocketed. Case closed.
Not so fast.
Here’s how we know this never happened. Even though Walmart knew the what (high correlation between diapers and beer purchases), they couldn’t possibly understand the why through data mining alone. The question would have required user research to understand the phenomenon. That said, UX isn’t the end-all-be-all. It would have been almost impossible for a UX team to know this phenomenon was happening at all without the help of data mining techniques.
This hypothetical demonstrates the value of combining user research with the findings unearthed through data mining. When UX and data science teams work together, they can ask better research questions and come up with more complete answers.
Users Say One Thing and Do Another
UX teams are especially familiar with this challenge: Users say something but actually mean (or do) something else. For this reason, if UX and data science teams work separately on their own data, there’s a good chance they will come to contradicting conclusions.
Spotify is one of the famous companies that practice bringing data science and UX research teams closer for this reason. They noted that the contradicting conclusions were because their DS team and UX team were not truly collaborating and the research was not conducted at the same time with the same user groups/segments. So they concluded that the two teams should collaborate by:
- Defining the research questions together
- Applying complementary research methodologies
- Implementing all methods simultaneously rather than separately
There Are Always Data Quality Issues
The success of data science is very much dependent on data quality (remember “garbage in, garbage out.”). According to HBR almost all companies’ data do not meet quality standards. This means that while there are a lot of things to observe and conclude from big data, there will still be a need for confirmation from user research when it comes to making important product decisions.
Sometimes It’s Too Late for UX
Sometimes it’s too late for UX research or user study after a specific event. Data science’s main advantage in that case is the ability to predict such events based on historical data. We can take churn prediction as an example. With the help of data science, you can identify the at-risk users before they even decide to churn. UX can then come back into the picture and research such users to make them feel heard and reduce the churn risk while getting more valuable insights without the tensions and biases that come after churn.
AI/ML Models Go into Production More Effectively
As mentioned before, UX and DS teams are on a common mission to help with decision making and to improve the product.
On the one hand, the best practice in UX is to do continuous discovery, which makes UX research expensive, time-consuming and unable to scale (i.e., the cost grows as the size grows). On the other hand, when it comes to AI products such as recommendation systems, AI/ML models usually continuously receive user behavior information and improve based on that incremental training. In other words, decisions in AI products are made automatically based on data, and product improvement is inherently based on user interactions.
Now, where can the UX team help with AI products? Data scientists are always asking: How do I know when my machine learning model is good enough? Is the model optimizing the right metric? UX researchers can help to understand how much the decisions made by machines are acceptable to end users, how much trust users have in the AI and what their concerns are with respect to privacy, bias and explainability. The answers to these questions will help data scientists choose the best model and data according to what users need.
In the end, it’s clear: UX + data science = smarter product decisions.
