The Breakdown blog series explores all of our departments to give you a better understanding of all that we do at SiteCrafting and GearLab. Each blog gives insight into our design and development process and shows how each department works together to form usable, custom web products.
The Breakdown: Data Science
In our last Breakdown blog post of the series, we interviewed GearLab’s Data Scientist Joe Izenman to explain the process and benefits of using data science to understand user behavior.
What is Data Science?
Data science focuses on treating collected data scientifically, by using statistics to make user predictions. The predictions result from data gathered on site users and their behavior. From server logs to Google Analytics, a number of different data resources exist, and it’s up to the data scientist to take in as many data sources as possible, and then analyze the information for concrete, reliable insight.
The Importance of Data Science
Data science is important in order to better learn about the behaviors and patterns of your site’s users. By tracking user data such as pages viewed or buttons clicked, and combining it with demographics and other sources of information, data scientists acquire vital insight about the users and what they do. Joe explains that the website world provides a stream of readily available information that is often ignored or missed. “Without someone to analyze the data, no one is listening to the users,” remarks Joe. Businesses benefit from listening to their users in order to determine and provide what they need.
Our data scientist begins new projects by performing a data consultation. This initial step of discovery involves working with the client to determine the best ways to measure and evaluate their site. In addition, Joe partners with our UX team, GearLab, to specify the goals and purpose of the website and how the collected data can be measured numerically.
The second step in the process is to perform a data inventory in order to answer the question: What data do we have? Some common data sources include site surveys, Google Analytics, server logs, and in-house client databases, as well as external market research. Once all the sources of interest are identified, Joe can work with software engineers to pipe all the data into a central location for analysis.
Next, with user research, Joe works to design experiments to test functionality or design issues. Each experiment or test has a clearly defined metric, such as wanting to get more people to click on a particular item. A common method is A/B testing. In A/B testing, data is collected as users are randomly assigned to interact with one of two versions of a website. A/B testing is different than user testing (or usability testing) as it is not conducted in a controlled environment, nor are users given specific tasks like they would be in a typical user testing session. Instead, user behavior is monitored while users interact with the website however they want. One major benefit of A/B testing is that data can be collected from many more users than can be collected from individual user testing sessions.
The data collected is also used for predictive analysis. By making predictions based on past user behavior, data science can provide product recommendations, targeted content, and other personalized features.
What Skills are Necessary for a Data Scientist?
A data scientist must be proficient with statistics, programming, data analysis, and experimental design. In addition, this field requires that one be comfortable with uncertainty as the goal is to understand the level of uncertainty from the available information.
An important and overlooked skill for data scientists is communication. They need to be able to read and interpret the numbers but also clearly show and communicate that data in an understandable way to people who tend to make decisions more intuitively.
The Role of Data Science
Data science plays a key role in the research phase of web design and development. After a website launch, the data scientist maintains an active role with iterative improvement based on post-launch research. Collected data informs what is working effectively and suggests trying something new.
Rewards and Challenges
For Joe, one major rewarding aspect of working in this field is the opportunity to continuously learn new things about the world, but to do so in a thought-out, rigorous manner. Joe finds his work both interesting and challenging as he works with so many different domains, such as healthcare and wineries, rather than a single industry. Sometimes, this can be a challenge as he must constantly change gears when tackling new problems with each client as all clients have different users. However, Joe enjoys working with our clients by helping them better understand their customers, and he derives satisfaction from helping our clients achieve success.
Joe explains that the biggest misconception surrounding data science is that people believe that data science and machine learning are somehow magic. Data science can sound like a one-size-fits-all solution for those who have heard the terms and see Amazon’s ability to predict things you buy or an automaker’s ability to make self-driving cars. However, data science is not quick, easy, or magical, but comes down to focus, process, and understanding both data and human elements in order to be successful. It takes a lot of work and understanding the user to get value from collecting and analyzing data.
As Joe explains, “Data science is really about finding what fails and what’s wrong. But, if we haven’t learned that something is ‘wrong’ yet, we’re going to treat it as ‘right.’”
Still curious about data science? Check out these resources from Joe:
The Stitch Fix Technology blog: http://multithreaded.stitchfix.com/blog/
The Linear Digressions podcast: http://lineardigressions.com/
Favorite Twitter Follow: Mara Averick, @dataandme
To learn about GearLab’s data science offerings, visit GearLab’s website.
And with that, our Breakdown series has officially come to an end! We hope you’ve enjoyed learning about each department and its unique process and role.
If you missed any of the previous Breakdown posts, you can check them out here:
Be sure to connect with us on social media to keep up with our team, blog posts, projects, events, and more.