What comes to mind when you think about the term “data science”?
Most of us have heard of it, but have trouble translating it into something useful. We struggle to describe data science and actually use it to solve problems, even as it becomes one of the hottest new areas in business.
Brad Klingenberg, VP of Data Science at Stitch Fix, has no such problem. And during his Super Forum keynote last month, Brad cleared up some of the confusion around data science, explaining what the term means and how organizations can start using it.
We may be uncertain about data science because there is no clear definition. It’s a busy field combining math, statistics, and technology with larger concepts like artificial intelligence and machine learning. Data science includes everything from experimentation and statistical modeling to algorithms and data engineering.
All that may be hard to wrap your head around, but applying data science can be simple and accessible.
For the last four plus years, Brad has been using data science in relatable business applications. He’s been instrumental in changing the masses of information gathered by his company, Stitch Fix, into actionable insights. That’s a big part of what’s driven the Silicon Valley startup to become one of the country’s premier subscription and online shopping services.
Brad laid out what hands-on application has taught him about data science and what other professionals need to know about using it in their own organizations. Here are five lessons to help any business in any industry get started with data science.
As Brad joked, no one among his friends and family would ever have imagined that he – with a PhD in Statistics – would someday be part of the fashion industry.
But data science is being used in surprising places, fashion included. Brad currently applies data science at Stitch Fix to do everything from creating demand models (how many customers will be asking for service next week?) to making personalized clothing recommendations (this shirt will look great on you!)
It’s all about what you want to learn and how you want to improve. For example, one of Stitch Fix’s newest initiatives is to use data science to actually design the clothes they’re selling.
What interesting problems do you think data science can help you tackle?
Personalization itself isn’t new. Most services were personalized in the past simply because we had to do things one at a time. Just think of going to a tailor to get your clothes fitted – that’s traditional personalization.
Data science allows us to provide similar services, including high-quality customization, at scale. Data is the start of this, but remember that data can’t do everything on its own. To make something out of your data, you need analysis or, in data science, algorithms.
Algorithms change data to data science, because they are iterable, testable, and replicable. Using algorithms, you can find an improvement in one area, then easily apply that improvement to all your clients, not just the one.
That’s why data science has become such a buzzword. It’s integral to what everyone wants: large-scale personalization in a digital world.
Big data is an engineering problem. It’s overwhelming. On top of that, big data often becomes small data through slicing or segmentation, so success often has little to do with the volume of data you have. Instead, data science relies on quality data. Look for data with high information content and predictive value, so the information you’re gathering is actually useful.
Brad has implemented procedures at Stitch Fix that use data science to determine what data they should be collecting and tracking in the first place. The high-quality information they get as a result helps them focus on what moves the needle for the business instead of vanity metrics.
Don’t be afraid of being wrong, be afraid of not getting started. Once you get started, you can test your approach and improve.
For example, you may start looking at a problem your business is facing, move forward with a potential solution, and find out that it isn’t working. That’s okay. As long as you’re moving forward, gathering data, and testing your strategy, then you’ll be able to identify and fix problems.
Through that process, data science works with you to provide additional information, helping you see the problem and potential solutions in a new light. Then you can effectively evaluate your business strategies and make improvements.
Data science can be overwhelming, so don’t try everything at once. There are major advantages to starting slow, with simple models.
Simple models are easy to interpret, clear, and you can confidently understand what the model does. That makes it easier for you to apply the model to your business and communicate how it’s helping stakeholders like your executive team and business partners.
Once you’re comfortable with simple data science models and have clearly demonstrated their value, then you can move on to more complicated algorithms.
Don’t complicate data science. Brad is using it to design clothes, make recommendations on what customers would like, and incorporate client feedback so they can improve in the future.
Your applications for data science will be different. Instead of recommending clothes you may recommend the best benefits for association members or use feedback to determine what product updates to prioritize.
However you decide to start using data science, start small. Use data and algorithms to learn more about your customers or members, then take things to the next level by making predictions and improvements.
And remember: don’t be afraid to make mistakes. Data science is all about learning and improving based on what you already know and what you’re trying right now.