Keynote: Data as DNA: Building a Company on Data
Abstract
Creating a data driven culture is no simple task. Many companies say they are data-centric but few are taking full advantage of the data they have and spend massive amounts of time aggregating it. What does it take to create a data-driven culture? And, can you take it too far? With stories from Stitch Fix, Salesforce, and Amazon, Cathy will talk about how giving people access to the data they need and aligning everyone to the right metrics can transform an organization.
Cathy will discuss the early evolution of A/B testing at Amazon and share how Salesforce built a telemetry system for real-time monitoring and decision making in its large-scale search infrastructure. In addition, she will discuss the data-rich culture at Stitch Fix and share how the company’s 70-person data science team drives every decision and function inside the online personalized styling service. From styling recommendations and inventory allocation to logistics and demand modeling, data is the DNA of the company. Cathy will share the highlights and pitfalls to avoid in the journey to becoming a data-centric company and she will provide suggestions on how other organizations can harness their data to do the same.
QCon: Tell us about the keynote that you want to do at QCon New York.
Cathy: The title is "Data as DNA: Building a Company on Data." Every company is becoming a data company, and not everybody understands how to create a data-driven culture. There's a lot of data, but people are not taking advantage of it in the right ways, and sometimes it's not easy to do.I’ve been thinking about ways that I've seen data being transformative in different companies.
I started my career at Amazon and I have a lot of interesting stories to share about how tests were created in the early days of the Internet. Sometimes, people were not understanding what the data they had really meant, and how to build a good experiment with the questions they had. When I came to work at Salesforce, the team was building a brand new large-scale search infrastructure. One of the struggles of the team was how to deploy that without interrupting any of the customers. We likened it to ripping out an engine in mid-flight because everything was so dependent on search. When we built a full-scale telemetry system, and we measured how the system was performing compared to the legacy one, we got the confidence to switch over systems. So, I’ve seen data being used from an Internet company to a large-scale infrastructure project.
Then it was Stitch Fix, where I started six months ago. Stitch Fix was created as a data company from scratch. Our clients fill out a detailed style profile about their price, style and size preferences and then a stylist selects clothing and accessories that best meet the needs of each client. It's the perfect balance of art and science: we use data science to deliver personalization at scale, combined with an expert human stylist, who deeply understands each client’s unique needs and is able to connect on a personal level. A machine helps customers find out what they love and refine the selection of merchandise for each stylist, so our stylists can focus on building important client relationships.
One of our first hires was to lead data science, and data is essential for everything, from picking the right merchandise, to recommendations for stylists to match the right stylist to the right clients, to helping with customer support requests. I love this idea of how to be a data company, how to use data in the right way and how to avoid some of the pitfalls of conducting the wrong experiments. We've even seen things that take too long sometimes. I think that's a really interesting aspect:to know when to use data and when is it better for expediency.
QCon: I love using art and science together, and particularly the way StitchFix uses data to help the stylist serve the customer.
Cathy: I think the art and the science machine is an interesting topic right now. It took a long time for any chess player to beat a human. Then, with Deep Blue, machines evolved. Many master chess players have used computer aided chess to beat the machines. So, the machines are better than humans by themselves, but the blend of humans and machines is better together than machines are solo.
QCon: What do you want someone who comes to your keynote to walk away with?
Cathy: I want them to understand how important it is to use data to solve business problems, and that there’s an abundance of data in the world. Also, it's not just about being a data science company, it's about how you can answer questions and improve your businesses by leveraging data in the right way. I want them to learn how to build data into the DNA of their company and how to create the right culture to solve problems.
QCon: Are there any stories you can share?
Cathy: In the beginning, we used percentile data for performance metrics and we're missing so much of what was happening on the fringes until we started looking at the 98th percentile. Then, we were able to dig into some of the issues that were going on.
