A breakthrough in machine learning would be worth ten Microsofts.
—Bill Gates1
While Katrina Lake liked to shop online, she knew the experience could be much better. The main problem: It was tough to find fashions that were personalized.
So began the inspiration for Stitch Fix, which Katrina launched in her Cambridge apartment while attending Harvard Business School in 2011 (by the way, the original name for the company was the less catchy “Rack Habit”). The site had a Q&A for its users—asking about size and fashion styles, just to name a few factors—and expert stylists would then put together curated boxes of clothing and accessories that were sent out monthly.
The concept caught on quickly, and the growth was robust. But it was tough to raise capital as many venture capitalists did not see the potential in the business. Yet Katrina persisted and was able to create a profitable operation—fairly quickly.
Along the way, Stitch Fix was collecting enormous amounts of valuable data, such as on body sizes and style preferences. Katrina realized that this would be ideal for machine learning. To leverage on this, she hired Eric Colson, who was the vice president of Data Science and Engineering at Netflix, his new title being chief algorithms officer.
This change in strategy was pivotal. The machine learning models got better and better with their predictions, as Stitch Fix collected more data—not only from the initial surveys but also from ongoing feedback. The data was also encoded in the SKUs.
The result: Stitch Fix saw ongoing improvement in customer loyalty and conversion rates. There were also improvements in inventory turnover, which helped to reduce costs.
But the new strategy did not mean firing the stylists. Rather, the machine learning greatly augmented their productivity and effectiveness.
The data also provided insights on what types of clothing to create. This led to the launch of Hybrid Designs in 2017, which is Stitch Fix’s private-label brand. This proved effective in dealing with the gaps in inventory.
By November 2017, Katrina took Stitch Fix public, raising $120 million. The valuation of the company was a cool $1.63 billion—making her one of the richest women in the United States.2 Oh, and at the time, she had a 14-month-old son!
Fast forward to today, Stitch Fix has 2.7 million customers in the United States and generates over $1.2 billion in revenues. There are also more than 100 data scientists on staff and a majority of them have PhDs in areas like neuroscience, mathematics, statistics, and AI.3
According to the company’s 10-K filing:
Our data science capabilities fuel our business. These capabilities consist of our rich and growing set of detailed client and merchandise data and our proprietary algorithms. We use data science throughout our business, including to style our clients, predict purchase behavior, forecast demand, optimize inventory and design new apparel.4
No doubt, the story of Stitch Fix clearly shows the incredible power of machine learning and how it can disrupt an industry. In an interview with digiday.com, Lake noted:
Historically, there’s been a gap between what you give to companies and how much the experience is improved. Big data is tracking you all over the web, and the most benefit you get from that right now is: If you clicked on a pair of shoes, you’ll see that pair of shoes again a week from now. We’ll see that gap begin to close. Expectations are very different around personalization, but importantly, an authentic version of it. Not, ‘You abandoned your cart and we’re recognizing that.’ It will be genuinely recognizing who you are as a unique human. The only way to do this scalably is through embracing data science and what you can do through innovation.5
OK then, what is machine learning really about? Why can it be so impactful? And what are some of the risks to consider?
In this chapter, we’ll answer these questions—and more.
Leave a Reply