When Neuberger Berman’s fundamental consumer sector analyst began to see positive signs coming out of the apparel company Lululemon Athletica, he turned to the data science team to find out whether more recent sales confirmed the apparent trend for growth. We scoured six billion credit card transactions and also worked with a data vendor in Asia to scrape SKU-level data on volumes and prices from the Chinese ecommerce site Tmall.
The findings were promising, and these growing ecommerce sales were reflected in its quarterly results two months later.
But our analysis was able to go still further and look into factors underpinning the sustainability of this sales growth.
Our researcher wanted to examine the company’s claim that it was quietly growing its menswear sales, having built its reputation in womenswear. This sort of question is often posed by data scientists, who apply artificial intelligence and machine learning to uncover patterns in large datasets. Data scientists have learned to identify bank accounts that are likely being used for organized crime, for example: they don’t write checks, they do a lot of cash-only transacting, they get mysterious deposits from charities, and so on. We can also identify with some confidence whether a bank account or credit card is owned by a baby-boomer or a Millennial, or by a man or a woman, based on their shopping habits. And it turned out that credit card transactions revealed Lululemon not only increasing sales but also selling to a growing number of men.
This example shows how alternative data and machine learning can combine to discover not only whether a company’s sales are growing, but also how its customers are changing over time.