The fashion industry is pretty visible in terms of high-profile events held in cities like New York or Milan. Fashion designers drape models with their latest creations, and then those models stroll down the catwalk while photographers turn night to noon with flashes. However, while that might be the public image of the industry, fashion companies face significant challenges. Rather than leading consumers with new designs and styles, these companies still must understand their audience, what drives them, and what consumer behaviors affect their success. Data science is being used to do just that.
Pivotal and Fashion
The data science team from Pivotal worked hand in hand with an unnamed fashion company to determine just what insights big data could provide to the fashion industry. The results were significant, and provided much more value than the client had anticipated.
The challenge here involved determining exactly what value analyzing big data could have for an industry generally considered to lead, rather than follow the whims of consumers. Between women’s and men’s fashion, hundreds of billions of dollars in revenue are generated annually in the US alone. Taken globally, it’s significantly more lucrative.
Traditionally, fashion businesses employed teams of consultants whose only job it was to focus on emerging trends in the industry as a whole – the search for the “next big thing.” Some companies utilized outside teams, while others had in-house consultants, but they all operated in relatively the same way. This information was amalgamated with sales trends, out of stock items, overstock levels and the like to help companies make decisions on where to go. In most of these instances, the information was captured manually, which is incredibly inefficient. Big data promised better results.
The team from Pivotal determined a roadmap and a focus for their efforts – in this instance, the goal was to determine what online retailers were doing with their websites, and how that data could be utilized.
Ultimately, the project was determined to be “trend forecasting and intelligence.” To achieve the goals of the client, the team would need to capture unstructured data from the vast number of ecommerce fashion sites currently operating. In the end, the model would allow self-service for users, allowing them to interact visually with:
- Product categories
- Product subcategories
- New merchandise
- Out of stock items
One of the most important end results here would be the ability to see how merchandising changed over time for ecommerce companies. By being able to visualize the entire industry and how it changed over the course of each season, the entire supply chain could be improved, offering greater efficiencies, better profitability, fewer out of stock items, and a reduced number of markdown products being sold for little to no profit. Intelligent decisions regarding retail partners could also be made (by judging which retailers were selling more of what products during each season of the year).
The primary hurdle with the project was the sheer scope. There were over 100 different websites, and initially data was aggregated for 18 months (daily information). Even once that initial amount of information was gathered, further websites had to be scraped. And this was only the initial element – the data still had to be analyzed, parsed, processed, numerical metric extraction performed, and then insights developed and displayed.
In the end, the project was hugely successful, managing to deliver significant insights and knowledge. Users were able to ask (and answer) questions ranging from what buying trends currently dominate the industry to how different markets vary in terms of buying behavior and consumer preferences. Retailers could be categorized by performance and better decisions made with partnering. In truth, it resulted in brand new intelligence being provided, and was only enabled by big data and data science. Moreover, this information was provided to decision makers not in weeks, days or even hours, but often in fractions of a second.
Other advantages gained by the client included significantly greater agility, better speed to market and the ability to quickly make informed, accurate decisions based on hard data rather than conjecture or sentiment.
If you found this Use Case interesting, why not review the other Use Cases in our archive.