Smarter e-shopping with The North Face and Watson
“Hi! Can I help you shop for a jacket today?”
This was no an ordinary sales assistant greeting me at the store. It was Expert Personal Shopper, a robot by The North Face that would help me choose “the perfect jacket for my next adventure” [1].
—
Global e-retail sales surpassed 2.3 trillion U.S. dollars in 2017, accounting for 10% of retail sales worldwide [2]. As this market continues to grow, retailers must provide improved buying experiences for customers, who are overwhelmed with sellers’ efforts to attract them to their sites. Once you get a customer’s attention, however, there’s much to do before converting a visit into a purchase. Some research suggests that too many options can reduce sales (through “choice paralysis”) and might reduce people’s satisfaction with their decisions, even if they were good [3].
The North Face (TNF), a California-based outdoor product company, has been a pioneer in using Machine Learning (ML) to solve these issues and grasp a competitive advantage in customer conversion. In 2016, TNF developed Expert Personal Shopper (XPS), a personalized buying experience that leverages on IBM Watson’s AI capabilities to help customers browse and choose through hundreds of products. After a pilot program, the tool is online and available to use at https://www.thenorthface.com/xps.
The idea behind XPS is simple. In TNF stores, you’re usually greeted by a sales assistant that will guide you through the store. To many customers, she will be an educator on the technologies embedded in TNF’s gear and a guide in choosing an appropriate product. This experience is difficult to replicate online but is crucial to sustain e-sales growth.
Through XPS, TNF is trying to mimic this type of customer support, engaging in a dialogue to find the right product for you. Questions might be simple, such as “How much rain are you expecting?”, or more complex -leveraging on Watson’s processing power- such as asking where and when do you intend to use this product. Through Natural Language Processing, XPS determines temperature and wind estimates for that location in that season and ranks the products accordingly.
Cal Bouchard, The North Face’s VP of digital commerce and experience, thinks this initiative is “game-changing” and is encouraged by the initial results. Customers spend 40% more time onsite when they interact with the solution, and 75% of consumers who tried it said they would use it again [5].
In the short term, she acknowledges that there’s a lot of refinement pending. The application has some bugs and, for now, is limited to jackets. She expects to add deeper layers of data, which will make XPS more intelligent [6]. But, after a year of testing, TNF is also considering new possibilities for this technology in the medium term. Departing from being just a product recommendation engine, Cal believes that it could provide content such as “product education, content about where you might be going or what activity you might be doing” [7].
TNF has taken a bold first step by opening XPS to the public. A key advantage is that it provides a scalable solution which will learn and improve throughout its use. However, TNF should be cautious. XPS might serve as a ‘proof-of-concept’, but as a pilot program it is weak. After a few tries, its limits and bugs were very evident. TNF should improve it significantly before taking a new step forward. IBM has been recently criticized for Watson’s performance in the cancer treatment space [8], this is a warning light to its application in the retail industry.
Looking forward, “echo-chamber” effects could develop by iterative feeding on data produced by the algorithm itself, which might be detrimental for sales. And, more fundamentally, TNF should really focus on its core -providing good gear- rather than using ML to compete in the travel-recommendation space, where others have greater competitive advantages in terms of data and users. TNF should use machine learning to design and develop clothing and equipment that will perform better, by leveraging on its broad worldwide use. To do so, it must collect much more data from its current users, incentivizing the sharing of it -for example- through stronger loyalty programs.
Finally, a few questions remain regarding the use of machine learning algorithms by The North Face:
- Can this technology can transform TNF’s in-store buying experience, or should it be limited to be an online shopping assistant?
- Should this technology be limited to TNF direct-to-consumer channel, or should it expand to encompass its retail partners as well? In the latter case some challenges arise, for example; should it adjust to offer only the SKU’s served by the retailer?
Word count: 775
References
[1] The North Face, “XPS, https://www.thenorthface.com/xps, accessed Nov 2018
[2] Statista, “E-commerce share of total global retail sales from 2015 to 2021“, https://www.statista.com/statistics/534123/e-commerce-share-of-retail-sales-worldwide/, accessed Nov 2018
[3] Barry Schwartz, “More Isn’t Always Better”, in Harvard Business Review, https://hbr.org/2006/06/more-isnt-always-better, accessed Nov 2018
[4] Statista, “Online shopping cart abandonment rate worldwide from 2006 to 2017”, https://www.statista.com/statistics/477804/online-shopping-cart-abandonment-rate-worldwide/, accessed Nov 2018
[5] Brittney Dorr, “Implementing AI and Cognitive Technologies as Part of Your Digital Strategy”, in IBM Blogs, https://www.ibm.com/blogs/watson-customer-engagement/2017/01/26/implementing-ai-cognitive-technologies-part-digital-strategy/, accessed Nov 2018
[6] Matt Marshall, “The North Face to launch insanely smart Watson-powered mobile shopping app next month”, https://venturebeat.com/2016/03/04/the-north-face-to-launch-insanely-smart-watson-powered-shopping-app-next-month/, accessed Nov 2018
[7] Lauren Johnson, “5 Bleeding-Edge Brands That Are Infusing Retail With Artificial Intelligence”, https://www.adweek.com/digital/5-bleeding-edge-brands-are-infusing-retail-artificial-intelligence-175312/, accessed Nov 2018
[8] Daniela Hernandez and Ted Greenwald, “IBM Has a Watson Dilemma”, in The Wall Street Journal, https://www.wsj.com/articles/ibm-bet-billions-that-watson-could-improve-cancer-treatment-it-hasnt-worked-1533961147, accessed Nov 2018
I was really intrigued by this piece – fashion is one place where I didn’t expect to see relevance for machine learning, but clearly TNF is finding ways to drive sales and upsell customers with precisely these tools!
To address your question, I think TNF should only consider bringing machine learning to its brick-and-mortar stores in a very conservative way. I’d expect customers who come to a retail outlet in person are more traditional shoppers that may be off-put by non-human recommendation engines telling them what to try on or buy. However, if employees in the store were given tablets that can provide this information seamlessly (e.g., a man+machine team), that could be a high quality customer experience and a way for employees to focus on selling rather than staying up-to-date on every single line in the TNF product assortment.
In response to your first question, I believe this should be leveraged to transform the in-person shopping experience. I see this technology as the entry point for the customer and as a way to customize the shopping experience without burdening the organization with high employee costs. The problem I see with this is if all humans were eliminated from the in-person shopping experience. Although we could use this technology, I do not believe we should entirely phase out humans. Particularly with the clothing industry, consumers find customer service associates to be a differentiation in the marketplace.
In response to your first question, I believe the XPS experience can be a step towards redefining the in-store shopping experience as well. As mentioned here (https://www.forbes.com/sites/stevendennis/2018/03/19/physical-retail-is-not-dead-boring-retail-is-understanding-retails-great-bifurcation/), retail isn’t dead. Despite the rise of e-commerce, companies like TNF will need to figure out how to utilize the data they collect on their consumers and overall industry trends to create a superior buying experience. Customers still want the trialability of a technical jacket in-store. I think they should use ML to better understand their customers’ online and retail buying trends to push more consumers into their stores and more properly stock the items they are likely to buy.
Regarding your second question, I see tradeoffs for both options but ultimately I don’t think TNF should expand the technology into other retail partners – while expanding the technology to retail partners could help TNF increase their data inputs and potentially help diminish the “echo chamber” risk, I worry that expanding too soon could have two detrimental effects. One, if the technology is expanded before the bugs have been worked out, the technology could have the opposite of the intended effect by frustrating both customers and retail partners alike. Second, in the case that TNF manages to optimize the technology, having it on their own website and not on retailers would enable them to have a unique competitive advantage, driving on site sales, which could help drive a hugely profitable channel.