ASOS.com is an independent online fashion and beauty retailer in the UK, which has sold ASOS and third-party brands to young adults Y2000. In addition to its iconoclastic, fashion-forward but budget-friendly range of clothing, it is known for offering value to its younger demographics, with free shipping based on order value thresholds and discounts for students. As an ecommerce portal with a specific product footprint, recommender systems are intrinsic to both customer and business value for ASOS.com.
We can conduct two scenario design exercises, the first focused on customer (shopper / site visitor) value, the second on business (shareholder) value. Recommender systems
Target users: Visitors to ASOS.com fall into one of several buckets: current customers, previous (churned) customers, new visitors, and erroneous traffic (which we can discount). While we can infer interest in fashion for anyone deliberately visiting the site, each of those target users may be in different stages of their purchase journeys.
Key goals: Depending on where they are on purchase path, site visitors may be checking the new seasons looks, trend-scouting, comparison shopping, pricing out merchandise, replacing existing pieces, bargain-hunting, checking on previously sold-out merchandise, etc.
How to help them accomplish those goals: Recommendations help with a variety of these behaviors and goals. ASOS.com already employs a “style feed” in a scrollable carousel along with conventional search bars to reduce the friction of IA page hierarchy navigation. “You might also like” and “buy the look” modules on section and product pages explicitly recommend products, and these could be augmented with further personalization. This personalization should build on the similarity matrix that matches either on-site browsing behaviors based on cookies from previous sessions or previous purchase / shopping behaviors signed-on account holders. Surfacing recommendations based a lookalike model will improve the salience of personalized “look” creation" as well as on-site SERP, providingshoppers a better experience.
Target users and key goals: Tantamount to a mirror image of the above for customer value. Given the importance of repeat purchase and the high cost of finding new customers, there’s likely greater emphasis on remaining relevant and top-of-mind to customers who have already made a purchase, who require smaller nudges to purchase again. Additionally, business value is predicated on legibility of user behavior across platforms and the ability to inform customer experience across those platforms through appropriate data capture.
How to help them accomplish those goals: Customer lifetime value (CLTV) is an important concept to many marketing-driven businesses, and to ecommerce in particular. It’s based on the idea that customers can be segmented by estimates of the total number of purchases and transaction value (i.e. profitability) that will occur over the course of their relationship with the business (i.e. lifetime). This framework is often used to identify existing customers and “nurture” them using content and promotions in order to maximize those customers’ value to the business. It can also be used to prospect potential customers, right-sizing marketing, advertising, and lead development expenditures based on the anticipated consumer value.
These two use cases - retention (churn mitigation, basket…) and acquisition - are supported by web technologies that instrument sites, apps, and inbound marketing channels for user-level tracking. These analytics technologies are used to track people (cookies, actually) who have visited the site and come into contact with advertising or media; and then, once accounts are created by those who “convert” (presumably for a purchase), to connect those cookies with marketing activity, preferably with history to date as well as on a go-forward basis.
How to help them accomplish those goals: Researchers have identified sophisticated, state-of-the-art approaches to evaluating the business value of customers and customer segments. Per a paper published in 2017, ASOS uses a combination of handcrafted features, ensemble regressors, and embedding of customers to factor in a dynamic product catalogue when providing estimate of future value of every customer on a daily basis. This is used to inform the activities and goals above, including personalizing offers, promotion value and timing, customer re-engagement, site personalization, etc.