The death of size: why high fashion and high technology are the perfect fit
LONDON, United Kingdom, October 31 – High fashion meets high technology at the launch of Fit Match, a smart search engine add-on from Rakuten Fits Me. Company CEO Vicky Zadeh, CTO Mike Kimberley and guest keynote speaker Alexandra Schulman OBE (former editor-in-chief of the British edition of Vogue magazine) outline key challenges for online fashion retailers at technology launch event at the Tate Modern. Artificial intelligence, shape-shifting robots and data science are combining with improved knowledge of the customer, body shape modelling and ‘behaviour of fabrics’ into highly innovative algorithms. A ‘show me what fits’ filter seeks to solve an enduring challenge. With a lack of standards in garment sizing and variation in cut and style, how can customers confidently buy fashion products online; minimising returns and maximising value for the customer and the retailer?
Alexandra Schulman elucidated the sizing consistency challenge personally and succinctly. She described being “a 10 in Marks and Spencer, a small in Gap, an 8 in America, a 6 in Dolce and Gabbana in Italy, a 12 in Marc Jacobs…” “Buying clothes is an exercise in experimentation” and “overly time consuming and frequently frustrating.” The reasons for such diversity between brands, even within the same brand across geographies has complex sociological explanations. In this context, Schulman posed an interesting and disruptive question “would it be helpful if we could ditch the idea of sizing and move more towards the concept of fit?” Size is somehow constraining, conforming, imposed by the garment on its wearer. Fit somewhat ‘flips this dynamic’, the garment conforming to the wearer. Is the concept of size therefore anachronistic?
In thinking about a traditional online search experience, a customer might typically use brand, colour, cost and size as key attributes. But as Schulman insightfully questioned is ‘size’ simply too confused and too constraining to be useful? Is the customer likely to end up with a poorly fitting garment, tedious returns processes and an ultimately disappointing experience? Filtering search results by ‘what fits’ is an interesting (if technically challenging) response, a step towards an automated personal shopper, au fait with personal shape, taste and flattering products.
In imaging such an intelligent personal assistant, Schulman cautioned that the algorithms should not constrain choice and personal creativity. Maintaining the balance of ‘who is in charge’ is important, but the risk of a hegemonic assistant is perhaps unfounded. Exposure to fashion ideas in the high street, through friends, magazines, traditional and social media offsets the risk of ‘constraining the search beam’.
The technical challenges of algorithmically determining ‘fit’, were unpacked by CTO Mike Kimberley. The company, founded in 2009 in the Estonian tech-scene has been using highly sophisticated robotic mannequins to simulate over 100,000 different body shapes. A large sample of garment types (170,000 garments from 900 retailers) and expertise from fabric specialists being a key input in defining the data points and data science algorithms. This is a complex problem space, even considering the basic characteristics of fabrics from natural to man-man fibres (such as stretch factor).
Age, height and weight are the key data points used to ‘model’ the customer, along with a body type selector and data from historical purchases. Considering adult shoppers, height is likely static (but not so in children), age is linearly predictable, but weight may be highly variable. Pregnancy introduces an interesting ‘fit’ challenge, and tracking weight data points could highlight improving or deteriorating health (in all customer demographics). A common concern in data science is the range of acceptable inference, balanced with customer privacy and consent. Fit Match anonymises data, thereby negating many of these concerns.
Technically enabling the ‘fit’ filter on top of search is non-invasive and the retailer retains full control of the search experience. In trialling (or launching) it is also simple to create a ‘split test’, whereby a group of customers receive unfiltered search results and another group receive results filtered by ‘fit’. In this way, a retailer could test the efficacy and benefits of the filter, looking at conversion rates, returns and customer satisfaction as key measures.
Technological disruption in online fashion retailing is vibrant. Rakuten, the 3rd largest commerce Marketplace in the world bringing the Fits Me product to market is an important signal. Competition in online retailing is fierce and the ‘arms race’ for smart technologies ever accelerating. Improving customer satisfaction and driving out inefficiencies (such as the frustration of returned goods) is a win-win for both the retailer and the consumer. The underlying reasons for size variation and size obsession are perhaps untouched. However, a focus on ‘fit’ and away from ‘size’ might provide an important inflection point.