When the details of an object are mapped with fine-grained specificity through tagging, a search to find that object is made exponentially easier. More than other objects, the complexity of fashion renders tagging an exhaustive task. To optimize innovative CX solutions for the realm of fashion, the foundation of tagging needs to be exceptionally thorough.
Ecommerce search technology geared towards fashion is presented with major obstacles due to the nature of shopping for fashion. Individual preferences of style, fit, and feel are notoriously difficult to predict. Naming apparel can be subjective - what is a romper to one may be a jumper to another. The depth of details is vast and part of the complication of factors in fashion search that come into play. Limited by time, resources, and fashion-related language, retailers listing items online constrain products into categories and a short list of attributes. That is why the searches via a search query for a piece of apparel on the site of a retailer’s online store inevitably pulls up some unrelated items. For example, a “red bandage dress“ may surface a “ fitted black dress” amongst other red dresses made of stretch bodycon material. Those seemingly random items most likely have certain keyword attributes in the original query such as “red” or “bandage”. The search algorithm may additionally pull up items that are not dresses. Ultimately, customers are still being shown items that they hadn’t intended on finding.
The sense of dissonance that generally follows after a couple of failed search attempts leads shoppers to abandon their carts. They may bounce to the site of a completely different retailer hoping the next store will carry their item or offer a better customer experience with easier search functionality. These behaviors leave many retailers with lost sales and a low search to basket conversion rate. Since customers that engage with search have a higher intent to purchase, savvy retailers will want to maximize the solutions for those buyers.
To consider the deceptively simple act of producing accurate results for a person who is searching online within a fashion site, there are two major problems to be solved in order to make it work. The first is understanding the customer's query and the features they are looking for. This could range from length, color and cut to more advanced categories such as style, season and trend. The second is to identify the retailer's fashion items and the features of those items. For example, a shopper may type “lace spaghetti strap dress “ in order to search for this or a similar dress on Superdown.com.
The traditional way of pulling up this item for the shopper is with tags to identify the key categories and a set of simple attributes. Superdown would tag the dress with attributes or keywords of “white,” “lace,” and “spaghetti strap” and of course, the category of “dress.” But what about the pattern, type of lace, style (i.e. bustier) or a specific length style (i.e. mini dress)? Shoppers wouldn’t be able to search for those exact details.
Velou in collaboration with fashion experts, found a precise method of tagging that analyzes every part of clothing, including the type of weave and style of neckline along with hundreds of other style attributes. This level of hyper tagging, through a solution called Product Data Enrichment, is extensive and if it were to be done manually, would be extraordinarily time intensive and slow. With artificial intelligence and machine learning, data scientists are able to automate the process of “enriching” a product with its details, thereby readying the products for listing online as quickly as the retailer wants to make the items available for purchase. This means that retailers can speed up time to market and scale up the number of items in their online store, bringing a larger fashion inventory to their customers at a much faster rate. In addition, having correct and complete metadata for a product page optimizes SEO for web crawlers and boosts traffic from customers looking for similar products. The amount of items matching the shopper’s search and the ability for a shopper to be able to quickly locate the exact items they want ultimately leads to increased sales.
As if tagging products weren’t a challenge enough, “ordering” the search then becomes a completely different ball game. What are the items that should be displayed for a particular customer to help narrow their search and in what order should those items be displayed? To this, Velou’s CTO Sumith Gunasekara responds, “to understand the items, we use both text analysis and image analysis using Computer Vision technology. We are in the process of continuously fine tuning these elements through training algorithms or what we know as machine learning.” Providing the relevant items isn’t enough. The technology continues to evolve as it “understands” the customer’s preferences through their search behavior. In the universe of potential fashion available out there, having a machine assisting customers with the items that best fits their preferences is a competitive advantage for any retailer.
With artificial intelligence, technology can predict the right items in the right order for an individual shopper each and every time. The live results are shown as the shopper types out the words or phrase in real-time on the site. The search queries, along with the assistance of any facet checkboxes selected and suggestions, become data for personalization. The value in providing a personalized UI experience is that it minimizes the shopper's effort in having to further define and refine their search.
Considering all these factors, fashion search is not an easy problem to solve. It combines the technology with fashion-related domain knowledge. After all, it does come down to human behavior (involving considerations of fit, style, personal preference, the feel of the garment when worn and fashion sensibility). That is what makes finding clothing for anyone a tough proposition. The more attributes identified as part of a product's anatomy, the easier it is to find the product that a shopper intends on purchasing. This is key to increasing search to basket conversions and revenue for ecommerce retailers.
Velou’s mission of harnessing next level technology to create meaningful ecommerce experiences will have shoppers discovering the items they love in no time.