One of the best ways to find the latest fashion trends is to walk into a store at the start of a new season. Almost every brand follows a similar layout — you have your discount racks at the back, the staple essentials in their usual places, and front and centre you have the latest styles and colours of the season. The ones that make it so much harder to walk by without adding them to your cart. It’s easier to resist if you’re closer to the end of the month, of course. Rebecca Bloomwood would know. Store windows and stylishly dressed mannequins make it easy to showcase the latest products when you’re in retail. When you’re online though, it’s a whole new ball game. Especially in the context of AI and machine learning, where products are ranked based on performance, how easy (or difficult) is it for brands to encourage product discovery for fresh, new arrivals in a product catalog that runs into thousands of SKUs?
Relevance, as Solved by AIOne of the biggest challenges faced by eCommerce is to get shoppers to discover the products they are most likely to purchase. Being online has its obvious advantages, but is wrought with a bunch of challenges too. You don’t have a store associate gauging the shopper’s every move, coaxing and guiding them towards a product they’re already weak on. With eCommerce, you have to figure that bit out on your own with the data you have about the shopper. But how do we get to a place of knowledge about what each one of those thousand shoppers is looking for without asking them? We know it, because they tell us. Semantic search, involves understanding the context of what is typed into a search box. It goes beyond a simple text match algorithm that simply picks up products that match the terms entered by the shopper, but intelligently understands the context and intent behind the search. Here’s an example of intelligent search. While looking for a ‘green wrap dress’, the system breaks down the search term to understand that green, wrap, and dress are three different attributes of the query where green is the colour, wrap is the style and dress is the category. The products that come up are ranked based on how close they are to what the shopper was looking for.
AI and Merchandising: the new Mac & CheeseBut what happens in the absence of sufficient data? If the shopper is visiting the site for the first time, or if there are new products that simply haven’t had the chance to climb the ranks in terms of popularity. Say hello to AI-assisted merchandising. In a typical search scenario, new products are automatically added at the front or back of the results making it harder for them to surface during search or while browsing through categories, unless there’s a dedicated page for new arrivals. There isn’t enough search volume built for these products and as a result, they are buried under existing products that are more popular. AI-assisted merchandising helps tackle such situations. At Unbxd, we assign a Freshness score to every product in the catalog.
What does a freshness score do?It gives a fighting chance to every new, freshly added product in the catalog. Newer products are assigned a higher freshness score giving them a head start when they get added to the catalog. From there on, visitor feedback is aggressively directed to products with higher freshness scores to gauge their performance. Over a period of time, the newbies get significant exposure to pit them against old timers in the catalog. Once there is sufficient feedback on performance, the results are recalibrated based on set performance indicators such as conversion rates, overall popularity, and other similar metrics.