Here’s a scenario for you: You’ve found the perfect “match” on Tinder, and want to make an impression on your first date. You want to start off by looking for a shirt and find one that you really like. So you click on it to look at its product detail page.
You start looking at the complete image and love everything that the model’s wearing — from the pants that work so well with that shirt, all the way down to the sneakers that look EXACTLY like the ones you’ve wanted for a while. You have no idea how to look for those sneakers — you don’t even know which brand they belong to! All you have is a vague description of the color and the style.
So you go through the painful process of clicking through innumerable pages and search results… without any luck! So after all that effort all that you’ve done, really, is waste lots of time without even buying that shirt. And don’t even get me started on the sneakers!
If only finding the right look was as easy as swiping right!
I’ve been there. And if your shoppers are anything like me, then you should know that it is extremely frustrating for them to see the products that they want, without being able to buy them.
Here’s the good news. We’ve managed to solve this problem with our Complete the Look module which provides cross product recommendation engine, that can power not just your product detail pages but also your cart and checkout pages.
Complete the Look understands your shoppers’ behavioral patterns and visual affinities to products, and also the rules of fashion, to deliver 1:1 personalization (personalization engine). So your shoppers see curated outfits that are foolproof — in terms of style, color, pattern combinations. Complete the Look acts as your website’s AI-powered styling tool, to showcase curated looks that they’ll love, taking visual merchandising to the next level.
We’ve noticed that while shoppers are inspired by the different looks offered as a part of the product images, they also want the individual elements making up the look to be “shoppable”.
On further observation, we realized some other aspects that are important to consider when targeting shoppers looking to buy ensembles:
Shoppers like to see accessories and other items of clothing that complement and complete a look, even if it’s not what the model is wearing. This is why cross product recommendations can significantly improve shopper engagement and have a direct impact on the basket size.
A little “visually similar” variety never hurt anybody. It is our understanding that it’s not just an item’s brand or price that shoppers fall in love with. More often than not, it’s also the visual merchandising that draws them. If a product within a particular look is sold out, they’re more likely to continue shopping on your site when you show them something that’s similar to products they like.
When in doubt, offer more choice. Shoppers want to see alternatives to the products showcased in a “look” even when all the original products are still available. This makes it easy for them to shop if something isn’t available in their size or preferred price range.
The versatility of products helps with decision-making. Shoppers like to see how versatile the product they are viewing really is. When you show them a product page that includes pairing suggestions in different colors — they’ll view more, click more, and buy more.
Minimizing navigation is key. While shoppers would like to gather different products within Complete the Look, they would also like to sift through them and add them all to their cart at once. This reduces the risk of the shopper abandoning your site during the course of their purchase journey.
Combining Behavioral Triggers With The Rules Of Fashion
As we developed an understanding of shopper behavior, we combined it with the rules of fashion to zero in on the parameters which would eventually allow our algorithms to curate ensembles without any manual intervention.
Visual affinity: This is a metric that is key to Complete the Look, and replicates the experience of having a personal stylist. We’ve developed a visual grammar by working closely with our in-house fashion and home stylists. This grammar allows the AI to pair the right top with the right bottom-wear, jewelry, and other accessories. This pairing is based on color temperature, the fit, the style, the patterns, the cuts, and much more. We approach it from a style perspective as opposed to just the attributes. So what you get in effect are formal ensembles for work, casual looks, 9 to 9 ensembles, evening wear ensembles and much more which are carefully curated by our artificial intelligence and computer vision engine.
Inter-product Correlation: This provides us with an understanding of how strongly two products are associated — whether they are usually bought together or independent of each other.
Price affinity: This allows us to group products that are termed similar in terms of their prices positioning within their categories.
For example, For example, a consumer who purchases expensive items in one category is likely to buy an expensive item in another category as well.
Brand affinity: This considers the shoppers’ preferred brands, and computes the similarity between two brand names to show relevant recommendations.
Use Case I: Getting the Ensemble That Compliments the Product
The shopper can view the curated looks and ensembles on the Product Detail Page (PDP) of each item. If an item is not available, the Vue.ai engine proposes a similar item:
Use Case II: Recommend Complimentary Items
The product recommendation engine also suggests complementary items to shoppers, based on their price preferences:
Use Case III: Recommend Similar Items
The product recommendation engine also suggest items that are similar to the original products the model is wearing within the display image. So you not only get to shop the look, but also get to see alternatives.
Use Case IV: Get Other Looks
This feature relates to the assumption that users want to see an article in more than one look/styling.
Complete the Look Works, And The Numbers Prove It!
We put the feature to the test, going live with it across some of our fashion, furniture and lifestyle customers. We learned that while it works well on the product pages, it yields incredible results on the cart/checkout page. If we were to look at some of the initial results of how it has performed, we learned that it has led to a 1.5x increase in the average order value (AOV) , with the average order size (AOS) doubled across the board.
Considering that this was a completely new feature for our customers, we’d say these numbers are very encouraging. And with a GMV contribution of 25K USD to the revenue from a single widget, we’re ensuring these retailers are not leaving any money on the table. And with our algorithms only getting smarter, we can safely state that AI curated ensembles are the future of retail.