After the holiday season it is common to experience a post-shopping hangover of sorts. You know, the aftermath of a feeling that you had to engage too much in commercial consumption. And the feeling of frustration in shopping is most often associated with the relentless searching for the things you want and then being unable to find them.
With these thoughts in mind we have been building CartSkill — an Artificial Intelligence based system to make the online product discovery process easier for people (you can access the demo here). With the choice and selection that exists today it is hardly the issue that certain products, that people want, do not exist online. They most certainly exist and increasingly can also be shipped to destinations worldwide. The challenge remains in finding the product that one wants.
The problem of product discovery (especially in fashion) boils down to a simple point — you only know what you want when you see it.
It is as if seeing a product removes the haze in one’s mind and makes it crystal clear that this is the exact product that I am looking for.
There is something in seeing a product and processing its looks and qualities after which we make a connection whether the product is something that we want or not. It is present in any category where you buy items based on their beauty — jackets, handbags, shirts, shoes, glasses, furniture — just to name a few. The end goal for finding many of these products is just to find something that you like.
To make a point, I was recently looking for a watch to buy online and I have no doubt that Amazon had the watch I was looking for. However, to get to the product that I was looking for, proved to be difficult. Sure Amazon has a vast array of checkboxes that one could make a detailed selection of — band material, band color and width, style and case diameter. But the simple truth when looking at these categories is that it is very difficult to make a selection. Why? Because I simply don’t know. I have a vision of the watch I am looking for in my head but I can only know it when I see it.
And therein lies the point. We are visual creatures and determine whether we like things or not in a fraction of a second while being unable to describe beforehand what we are exactly looking for. Describing products that a person wants with words or categorizing them with checkboxes is tough. And as such it is difficult to discover products online.
As an answer to the problem we have been training machines to see products like we see them and recommend us products we should like. Utilizing our deep learning networks we are helping the machine to understand what kind of products it is seeing. As it learns, it is also able to say which products are visually similar to one another. By doing so, it can recommend you similar products in a blink of an eye and suggest you products that you actually want so that next year Santa’s job will be hopefully easier 🙂