The wind is blowing, there is a screeching sound from under the floorboards and suddenly the lights go out. It’s that time of the year again, Halloween! Halloween is a fast growing global phenomenon with celebrations in Hong Kong Disneyland, Osaka’s Zombie Parade and over 1.7 billion USD spent in America on costumes alone.
The search for the perfect Halloween outfit can be a real scare. There are so many options. Do you go as a goblin, a witch or a participant in Squid Game? Squid Game appears to be the winner this year. So how do you make that decision? The first step is to take a look at what is on offer. Horror! You are confronted with thousands upon thousands of choices. One site we viewed had 91 pages of costumes in the “new” category alone. We love customer choice and for retailers that go all in on choice, they need to help their customers sort through the overwhelming variations. Categories and filtering may help, but there is a faster and more engaging way to deal with thousands of products and that is Visual Search. It is not a magic spell but can act like magic with how the results can delight and entertain your customers.
First, a little explanation. When you deploy a Visual Search solution, there are many decisions to be made. Software runs a machine, and a machine has to be told what to do. Do you classify the category first by sorting eyeballs in one pile and newt tails in the other? Do you look for matching color patterns? What goes well with blood splatters? Do you want something of a similar shape to a bone?
This seems like a lot to consider because as human beings, we jump from one idea to another without thinking deeply. First, you look at a nice color like orange. Then, you decide you want to dress as a pumpkin. Wait, no! You want to be Annoying Orange. As a human, we switch between categories without a second thought. This type of switching between distributions and forming new distributions to name something is easy for a person. Machines are not so good at that yet. You need to create very specific rules for the machines. And this is where a careful consultation with your AI service provider is helpful. By working together, you can decide how you want to engage your customers.
There can be lots of fun in the discovery stage. What if we like spider webs? A good computer vision model can sort through those thousands of images and find the spider web like objects of course. It will pick up the spiderweb lamps, spiderweb earrings and spiderweb decorations. It is not perfect and some silly results may appear. It may also pick up a lace dress for example, but that is the point! Serendipitous discovery is part of the fun. Humans love novelty but only if there are some familiar patterns to make the novelty interesting.
How else can we harness the novelty in Visual Search for the holidays? What happens if you search for googly eyes? No one usually wants googly eyes and most stores will not be selling them, at least not after Halloween but the search is really fun. Using our own technology, we searched through a set of fashion items and found some really fun results. As a human, we can quickly see the similarities between the googly eyes and the sunglasses (green background, off set black circles). You get lots of fun results, but what you can understand is that the results are not random. They follow rules that get us much closer to the product we are looking for even if we can’t say exactly.
Visual Search tools can be harnessed to create unique and fun experiences for clients and can also be harnessed to serve as a gateway to discovery for even the spookiest of holidays. Our team at Delvify is able to help your brand create that seamless (and fun!) experience for your consumers with our Visual Search and Recommendations tools. Enable yourself to quickly search for the products that you desire and even find inspiration from all the similar products.