Visual AI Blog Ecommerce Killed Textual Product Search… What Comes Next? Idan Pinto November 25, 2018 9 Min Read Product search innovations made outside of text search have made some incredible strides, but it’s only now that we can see the death of text-based product search square in our sights. Today’s visual tech is ready to take product search where it needs to go. Ecommerce pushed textual search over the edge because ecommerce sites cater to visual browsing, not text skimming. Textual search simply wasn’t built for the ecommerce consumer — it was built for crawling web pages with text on them. That’s why retailers have been trying their hand at ways to replace or enhance text searches for years. From the first barcode in 1974 to today’s futuristic AI, retailers have been pushing to make product search easier and more intuitive for ecommerce shoppers, but the successor to textual search wasn’t ready — until now. QR Codes and Barcodes: From Meatspace to Ecommerce Marketplace The basis of textual product search is that a shopper is describing something that exists in the real world in a way that will allow them to find that object online. The pitfalls to text search (and voice search) are obvious when it’s framed this way: language is subjective. One person’s red is another’s orange, one person’s settee is another’s loveseat. And when a customer has a very specific item in mind, the way they describe that item may not correspond with the way that item is tagged by a company for search. This is a problem that doesn’t really exist for shoppers in stores or shoppers who browse physical catalogs. In meatspace, customers can quickly scan racks or pages of clothing or ask a sales assistant to point them in the right direction. They have time, space, and conversation to guide them to items. To bring this ease to the ecommerce space, QR codes and barcodes offered a way out: if an item is fitted with a QR code or barcode, any customer can pull out their phone, scan the item, and pull up that item in digital space. Today, QR codes and barcodes are used to unlock discounts, comparison shop in-store, and save items encountered in-store for online browsing later. Take Zara’s app, which allows shoppers to scan and save anything they encounter in-store: [Source] This is a brilliant way to circumvent some of the basic issues behind textual product search — it seamlessly blends the physical and digital world for consumers. But there’s a big catch — it’s impossible to scan a QR code or barcode that isn’t right in front of you. QR and barcodes facilitate search, but many shoppers are not standing in front of an item with a barcode on it when they open a web browser to shop. Books, packaged foods, and nonperishables like candles and lotion might have a barcode stick around, but your friend’s shoes don’t, and the picture you save from Instagram doesn’t either. Product search with QR codes and barcodes are so popular in part because their technology has been accurate, available on mobile, and easy to use since it came out. At the end of the day, QR codes and barcodes aren’t an answer to the question of textual product search for ecommerce. They are great tools for melding meatspace and ecommerce marketplaces but will never be positioned to completely take over search. Quizzes: Just Right… for a Handful of Industries After you search for an item on a site, if you want more accurate results you typically are left at the mercy of search filtering. This can be extremely useful for some attributes, like price or size, which are concrete and easily defined. But layering on filters doesn’t always connect with the specificity of what customers want to search for, and it takes a consumer time and effort to search and then repeatedly filter until they find what they’re looking for. Quizzes turn this problem on their head. They are great for helping users find products that respond directly to their needs, especially when these might escape easy text search. For example, a quiz for hair dye color might show a range of skin tones and ask you to choose from those, then show a range of natural hair colors to choose from. This might result in a range of dyes not possible to search for via text alone. For example, L’Oreal’s hair dye quiz asks if a shopper currently has treated hair and by which process their hair was treated. This will automatically eliminate dyes that could interact with previous processes, something that isn’t going to come up in a text search: In this way, product quizzes roll search and filtering together in a way that’s gamified for the shopper. It gives more accurate results than a text search alone and allows a company to ask more detailed questions than they might be able to include in a search filter. Especially compared to QR codes or barcode scanning, quizzes are an easy byproduct of digital shopping because they don’t require any input other than the user answering questions. Anyone with an internet-connected computer can take one on an ecommerce website. For certain products, quizzes are probably going to rule the day for a long time. Hair dye, glasses, bras, skincare, and hair care products are all areas in which text search is probably always going to fall short. These are items that someone generally needs a personal consultation for, and that’s basically what a quiz is. Prose’s custom hair products are built entirely online with an easy quiz on their website. Quizzes are a knockout success for those areas but are not otherwise destined to take over ecommerce product search. A casual shopper browsing furniture, paint colors, sneakers, electronics, or bathing suits doesn’t need to take a quiz to have a look around. While quizzes can be a great way to direct people, they stymie browsing, especially for people who want to browse for many items at once, comparison shop, and quickly purchase. In addition, they aren’t as easy on mobile as they are on desktop. They can be heavy to load and clumsy to scroll through on a phone. For all its faults, text search is about as easy on mobile as it is on desktop or a tablet. Visual Search: The Best of Both Worlds Visual search has the ease of textual search, the mobile-friendly nature of QR codes and barcodes, and often has more accurate results than textual search. Why textual search has dominated for so long is that, until now, visual search AI hadn’t been accurate enough to usurp its text-based counterpart. No longer. Today, visual AI is accurate down to minute details, like the difference between a “glossy” and “shiny” fabric, hemlines, the metals jewelry is made out of, shoe type… anything that’s shown in an uploaded image that can be matched to an inventory item. This means that results from visual search are often faster and more accurate than text search. Plus, they don’t ask consumers to translate objects into text for a search anymore. If someone has a photo, snaps a picture, saves a screenshot, or favorites a post online, they can use it to search. This can happen on mobile or on desktop. Users can even snap a photo in-store and bring up comparison shopping online on mobile, just as they might with a barcode tool. Consumers think, shop, and spend based on visual cues and are more likely to be able to find reference images than they are to describe the exact item they’re looking for. So if someone is looking for a dupe of the lusted-after coat Idris Elba wears in Luther, they can just use a saved picture and find similar coats without having to describe the texture or color: Atterley showcases visual search based on a costume piece from Luther. Visual search is now easier, faster, and more accurate than text search, more versatile for product search than barcodes and QR readers. It’s readily available on mobile apps and easy to use on desktop. It is already encroaching on textual product’s search space, and major retailers have already put it in place. Visual search is poised and ready to take over. What About Voice Search? Voice search is on the rise, and it, like quizzes, has its own monopoly. Tech like Alexa and Google Home and, to a lesser extent Siri, etc., are designed to be operated by voice alone. There’s no competition between voice search and visual AI on Alexa because you aren’t going to use Alexa to snap a photo of a top to buy later. Voice search tends to suffer from the same downfall as textual search when it comes to finding a product because they’re both based on word descriptions. For this reason, voice product search is complementary to visual search at best. If you don’t know where to find a jacket you saw on your favorite influencer, you probably are not going to describe it to your phone and hope it can find a match. Conversely, when you need a quick answer to a cooking time question in the kitchen, you probably aren’t going to be using visual search — that’s voice’s domain. While it’s true that voice search is on the rise, visual AI and voice recognition tech are competing more against text search than they are each other. Get off a Sinking Ship Text search can still be a part of your site, but it should not be the only option for consumers anymore. On desktop and mobile, adding visual search with Syte is easy and fast and pulls your business ahead. Workarounds for textual search are over — the future is here.