How do you look for answers which are hard to verbalize? For example: What does this symbol mean? What’s the story behind that landmark at the corner of the street? Where can I buy the cool sweater I just saw?
For a normal person, visual cues trigger curiosity, and such questions follow suit. But, ‘technically’ translating them into verbiage and finding answers can be an uphill task. Visual Search does help to solve this in many ways. Images run as queries and image recognition tools return answers like Wikipedia pages, other similar images, text or video.
Search providers solve these problems using AI-based neural networks and large repositories of images and tagged image libraries. The accuracy on these searches has hence over time, improved tremendously.
With companies like Google, Amazon, Bing, eBay, and Pinterest launching major apps, visual search has taken off in a huge way. Compared to the overall hundreds of billions of text searches, only about 1 billion searches are image in any given month. However, the growth of this feature is predicted to accelerate in the coming years.
Pinterest cited a 46% improvement in relevance using “flashlight” searches, which let users crop search a part of a larger image. Further, it said, that visual searches went from 250 mil/month a year ago to 650 mil/month in Feb of 2018 on its platform alone.
What’s more? Over time search engines have become adept at predicting the implicit question behind an image search. A sign in a foreign language might be a request for translation, or the picture of parsley maybe a query for recipes or tips for gardening.
With all this great information, e-commerce and retailers are ahead of the curve in implementing use cases.
They are using visual commerce — where image searches, videos, AR and VR are serving as cues for purchase intent. As consumers have become more comfortable with images, and technology better at interpreting them, images have become more central in e-commerce. In a PowerReviews survey of US internet users in September 2016, 40% of respondents said they “always” search for visual content, and a further 32% “regularly” search for it.
At the backend, tagging images have always been a very crucial part in ensuring the most relevant ones show up when searched, and now retail outlets have added more depth by appending detailed, descriptive tags. If you’re looking for a pair of shoes, sites now try to decipher what about the shoes sets it apart — does it have heels? What color is the material and what’s the shape?
That most social media now allow links from photos to e-commerce landing pages where users can ‘Shop the Look’ at the click of a button is known and gaining steady popularity. However, the more interesting implementations have been from a number of individual retailers. Houzz has ‘Visual Match’ which allows users to click a real-world picture and search on their website catalog for similar products. One no longer needs to type out long drawn descriptions such as ‘White Coffee Table with Drawers’. Similar technologies are being implemented by retailers like Toy R Us and The Home Depot. However, the search time, search result quality and the factors that go into performing a search query are varied — meaning the experience is different for a customer with each outlet. Some have slower searches, but high relevancy. However, some others are quick but have a lower percentage of accuracy.
Another interesting application is using the image of a product to search for reviews and recommendations. There have been polarizing views on this — some users find it ‘cool’ with some others referring to it as ‘creepy’. But, the overall increase in adoption indicates its usefulness is undeniable.
Pinterest allows users to visually navigate through its content, which is to say users can search for similar visual aesthetics in home décor or even find hairstyles using images of skin tone. Amazon has tied up with Snapchat to show ‘Amazon Cards’ for products that a user snaps through the app.
From a business impact point of view, currently as we stand, the predicted ad revenue growth of visual search is 865B $ by 2022, however, this is only a fraction of the overall search. This, however, isn’t to say that visual search is not worth our attention. Far from it, many retailers like Forever21 have recorded almost a 20–30% increase in conversion rates for the product categories where they have tested visual search.
There are, however, two key factors that go into successful implementation. Firstly, the user flow needs to be intuitive and secondly, usage needs to be easy. If it ever comes to a point where brands have to explain to users how and why they need to use a particular technology, in this case, aspects of visual search, it’s like going backward.
The rate at which these technologies are catching up, it’s only going to get smoother and more sophisticated from here. And as more and more interesting ways of implementing visual search emerge — customer’s lives are set to become easier, and we have a lot to look forward to.