The human eye is a wonder of evolution. The human eye can see a single candle flickering 2.6 km away (Krisciunas and Carona, 2015) and the human brain can process images 60,000 times faster than text (D.R. Vogel, G.W. Dickson, and J.A. Lehman, 1986). What does this mean for your customer experience and how can visual search help increase action by your customers? We delve into some interesting research below.
As early as 1986, researchers from the Management Information Systems Research Center at the University of Minnesota and at 3M (Vogel et al., 1986) have shown decisively how good visuals are able to affect perception, attention and action. Based on viewers’ responses to presentation of visual aids, the results were able to show a positive change in viewers comprehension and action.
What would this mean for your eCommerce business? When a user reaches a decision point, browsing products for example, a visual element will be more persuasive and more likely to capture their interest and attention, which would lead to a positive reaction.
When searching for a product, your brain can not only process quickly but it can also easily spot what is important. The textual elements of any product are vitally important and are able to tell your customer technical and emotional details of a product. However, customers prefer to sort products using images before they reach the text.
The importance of visual search is evident in a wide range of eCommerce applications. One Airbnb-host-turned-consultant stated that “photographs are the single most important factor for marketing a vacation property”. Another study decided to test if visual information provides enough technical quality and content to boost demand (X. Wang, J. He, and R. Grewal, 2020). Using a convolutional neural network (an AI modelling technique), to process images across Airbnb, they investigated the effect of images on demand. Unsurprisingly, they found that the effect of simply showing an image of a living room, translated into a 13% increase in the booking rate, equivalent to about two additional nights of stay. Overall, they found that a clear image of the interior design of a property increased the booking rate by 38% over a 16-day, year-end holiday period. Furthermore, increased booking rates associated with images led to 784 USD more revenues in that same time period.
Machine learning approaches can help determine the best images for your customers, looking at quantifiable measures of an image’s characteristics such as clarity and brightness to determine the best images. Your customers will readily identify image content, and advanced AI tools can help determine the most pleasing images for a use leading to greater revenue and engagement.
The process of setting up the online space for your store can be complicated and will involve many manual processes. Manual processes are repetitive and boring tasks, which can invariably result in errors. For example, most sellers must manually tag all images uploaded to a site. If the image is mis-tagged or is missing a tag, most search tools will be unable to return a result. If your customers cannot find what they are looking for, then they won’t be able to buy your products.
What if you replace your manual tagging with visual search, will it be accurate? Any visual search tool is essentially an unsupervised learning task. Another study attempted to discover if the computer vision tools were up to the task (F. Li, S. Kant, S. Araki, S. Bangera and S.S. Shukla, 2020). They note that typical visual search algorithms such as autoencoders can be used to generate the latent features of images. They aimed to test if simple AI tools were able to accurately generate search results and what they found was a good degree of accuracy.
The model worked well in predicting the sub-categories of the products as well as predicting the article type. This result was close to a human level of accuracy. So, how do you improve this? A good AI practitioner can fine tune a model on your own data set and unique set of products to drive higher than human accurate predictions in many cases. Over time, a well-trained and retrained model will become even better. Visual AI search can be used to effectively speed your time to market and reduce errors by removing the tedious and laborious tagging processes associated with manual uploads.
Visual Search has evolved rapidly over the past few years. A skilled practitioner can transform your online images to powerful, actionable insights. Images have been shown to drive people to take action, improve their comprehension and increase satisfaction. The AI approach to image processing can serve the right images to your customers, increasing their happiness, your revenue and speeding your individual time to market.
Get in touch with our team today to understand how you can utilize Visual Search within your business.
Want to keep updated on the latest happenings in AI and Programmatic? Subscribe to our updates and stay in the loop