Leveraging AI to improve product discovery for fashion e-commerce

The pandemic has altered the way many consumers shop for fashion for good. Consumers are demanding more product variety and shopping more online, with apparel and accessories accounting for 29.5% of all commerce sales in 2021 in the US alone, according to Statista. However, with such a sudden increase in products offered online, one of the main issues consumers face is not being able to find the product they're looking for. 94% of online shoppers receive irrelevant results, with 72% of websites falling short of search expectations. What's more, 76% of consumers are likely to abandon a website if they cannot find the product they're looking for, with more than half choosing not to return after a poor search experience. Poor product discovery is costing e-commerce companies billions in lost sales as they fail to live up to consumers expectations.

Searching for and being able to find the right products remains a very frustrating process for customers when it should be an enjoyable, seamless experience. Customers expect to find the products they are looking for based on their search keywords, but things can already go wrong during their first search query. For example, when searching for a "short black dress" in the search bar or on a retailer's website, customers will likely see black dresses - in addition to black jumpsuits, skirts and shorts. Overwhelmed with all the options, customers are more likely to abandon their search or purchase several options and return them all due to poor product data.

Poor product data, often given to online fashion retailers by their vendors, is often incomplete, inconsistent and inaccurate. Online fashion retailers are faced with several challenges, from missing labelled attribute fields, such as lacking the name of a colour, to different product names to wrongly input data. Together this leads to a bad search experience, which leads to lost sales. A recent report from Google found that approximately $300 billion in sales are lost in the US alone each year due to bad online search experiences. Further data indicate that up to 94% of customers abandon shopping sessions due to irrelevant results, with as many as 72% of all e-commerce sites falling short of customers' search expectations. To ensure they finalise sales, online fashion retailers need to focus on improving product data, which begins with having the correct data. This is where product data enrichment platforms powered by Visual AI, like Pixyle.ai, come in.

AI fashion e-commerce tools and solutions are a leading necessity for retailers today for companies looking to secure their bottom lines. Helping fashion retailers improve the discoverability of their products, Pixyle's solutions enable customers to find exactly what they are searching for. Leveraging AI, machine learning and computer vision to detect fashion products in images, Pixyle can enrich product attributes from images.

Based on neural networks, the machine is taught by Pixyle’s AI algorithms to see and interpret images exactly how humans would. Building a proprietary dataset of more than a million carefully annotated images, all images are labelled in-house to ensure the highest level of quality control. Then using this dataset to train their machine learning algorithms, Pixyle's solutions can deliver detailed product attributes with up to 95% accuracy.

Offering three AI-driven solutions based on image recognition technology, each one is designed to assist online fashion retailers in strengthening product discovery. By implementing automatic tagging, online retailers can efficiently enrich their product data.

Starting with having a detailed product taxonomy in place, this is a form of structure through which Pixyle classifies and organises products in different categories and attributes in a way that reflects how customers search for products online. Constantly evolving and expanding, Pixyle works with leading fashion experts and data scientists to ensure its taxonomy is up to date with the latest trends. Currently covering more than 20,000 product attributes, Pixyle can capture any specific search query that a customer may have. The AI-powered image recognition system then accurately recognises fashion products in images and provides insightful data, letting retailers such as Depop and Otrium automatically tag thousands and thousands of photos in a matter of minutes. Replacing manual, often repetitive and costly labelling with Pixyle's visual AI helps retailers rapidly process their entire catalogue, saving time, resources and energy. The accurate and detailed descriptive product tags improve e-commerce store filters and offer customers precise search results while improving catalogue management.

With visual search technology, customers can find the products they are searching for by simply uploading an image. Transforming customers' wish lists and fashion inspiration boards into a seamless, intuitive online shopping experience makes it even easier for customers to find what they are searching for. Text searching does not always provide the exact context, and convenience customers expect when searching for a particular product online, partly due to the vocabulary gap caused by different words and descriptions for items. Visual search lets customers upload an image of whatever product they want and instantly find the most similar item from the online retailer, thereby increasing conversion rates.

At the same time, visual searching can also provide customers with the most similar-looking products available on a fashion retailer's website. 80% of executives surveyed in Deloitte's 2022 Retail Industry Outlook found that customers prioritise stock availability of products over brand loyalty, highlighting how vital product discovery is for sales. If the exact product is unavailable, similar-looking items should be offered as alternatives to maintain the sale. Visual searching also enhances customers' experience by creating a more visual and optimised shopping journey by removing any potential friction points.

On a mission to transform the way people discover products online by helping retailers build better search with visual AI, Pixyle has just begun its journey. "There is still so much we can do in product attribute enrichment and e-commerce product discovery," says Svetlana Kordumova at Pixyle. "We will continue building and expanding our (already extensive) fashion taxonomy to stay current with the latest trends while keeping our accuracy at the highest level possible."

In line with Pixyle's forward-looking view, the company is set to launch a new product shortly that will allow retailers to "read'' detailed information from a label. Using object character recognition (OCR) technology, the company will make it possible to recognise brand name, material composition and size, simply by scanning the label.

Learn how you can leverage AI-powered product attribute enrichment to connect shoppers to the items they're really looking for by implementing Pixyle's solutions today.