Ecommerce has become highly sophisticated, and the most successful retailers and marketplaces use a combination of big data analytics and emerging Internet based technologies such as real-time customer service, dynamic pricing and personalised offers to help ensure the commercial success of these operations. But if there’s one thing guaranteed to frustrate customers and prompt them to move to a competitor’s site, it’s the inability to find the product or service they are looking for.
Improved interaction with customers is not just about knowing their preferences or providing them with great delivery options, they also need a fast and intuitive response to their search enquiry during the search and buying process, and more importantly, for the resulting product or service information to accurately match their needs.
It’s not an easy task. In the last ten years’ the number of products listed on an average ecommerce site has increased from hundreds to hundreds of thousands, even millions, and that number is only set to increase further. On top of that, ecommerce retailers and marketplaces have to serve the multiple channels that customers are using for their interactions with the same degree of efficiency, and refresh their product cycles more and more frequently.
To meet these increasing demands, companies have invested in larger teams to manage their analytics and customer experience programmes. They have also increased their manpower budgets to allow for manual inputting of product data, speed up their time to market and stay ahead of the competition. It’s no wonder that retailers are increasingly asking how they can continue to scale up, manage the weight of that product data and still tick all of their customers’ boxes.
Which is where artificial intelligence (AI) comes in. AI powered platforms that enable data to be cleansed, identified, categorised, de-duplicated and enriched are not new. They have been around for the past four or five years but putting them to practical use in business, particularly in ecommerce has been limited; the system processing speeds needed to support these platforms have only recently caught up.
AI solutions designed for ecommerce use machine learning algorithms to provide the tools that the sector needs to meet customer demands, save time and save money. The solutions come as a blank sheet, and are taught and modelled to solve challenges with neural networks that simulate human decision making. They don’t come out of a box ready to revolutionise a retailer’s Product Information Management (PIM) issues, instead, they have to be trained to recognise categories, descriptions and labels.
The more that AI ecommerce solutions are fed, the more intelligent and useful they become. A customer, for example, could click on a picture of a particular pair of jeans when browsing on a site, and the AI solution would recognise all other jeans of that kind, and the colours and different styles available, and display them in milliseconds. The addition of text, for example, ‘straight leg jeans’ would add further clarification, and help to more closely pinpoint exactly what the customer is looking for.
AI solutions can also supplement a retailer’s own product information with additional facts scraped from the Internet in order to enrich their core data and ultimately allow retailers to more accurately categorise their products to improve the customer experience on their website. Crucially, they also continually check that the categories are still the right ones for a retailer so that products are given maximum exposure. This is particularly important for vendors selling products on marketplaces such as Amazon and eBay, where product details have to be mapped to the right classifications and labels utilised by those marketplaces. A simple labelling mistake can lead to the loss of vital selling opportunities as customers get frustrated with the lack of an intuitive response and move on.
As the core data grows, the easier it becomes to align product information with the channels that customers are using from mobiles and tablets through to apps. Each of these channels demands different forms of categorisation, and the AI system will learn to respond appropriately and in addition, display augmented information that allows ecommerce retailers to upsell with more accuracy.
For ecommerce managers looking to implement AI into their own product information management, it will be important to minimise disruption to their existing processes. They need to identify solutions that do not impact on their existing architectural landscape and can be accessed quickly and easily through the simple downloading and uploading of a CSV or an API to the product database.
But apart from the highly beneficial effect that AI can have on accurately categorising and enriching product data and therefore improving the customer experience, there is an additional bonus. AI solutions can go a long way to helping online retailers make better use of the vast amounts of additional unstructured data that is being collected and stored but not utilised to its full potential. Those same algorithms that are being taught to recognise and categorise products, can also be applied to many other back office functions within retail operations, that are simply too cumbersome to be managed by humans.
Retailers understand the need to leverage product data to achieve commerce success, and they know that the more clearly a product is described, the more likely it is it will be found by interested buyers. But bringing that clarification to retail data of all kinds through AI has even more powerful potential.
Henk-Jan van der Weide is co-founder of Artifam