Disappointing 'search relevancy' cost retailers over $2 trillion globally

Google's reputation for delivering highly relevant search results has set an extraordinary benchmark for user expectations across the internet.

Whether our search words are long or short, simple or complex, there's a sense that Google 'gets us' and can intuitively provide exactly what we're looking for. That means there are high expectations for search, and people anticipate this excellence across the internet.

However, online retailers are failing to live up to them. According to research from Kin + Carta and Google Cloud, "search relevancy" is the main difficulty for 85 per cent of the UK's leading retailers when it comes to product discovery on their e-commerce sites.

These difficulties in delivering relevant results are causing search abandonment concerns, which are costing most e-commerce retailers millions of pounds. According to Google data, it costs retailers more than $2 trillion every year globally.

On how to fully harness the potential of AI, Karl Hampson, Chief Technology Officer, Data & AI, Kin + Carta highlighted some strategies for optimising Google searches. One approach is using 'tail' queries to interpret.

Hampson stated that when users enter long-tail search searches with three or more terms, they frequently use natural language, citing 'date night dress in small' as an example. Longer questions like these are becoming more common as customers become more accustomed to interacting with AI chatbots in a conversational manner, so it's critical that stores' search functions can handle them successfully, he said.

Hampson emphasised the importance of bridging the gap between user intent and product information. Traditional search functions often struggle to understand the language used in product catalogues, making it difficult to produce accurate results. However, employing large language models (LLMs) can help improve search relevancy.

According to Hampson, finding motivation is another key aspect of search. He said users are also increasingly turning to search functions for inspiration, essentially looking for product inspiration using terms like 'Christmas gifts for dad'. This is almost the inverse of the problem with long search terms: the lack of precision makes matching the query to relevant products in the inventory exceedingly difficult, he added. Additionally, Hampson noted that generative AI can help shops in this situation. LLMs can accurately grasp the intent behind these search queries and offer the best answers since they've been trained on such large data sets.

Addressing typographical errors is also important. The occasional typo or spelling error is unavoidable in online communication, especially when users are pressed for time. The Chief Technology Officer, Data & AI, Kin + Carta said while a friend may be able to overlook a misspelling via WhatsApp and still respond to the user's message appropriately, traditional search engines may struggle.

The consequences of a failed search function are enormous. According to Google's research. Three out of every four clients will go elsewhere. 77 per cent of clients do not return. The research noted that If they can't find even one item, 52 per cent abandon the cart. Generative AI is the ideal approach for resolving the vast majority of these issues, Hampson stated. However, the advantages go beyond simply correcting errors and making search more user-friendly.

Hampson further stated that LLM technologies, like Google Cloud Retail Search, can increase conversion stats, even further by personalising results for each user, resulting in a more intelligent experience – a significant distinction vs legacy search. When compared to their traditional search counterparts, he stressed that retailers are reporting up to a 20 per cent increase in revenue per visitor when using this technology.

It also means that search and browsing results will be ranked more intelligently, he added. Hampson also mentioned that the results will determine what a customer is most likely to buy and what generates the most income for the business. With time, it can become hyper-personalised to the individual client rather than just a cohort, lowering cart abandonment while raising the likelihood of repeat business.

With both (generative AI and LLMs) of these functional pieces in place, Hampson said retailing becomes more about steering search to what you want to sell and less about attempting to plaster your search engine with hard-coded rules. With generative AI in place, he stressed that businesses can begin delivering these 'wow' moments rapidly, reducing pain points from their existing search to create an experience that simply works.

Hampson also noted that LLMs are quickly becoming an essential component of e-commerce search systems, requiring little to no human involvement once implemented. He warned that retailers who fail to keep up with this transition risk losing millions of pounds in income.