At this week’s NRF Big Show in New York, Google Cloud introduced four AI technologies to help retailers transform their in-store shelf checking processes and enhance their ecommerce sites.

A shelf checking AI system, built on Google Cloud’s Vertex AI Vision, uses Google’s database of facts about people, places and things, giving retailers the ability to recognise billions of products to ensure in-store shelves are right-sized and well-stocked.

In an update to its Discovery AI, Google Cloud introduced a personalisation AI capability and AI-powered browse feature to help retailers upgrade their digital storefronts with more dynamic and intuitive shopping experiences.

Finally, Google Cloud’s added to its Recommendations AI machine-learning capabilities that empower retailers to optimise product ordering and recommendations panels dynamically on their ecommerce pages and deliver personalised suggestions for repeat purchases.

“Upheavals over the last few years have reshaped the retail landscape and the tools retailers need to be more efficient, more compelling to their customers, and less exposed to future shocks,” said Carrie Tharp, vice president at Google Cloud. “Despite uncertainty, the retail industry has enormous opportunity. The leaders of tomorrow will be those who address today’s most pressing in-store and online challenges with the newest technology tools, such as artificial intelligence and machine learning.”

The problem of low or no inventory on in-store shelves is a troubling one for retailers. According to a NielsenIQ analysis of on-shelf availability, empty shelves cost US retailers $82bn in missed sales in 2021 alone. While retailers have tried different shelf-checking technologies for years, their effectiveness has often been limited by the resources needed to create reliable AI models to detect and differentiate products, from the different flavours of jam and jelly, to the dozens of types of toothbrushes.

Now available in preview globally, Google Cloud’s AI-powered shelf checking can help retailers improve on-shelf product availability, provide better visibility into what their shelves actually look like, and help them understand where restocks are needed. Built on Google Cloud’s Vertex AI Vision and powered by two machine-learning models – a product recogniser and tag recogniser – the shelf checking AI lets retailers solve the problem of how to identify products of all types, at scale, based solely on the visual and text features of a product, and then translate those data into actionable insights.

Retailers don’t have to expend time effort, and investment into data collection and training their own AI models. Leveraging Google’s database of billions of unique entities, Google Cloud’s shelf checking AI can identify products from various image types taken at different angles and vantage points. Retailers will have flexibility in the types of imagery they can supply to the shelf checking AI. For example, a retailer can use imagery from a ceiling-mounted camera, an associate’s mobile phone or a store-roaming robot on shelf-checking duty.

Now in preview, this technology is expected to be generally available to retailers globally in the coming months. Importantly, a retailer’s imagery and data remain their own and the AI can only be used for the identification of products and price tags.

To help retailers make the online browsing and product discovery experience more modern, faster, intuitive and fulfilling for shoppers, Google Cloud introduced an AI-powered browse feature in its Discovery AI for retailers. The capability uses machine learning to select the optimal ordering of products on a retailer’s ecommerce site once shoppers choose a category, such as women’s jackets or kitchenware.

Over time, the AI learns the ideal product ordering for each page on an ecommerce site using historical data, optimising how and what products are shown for accuracy, relevance and likelihood of making a sale. The feature can be used on various ecommerce site pages, from browse, brand and landing pages, to navigation and collection pages.

Historically, ecommerce sites have sorted product results based on either category bestseller lists or human-written rules, such as manually determining what clothing to highlight based on seasonality. This browse technology takes a different approach, self-curating, learning from experience and requiring no manual intervention. In addition to driving improvements in revenue per visit, it can also save retailers the time and expense of manually curating multiple ecommerce pages. The tool is generally available to retailers worldwide supporting 72 languages.

Research commissioned by Google Cloud found that three-quarters of shoppers prefer brands that personalise interactions and outreach to them, and 86% want a brand that understands their interests and preferences.

To help retailers create more fluid and intuitive online shopping experiences, Google Cloud has introduced an AI-driven personalisation capability that customises the results a customer gets when they search and browse a retailer’s web site. The technology turbo-charges the capabilities of Google Cloud’s browse offering and existing retail search.

The AI underpinning the personalisation capability is a product-pattern recogniser that uses a customer’s behaviour on an ecommerce site, such as their clicks, cart, purchases and other information, to determine shopper taste and preferences. The AI then moves products that match those preferences up in search and browse rankings for a personalised result. A shopper’s personalised search and browse results are based solely on their interactions on that specific retailer’s ecommerce site, and are not linked to their Google account activity. The shopper is identified either through an account they have created with the retailer’s site, or by a first-party cookie on the web site.

Product recommendation systems are a critical component of any retailer’s ecommerce strategy for good reason: online retail sales are expected to reach more than $8tn by 2026. However, retailers have long had difficulty determining which panels to display on their web sites, how to arrange them, and how to coordinate content that is relevant and personalised. Google Cloud’s Recommendations AI uses machine learning to help retailers bring product recommendations to their shoppers.

Upgrades to Recommendations AI can make a retailer’s ecommerce properties more personalised, dynamic and helpful for individual customers. For example, a page-level optimisation feature enables an ecommerce site to decide dynamically what product recommendation panels to show a shopper. Page-level optimisation also reduces the need for resource intensive user experience testing, and can improve user engagement and conversion rates.

In addition, a recently added revenue optimisation feature uses machine learning to offer better product recommendations that can lift revenue per user session on any ecommerce site. A machine-learning model, built in collaboration with DeepMind, combines an ecommerce site’s product categories, item prices, and customer clicks and conversions to find the right balance between long-term satisfaction for shoppers and revenue lift for retailers. Finally, a buy-it-again model leverages a customer’s shopping history to provide personalised recommendations for potential repeat purchases.

Compared with baseline recommendation systems used by Google Cloud customers, Recommendations AI has shown double digit uplift in conversion and clickthrough rates in experiments controlled by retailers using the technology. The page-level optimisation, revenue optimisation and buy-it-again models are globally available to retailers.

Google Cloud’s shelf-checking AI tool is in preview globally. The ecommerce technologies, including the personalisation AI capability, browse feature and updates to Recommendations AI, are all globally available to retailers.