Sandra Calvo, customer engineer at Google Finland, and Mark Bryan, principal architect at Google Finland, talk about how they use and store data, and break it down to make it usable and understandable for all.
The science behind personalisation
"Vector embeddings will be the hottest thing in the next 12-18 months". Mark Byran describes vector embeddings as a "way to represent almost any kind of data—text, videos, music, whatever—as points in a space. We're trying to map data by seeing what's closely related to each other and find ways to understand the patterns intuitively".
Mark gives the example of a customer at a bookshop. What might they be interested in reading next? To find out, you can use vector embeddings to map their preferences, such as science fiction and specific books they have purchased in the past, and map books that are similar to their preferences to find clusters of related things. You can make accurate recommendations by overlaying the map of the consumers with maps of similar books. "This is the science behind how you get recommendations on Netflix, Spotify or YouTube," says Mark.
The biggest thing in commerce right now is tailoring experiences to your customers. Vector embeddings can help you do it. Personalisation is the best way to create customer loyalty, whether through a paid media ad, specific promotion, or content piece—because you're adding value to them specifically.
How does AI work?
AI works by using "vector search", says Mark. It finds similar embeddings in the space by calculating the distance or similarity between vectors.
Researchers have developed an Approximate Nearest Neighbour (ANN) technique to make it faster. It uses "vector quantisation" to separate the space into multiple spaces within the tree structure. You can do a vector search, such as the Approximate Nearest Neighbour search, and look for clusters of patterns in the data. "This is how generative AI is actually working behind the scenes", says Mark. "It's looking for similar clusters of relevant data points and data sets in an extremely fast way". Nearest Neighbour search helps you give recommendations, create target promotions or ads, power semantic text searches, answer questions and more.
Mark says, "Vector embeddings power everything you use today and can transform how you engage with your customers".
Generative AI has triggered a massive acceleration in the understanding of vector embeddings. Mark believes "if you don't already use vector databases, you probably will in the next 6-12 months".
Why you need a data economy
"We have way too much data", says Mark. We don't have the time to read it, understand it and make it useful. One of the biggest problems is that data sits in silos, whether that's data silos or business silos. How do you fix that? Mark believes "you need a data economy". It's about adopting principles and practices to ensure data can be published, discovered, built and relied on.
Mark gives the example of a traditional business setup organised by department or group. "Your organisation will have business cases, use cases, data tools, and operations functions". There's a lot of data that's organised by silos—whether that's group silos, data location silos, tool/format silos, processing silos or more. Mark says that means you'll have a lot of repeating work. How can you create more efficient teams and ensure you have good data on which to base decisions? "Remove all the noise and focus on what's unique: the business case, data pipelines and data", says Mark.
Make it accessible so your other business units rely on your data products to make decisions. "If you're not tailoring your products and solving for your customers, then digitalisation, modernisation and transformation are not strategies; they're outcomes", says Mark.
Discovery AI helps retailers solve product discovery challenges
Sandra Calvo, customer engineer at Google Finland, explains the small steps you can take to start solving for your customer. "At Google, we're on a mission to make things easier. You don't need to be a researcher or coder to do things with data". Sandra speaks about three tools to help you use data to meet customers' needs. Discovery AI helps retailers solve product discovery challenges with a "multi-modal" approach that focuses on:
Retail Search and Browse powers your digital commerce site or app with search and browse capabilities.
Recommendations AI delivers highly personalised content recommendations at scale.
Vision Product Search lets customers search for products with an image and receive a ranked list of visually and semantically similar items.
"All together, these create a very powerful combination—and you can start using it today. It uses simple data—your product catalogue or user events data—and looks at the things in your portfolio and how people move around on your website", says Sandra.
Instead of building a data model from scratch, you can use Discovery AI and the data from Tag Manager and Google Analytics to build a recommendations model. Whether you want to increase revenue or improve conversions and click-through, Sandra says you can easily select and train the model and implement it on your platform. You don't need extensive technical skills.
Generative AI helps retailers improve customer experience
Generative AI can help you create text, audio, video or code, and you can use it for everything from customer service to marketing campaigns. Sandra gives the example of a marketing campaign that needs to be adapted to specific audiences, countries and languages. "With Generative AI Studio, you can test everything without having to code". You can feed Generative AI Studio a command to create an image of something you want to sell, and then finetune it. You can also use real imagery, such as your product, and change the colours and background—adapting the campaign to your target audience.
Search and Conversation AI revolutionalises commerce
There's a lot of information out there, and finding the answers you need can take time. You can help employees and customers find information quickly by combining enterprise data with Google Search and conversational AI. Putting all that data in cloud storage and connecting Search can make information more accessible. With it, you can ask questions in natural language and get generated answers through a Large Language Model (LLM). You can build custom chatbots and voice bots and enjoy enterprise-grade scalability, data privacy, security and control.
How do global leaders use data and AI to boost ecommerce?
Sandra names a few leaders in ecommerce that have successfully used data and AI to boost performance:
Macy's increased conversions by 2% by using data to improve how shoppers find its products.
Ulta Beauty grew its loyalty membership by 8% by using data and AI to create more personalised experiences for loyal customers.
Ikea made product recommend products better thanks to Goodgle's Recommendations Engine.
Carrefour created new marketing campaigns in just one click by training AI to learn from past campaigns and improve them.
Ready to start your journey to composable commerce?
Composable commerce makes it easier to harness emerging tools and use cases. By breaking down complex architectures into smaller composable modules, composable commerce helps you curate a selection of APIs and microservices that can be integrated and swapped without the need for expensive developers.
Ready to rethink your ecommerce? Get in touch to book a discovery process and start your composable commerce journey today.