Google Cloud Recommendations AI, or Recommendations from Vertex AI Search, is a managed service that helps ecommerce websites personalize product recommendations for customers. It uses AI and machine learning to analyze user behavior, then recommends items that are most likely to be of interest to them.
Types of Models
Google Cloud's Retail API includes several pre-built AI models. Some models will require ingesting user event data for 60 to 90 days before they can be trained. Each model should be placed on its intended page type for best results.
Recently Viewed Items
The built-in, default model that shows products that users recently viewed. This model reminds customers of what they were interested in, saving them time and effort from having to search for those items again. This is especially helpful if they were interrupted during their browsing session or if they were considering multiple options.
Recommended for you
Predicts products that users will most likely engage with or purchase based on their shopping or viewing history. Recommended products act as a helpful nudge, highlighting options that fit the customer's needs and preferences, simplifying the decision-making process. Providing recommendations that feel like they were handpicked fosters a sense of connection and value and shows the customer that your store understands their taste and is invested in providing a tailored experience.
Others you may like
Predicts products that users will most likely engage with or purchase based on their shopping or viewing history and its relevance to specific products. Sometimes customers get stuck in their own search bubbles, repeatedly encountering the same products. These recommendations break this cycle, introducing them to hidden gems, complementary items, or unexpected delights they might not have stumbled upon otherwise.
Frequently bought together
Predicts product that are frequently bought together based on the product pages visited during the same shopping session. By showcasing items commonly purchased alongside the product the customer is viewing, it eliminates the need to search for these accessories or complementary items separately. This is especially helpful if they are unfamiliar with the product or its ecosystem.
Predicts products that have the mostly similar attributes to the products on the page. The customer may love the general idea of the product they're viewing, but perhaps something about it, like the color, size, or material, isn't quite right. Similar items show alternative options that address their specific preferences, leading to a potential purchase.
Buy it again
Predicts products for a user to purchase again based on their purchase history. Showing a customer their past purchases readily available creates a sense of recognition and personalization. It shows the customer that the store remembers their preferences and caters to their individual needs.
Recommends products that are on sale based on their current and original prices. The most common reason customers use this model is to find products at a discounted price. Saving money is always a motivator, and "On Sale" sections highlight deals that can lead to significant savings compared to regular prices.
The Page-level Optimization AI model can be applied to many page types and automatically optimizes the entire page with multiple recommendation panels.
Model Training Frequency
Training a model unnecessarily can create additional Google Cloud costs. We recommend using the default settings for each model and pausing model training if your storefront traffic becomes stagnant.
Some AI model types allow you to select different business objectives and will have different training data requirements. The business objectives are typically one of:
The model will recommend products that optimize for the highest number of clicks on the recommended products.
The model will recommend products that are most likely to complete a purchase on your website.
Revenue per Session
The model will recommend products that are most likely to increase the amount of revenue generated from each purchase.