Google has made its Recommendations AI tool available for e-commerce. The tool uses Google’s AI tools to send personalized recommendations to customers of retailers.
Customers can use the Recommendations AI tool to create different models to make recommendations. The tool uses data from various Google Cloud services such as Google Analytics 360, BigQuery and Merchant Center allowing retailers to focus on specific customers. New models are trained within two to five days, depending on the complexity, after which they can be used to send recommendations.
According to Pallav Mehta, Google Product Manager, retailers can now focus more on individual customers instead of individual products. Recommendations AI uses the purchase history and browsing habits to find products that are of interest to the customer.
The service uses deep learning models that analyse the metadata of articles and users in order to gain insights about millions of articles on a large scale. “Recommendations AI is also capable of correcting for bias with extremely popular or on-sale items, and can better handle seasonality or items with sparse data.” According to Google, these models can be re-trained every day to gain new insights.
Recommendations AI was published last year in private beta. Users of that version reported a significant increase in sales due to the tool. Sephora, a retail chain specialising in perfume and cosmetics, saw a 50 percent increase in CTR (click-through rate) and the conversion rate (how many visitors actually buy something) increased by 2 percent. Other companies see similar results.
Google now makes the Recommendations AI tool available (public beta) at three price tiers, plus separate costs for training and tuning the deep learning models.