How do you make product recommendations when you have 150 million products and millions of customers? If you stick to best selling items you’ll never surface the rarer products which might be of most interest. This is a problem eBay has struggled with for years, hence our enthusiasm for the new eBay Feed coming to the UK next year.
Alibris, who are owned by Monsoon Commerce had exactly this problem. Selling new and used books and media they’ve got 150 million products which each individually rarely sell to their millions of customers. Helping customers find relevant products with a more personalized shopping experience is both challenging and essential according to Casey Carey, vice president of marketing for Monsoon Commerce and general manager for Alibris Marketplace Services.
Alibris’ data was vast, but sparse and in this situation, conventional machine learning models don’t work because a training subset can’t cover all the products. That’s where Simularity comes in. Alibris is now using Simularity technology to bring shoppers the joy of discovering products that are destined to become their new favorites.
Simularity’s technology uses highly scalable correlation algorithms to provide real-time similarity analytics over trillions of data points on a surprisingly inexpensive amount of commodity hardware. This is now enabling Alibris to power recommendations across their millions of products and make sense of their vast but sparse data.
There’s no real surprise that Alibris are using Simularity, Liz Derr their Founder and CEO was VP of Engineering and COO at Monsoon Commerce for seven years until she left in 2011 to set up the new company.
Alibris now have better and faster recommendations, powered by all our data, on commodity hardware which is good news for both buyers and sellers.