If you've done a search on Amazon, you've used the
A9 Product Search engine.
Our work starts long before a customer types a query. We've been analyzing data, observing past traffic patterns, and indexing the text describing every product in our catalog before the customer has even decided to search. And as soon as we see the first keystroke, we're ready with instant suggestions and a comprehensive set of search results.
Understanding the Customer
Great search can seem to customers like it is reading their minds. We start the search experience by giving customers suggestions on how to formulate their queries as soon as they start typing.
The better we understand the meaning of a query, the better we can help customers find the products they want. So we focus on the words and the intent behind those words. When a customer tells us they are looking for "Harry Potter in books", we distinguish in their query the title: "Harry Potter" from the category information: "in books".
Customers rely on us to find the right product based on context, even if a query is misspelled: if someone searches for Timbaland in Music, we return results for the rap musician; but the same search in Shoes will be corrected to return Timberland boots and jackets.
Results and Ranking
One of A9's tenets is that relevance is in the eye of the customer and we strive to get the best results for our users. Once we determine which items are good matches to the customer’s query, our ranking algorithms score them to present the most relevant results to the user.
Our ranking algorithms automatically learn to combine multiple relevance features. Our catalog’s structured data provides us with many such relevance features and we learn from past search patterns and adapt to what is important to our customers.
We strive for continuous improvement of our ranking algorithms. We continuously evaluate them using human judgments, programmatic analysis, key business metrics and performance metrics.
Search. At Scale.
We index millions of documents worldwide, and deploy them on highly scalable massive fleets of servers. We reflect millions of price and inventory changes in real time, and we return relevant results in milliseconds.
Our core Search Infrastructure team runs a critical service that is powerful, flexible and efficient. Since both query traffic and the size of our product catalog are always growing, we need to continually innovate to prevent geometric growth in the number of servers in our fleet. We ensure lightning fast experiences for customers, regardless of traffic patterns at a given time.
Our services are deployed internationally, powering Amazon sites and applications in America, Europe and Asia. We apply principles and solutions that cross boundaries: for locales where those principles don't enhance the customer experience, we adapt them to be effective in a given culture and language.