How can the user experience affect the business?

Zappos creates breakthrough customer experiences with AWS

Looking for the perfect fit

Zappos knows that accurate recommendations are key to an efficient shopping experience. The company empowers its customers with a generous return policy and fast, free shipping, but this expensive service is industry standard today.

"We always ask ourselves: How can we stand out further?" Says Kazerouni. "How do we optimize the response rate without negatively impacting the customer experience? We want to solve these kinds of challenges with the help of machine learning and analysis on AWS."

For the search phase of the customer experience, the company wanted to ensure that personalized recommendations appear during the search in order to increase search relevance. Instead of using a generic search algorithm, Zappos tries to understand customers personally and deliver unique search results for a specific term. (For customers who do not want this degree of personalization, an opt-out button is also clearly visible).

At the same time, it cannot afford to noticeably slow down search performance. "We had to keep the time required for the additional processes as short as possible," notes Kazerouni. "So we combine high-performance caching, strategic pre-calculation of certain results and holistic machine learning approaches that use several simple models."

More than the sum of its parts

The data pipeline begins with a lightweight client that sends relevant events to an ingestion API for processing. The API is in an auto-scaling group to handle large amounts of data. The API sends the data to Amazon Kinesis Data Firehose for inclusion in an Amazon Redshift data warehouse, which provides high-performance data access for machine learning research. Amazon Simple Storage Service (Amazon S3) acts as an intermediary between Amazon Kinesis Data Firehose and Amazon Redshift.

Zappos uses various technologies for training and model operation. It relies on Amazon SageMaker to predict customers' clothing sizes. These predictions are cached and then made available at runtime through microservice APIs for use in recommendations. Zappos leverages Amazon EMR to perform big data analytics at a fraction of the cost of traditional on-premises clusters. It also runs models with Graphical Processing Units (GPUs) in the Amazon Elastic Compute Cloud (Amazon EC2).

The company enables the ultra-fast search for pre-calculated forecasts with the help of two different services. Amazon DynamoDB stores precalculated results that are accessed at runtime. This fully managed key-value and document database provides single-digit millisecond performance on almost any scale. It can process more than 10 trillion requests per day and support peaks of more than 20 million requests per second. For even faster response times, Zappos uses Amazon ElastiCache for Redis, an in-memory data store, as the cache level. This service guarantees a latency of less than one millisecond if required.

The microservices that run models and consolidate results run on Amazon EC2 instances arranged in auto-scaling groups with location-based load balancers. Zappos uses Amazon Route 53 as its domain name system and routes traffic through the entire solution.

Don't go - run!

Creating and maintaining this complicated architecture using traditional development and deployment methods would be immensely complex. Instead, Zappos relies on Infrastructure-as-a-Code with AWS CloudFormation. "Every aspect of the solution is in the AWS CloudFormation templates," reports Kazerouni. "All we have to do to make a change is tweak the template. If we need to correct the way the services communicate with Redis, we don't repeat the change manually - we change the template and deploy it everywhere."

Kazerouni said it would be impossible to develop the solution without the plethora of AWS services available to the team. "Using AWS services as building blocks allows engineers to focus on improving performance and results rather than dealing with the overhead of DevOps."

Customers feel the love

Zappos delivers these improved search results to its customers with an almost undetectable increase in latency, with 99 percent of search queries being completed in less than 48 milliseconds. With the help of a similar architecture, the personalized size recommendations, which are based on simple surveys on fit and previous purchases, have also been significantly improved. As a result, the company has reduced repetitive searches and product returns. It has also achieved higher click-through rates from search to product and increased the position of customer selection in search results.

Kazerouni sums it up: "We see ourselves as a customer service company that happens to be selling shoes and clothing. Anything we can do to optimize service improves our business. Using AWS allows us to innovate the customer experience faster . "


About Zappos

Zappos started out as a small online shoe retailer 20 years ago. Since then, the company has grown to include clothing, handbags, accessories, and more. At the same time, it offers renowned customer service and innovative employee experiences. The company has been an Amazon subsidiary since 2009.

Benefits of AWS

  • Keeps search latency below 48 milliseconds in 99% of searches
  • Personalizes the search for a better customer experience
  • Achieves higher click rates from search to product
  • Get fewer returns due to improved size recommendations

AWS services used

Amazon EMR

It enables Apache Spark, Hadoop, HBase, Presto, Hive, and other big data frameworks to run and scale easily.

Additional Information "

Amazon Kinesis Data Firehose

Amazon Kinesis Data Firehose offers the easiest method to reliably load streaming data into data lakes, data stores and analysis tools. It can record streaming data, convert it and load it into Amazon S3, Amazon Redshift, Amazon Elasticsearch Service and Splunk, so that analysis with the business intelligence tools and dashboards you use is possible in almost real time.

Additional Information "

Amazon SageMaker

Amazon SageMaker is a fully managed service that enables any developer and data scientist to quickly build, train, and deploy machine learning (ML) models. SageMaker eliminates the toughest tasks in every step of the machine learning process to make building high-quality models easier.

Additional Information "

Amazon Redshift

Redshift supports analytical workloads for all types of organizations, from startups to Fortune 500 companies. Companies like Lyft have gone from start-up to multi-billion dollar company with Redshift.

Additional Information "

First steps

Companies of all sizes and in every industry are transforming their business with AWS. Contact our experts and embark on your own AWS cloud journey today.