Leveraging Technology and Automation to Keep Our Customers Safe
At Amazon, we strive to be Earth’s most customer-centric company, where customers can find and discover anything they might want to buy online. As we grow our selection and business to delight our customers, we also encounter complexity as we operate as a brand owner, a retailer, or as a supplier. We solve this complexity by constantly innovating on behalf of our customers to better serve and protect them, and we have implemented technological solutions within food safety for relatively straightforward tasks like slicing deli meat, as well as complex tasks like ingesting customer feedback and executing food recalls. We look at a high-risk process or event and ask how we can eliminate the risk while driving innovation that reduces the chance for human error.
Listening To Our Customers at Scale
We invest heavily into mechanisms to listen to our customers and detect when something has gone wrong. We make data-driven decisions but also respond rapidly to customer anecdotes. On the rare occasion when data and a customer anecdote disagree, we work on a solution with the tenet that our customer is right. Through automation, we aggregate 30 million pieces of customer feedback a week globally in over 40 languages. These interactions include customer contacts or feedback data such as product reviews, customer return comments, Customer Service chat, social media account, etc. We have leveraged automation to handle both the scale and the data extraction challenge. We rely on Machine Learning (ML) and complex software- based logic to understand context across languages and unlock the meaning of customer comments without the need for users to review them one by one.
This first pass of ‘labeling’ customer interactions is critical to separate the true signal from the large amount of customer interaction data. We then back it up through further automation in the form of ML to determine the feedback’s relation to a food safety concern. When the ML judgment has high confidence, we automate the action, otherwise, we rely on human Subject Matter Expert (SME) review. Our SMEs have accurately assessed safety risk hundreds of thousands of times over the last several years, and all outcomes of their judgment is then utilized to further enhance our ML automation efforts.
Instituting Technology to Ensure Food Processing Standards
Slicing deli meat is a relatively straightforward task, however, the physical handling of ready to eat food is one of the highest risk process control points from a food safety perspective. We implemented a physical process control by leveraging software logic in our deli slicing operation to ensure our customers receive safe, correct product, at the correct time. Our software helps validate the receipt temperature, monitor the storage temperature with real-time alarming, and calculate a shrink date. When our customer orders a specific amount of sliced meat, we run the product on hand through an SKU transformation process to come with the same details for the new sliced meat customer order. We then leverage our software to generate the critical product label, where we display information like the item name, ingredients, and allergens. This logic has helped us avoid human error and gain our customer’s trust in food safety standards.
Orchestrating Complex and Rapid Product Recalls
Amazon handles thousands of recalls every year for food and non-food products. Through technological investment, we have built robust mechanisms to execute recalls across all our businesses. We developed technology to automatically identify a recall once it is available in public domain, and supplement it through competent manual reviews to ensure we don’t miss recall alerts. Our complex catalog presents us with product identification challenges, and we have solved this by utilizing syntax matching, image analysis, and query logic to identify all related products across Amazon’s various channels. Once we identify the product we orchestrate recalls through our recall automation tool. The tool initiates multiple parallel efforts to prevent customer orders, remove the product from sale, rapidly isolate the product in our supply chain, and alert customers with a personal email in a few hours.
Investigation and Traceability for Agile and Focused Solutions
We have invested and continue to invest in technology that allows us to track and investigate the virtual details of products or issues. We built a comprehensive investigation system supported by decision logic that adds tailwinds to our investigators and allows for a rapid scale to deliver safety solutions to our upcoming businesses. With this capability, we actively tag and track issues from the point of detection all the way to initiating suitable remedial action. As we have multiple business channels and operate globally, we have enabled our investigation technology to initiate checks to look across business channels and global regions for similar issues and act proactively.
Our customers continue to teach how to better serve them, and we will continue to invest heavily in automation and technology to delight them every day while ensuring their safety.