The Role of Machine Learning and AI in Amazon’s Data Monetization Approach

A brief overview of Amazon as a company and its dominant position in the e-commerce industry

Founded by Jeff Bezos in 1994, Amazon is a multinational technology company that started as an online marketplace for books and has since expanded to become one of the world’s largest and most influential companies. It is headquartered in Seattle, Washington, and offers customers a vast array of products and services worldwide through various online marketplaces in different countries, including the United States, the United Kingdom, Germany, France, and more.

What sets Amazon apart from other e-commerce platforms is its commitment to customer experience. It has revolutionized the concept of online shopping with features like one-click ordering, personalized recommendations, fast and reliable shipping through services like Amazon Prime, and customer reviews that help buyers make informed decisions. Amazon’s dominance in the e-commerce industry is undeniable, with a massive customer base and a vast network of sellers who use its platform to reach millions of potential buyers.

In addition to e-commerce, Amazon has diversified its business portfolio over the years, expanding into sectors such as cloud computing (Amazon Web Services), digital streaming (Prime Video), audiobooks (Audible), smart home devices (Echo, Alexa), and physical retail (Amazon Go, Whole Foods Market). This diversification allows Amazon to capture various revenue streams and leverage data monetization tactics.

Figure 1: Amazon development time analysis.

Amazon’s success is attributed to its relentless focus on innovation, operational efficiency, and long-term thinking. The company continually improves its services and expands into new markets by leveraging advanced technologies and data-driven insights. It invests heavily in data monetization, collection, and analysis, enabling it to understand customer preferences, optimize inventory management, and provide personalized experiences for more users. Furthermore, Amazon’s extensive data capabilities empower its advertising business, allowing brands to target customers precisely.

What do the new revenue streams by Amazon mean?

Based on the data, Amazon’s market share and current position in the e-commerce industry are remarkable. Amazon is one of the most valuable companies globally, with a market value exceeding $1.105 trillion in 2022. Its brand value of $350 billion places it second only to Apple, surpassing Google and Microsoft. Amazon’s dominance is evident in its user base, with over 300 million active users and approximately 197 million monthly visitors to Amazon.com . The company’s reach extends to more than 100 countries, excluding a few countries due to restrictions.

Amazon Prime, a subscription-based service, has over 200 million subscribers globally and over 157.4 million subscribers in the United States alone. Amazon Prime Video, the second-largest streaming service in the world, boasts over 205 million subscribers. This extensive customer base contributes to the significant revenue generated from membership services, amounting to approximately $25.21 billion yearly. In addition, Amazon’s annual Prime Day sales reached a staggering $11.19 billion in 2021.

Amazon’s market dominance is evident in multiple product categories, where it owns over 90% market share, such as batteries (97%), kitchen and dining (94%), home improvement tools (93%), golf (92%), and skin care (91%). Furthermore, the company holds a nearly 50% share of the US e-commerce market, surpassing its main competitors, including eBay, Apple, and Walmart.

The influence and success of Amazon are not limited to its direct operations. The platform hosts almost 2 million small and medium-sized businesses (SMBs) as third-party sellers, with approximately 56% of Amazon’s revenue in 2021 coming from these sellers. SMBs benefit from Amazon’s expansive customer base and services like Fulfillment by Amazon (FBA), which offers reduced shipping costs and logistical support.

Employment-wise, Amazon has seen significant growth, with 1.2 million employees in 2022, nearly double the number recorded in Q4 of 2019. The company hired 270,000 new employees in the second half of 2021 to meet the demand for its fast delivery services.

Finally, Amazon’s advertising revenue reached approximately $31 billion in 2021, reflecting the value of the company’s growing presence in the digital advertising space.

Importance of data in today’s digital economy

Figure 2: Intangible assets market value (Source: Ocean Tomo)

The Data Ecosystem at Amazon

Overview of the various sources of data generated by Amazon

Amazon gathers vast amounts of data from various sources, including its:

e-commerce platform,

customer transactions,

product details,

pricing and dynamic pricing,

inventory levels, and shipping information.

The company tracks customer browsing behavior, purchase history, and wish lists to provide personalized recommendations and targeted advertising. Customer reviews offer valuable insights into customer preferences, product satisfaction, and quality. Moreover, measurable business performance improvements and data monetization wouldn’t be possible without understanding customer behavior.

Alexa, the voice-controlled assistant, generates data through voice interactions, including voice commands, search queries, and user preferences. Advertising and marketing campaigns contribute data related to sponsored product ads, display ads, and affiliate marketing programs.

As a leading cloud services provider, Amazon gathers data from various sources within AWS, including usage patterns, resource utilization, performance metrics, and operational logs. The company’s extensive supply chain and logistics operations generate substantial data on inventory levels, warehouse operations, shipping routes, delivery performance, and fulfillment efficiency.

Amazon’s Kindle e-readers and e-books generate data on reading habits, book preferences, and user highlights. Third-party sellers listing products on Amazon’s platform generate data on product listings, pricing information, seller performance metrics, and customer feedback.

Additionally, Amazon collects data from its brick-and-mortar stores, such as Amazon Go and Whole Foods, including transaction data, customer behavior, and inventory management information.

To leverage this collected data, Amazon employs data monetization and embedded analytics. It utilizes the data to refine recommendations, optimize advertising strategies, improve service offerings within AWS, enhance supply chain operations, and provide seamless shopping experiences.

Furthermore, Amazon may directly share or sell data or perform in-depth analysis to extract valuable insights and drive decision-making across its business ecosystem.

Customer data: purchase history, browsing behavior, reviews, and dynamic pricing

For Amazon to make data-driven decisions, they must collect it first. All in all, Amazon collects around 1 exabyte of purchase history data from their consumer base. In perspective, 1 exabyte is about 10,000 miles short of reaching the moon (Source). They have a vast pool of customers to collect from, considering that the Amazon Web Store fields about 1.1 million requests a second. Still, purchasing information is one of many types of data they collect.

A BBC journalist found that Alexa had stored the transcriptions of all 31,082 interactions his family ever had with the device. This incident highlights the potential extent of data collection by Amazon’s Alexa voice assistant. In addition to the vast amounts of data generated by Amazon’s e-commerce platform, website interactions, advertising campaigns, and other sources, the storage of complex voice interactions further contributes to the overall data volume that Amazon collects.

Except for recommendations and campaigns, Amazon’s unrivaled expertise in dynamic pricing is evident as the marketplace giant adjusts its product prices an astonishing 2.5 million times daily.

You need a lot of data and measurable business performance improvements to do that.

Anchored on cutting-edge technologies like artificial intelligence (AI), machine learning (ML), and big data analytics, Amazon’s dynamic pricing strategy relies on a sophisticated repricing feature that employs advanced algorithms, data monetization, and understanding customer behavior. These algorithms evaluate and update prices for millions of products multiple times daily, considering various factors such as demand, stock availability, and customer behavior. By leveraging historical and real-time market data, the algorithms identify trends and make accurate predictions. Business and product teams summarize a lot of data, ensuring there is no better offer on the Internet.

While the specifics of Amazon’s pricing strategy are closely guarded, it can be inferred that the company considers specific parameters when implementing price changes. These parameters encompass both global values and user values.

Global values encompass factors like demand volume and stock volume, while user values include product visit frequency and time of purchase. Evaluating global and user values allows Amazon to optimize its pricing strategy and provide the most competitive prices.

Global values, such as market behavior, include demand and stock volumes. Amazon adjusts prices based on anticipated market demand for a product, considering whether the demand is seasonal and predictable and what motivates customer purchases.

If data stock units are running low and continued demand is expected, Amazon may increase prices if it does not hinder purchases from a particular consumer segment.

On the other hand, data user values, such as SKU visit ratio and day and time of purchase, help Amazon analyze consumer behavior. By tracking user behavior and collecting data from every online interaction, Amazon can determine the number of times customers have viewed a specific product, whether they viewed it at particular times, or if they tend to browse related products together.

This data enables Amazon to develop more appealing pricing strategies for potential buyers, resulting in increased sales figures and higher average transaction values. Similar to the tourism industry, certain products may be purchased more frequently on specific days of the week or times of the day. Amazon capitalizes on these moments when customers are more likely to browse, compare, and make purchase decisions by implementing price changes.

Figure 3: How Amazon enhances customer experience and delivers data monetization capability (Source Shiksha Online)

Seller data: sales performance, product information, inventory, etc.

It’s fascinating how Amazon has designed the whole supply chain to conduct data monetization at every step. We can learn a lot from their business users and analytics-enabled platform. First and foremost, existing data is fuel for measurable business performance improvements. To understand it, looking at the fulfillment process Amazon designed to meet its data monetization needs is fantastic.

With FBA (Fulfillment by Amazon), sellers can entrust Amazon with their storage, packaging, and shipping needs. The company handles all repetitive tasks. Sellers send their products to Amazon fulfillment centers, and the platform handles everything else, such as customer service, returns, and refunds.

Figure 4. How does Data Driven Fulfillment by Amazon work (Source: ProftWhales )

Amazon’s recently announced improved Fulfilled by Amazon (FBA) inventory capacity management system aims to reduce development time for sellers while effectively managing their inventory. By collecting more data and quantified feedback from sellers, Amazon teams have streamlined the system, offering a range of new data monetization methods.

Monthly Capacity Limit: Amazon has addressed sellers’ concerns by providing predictive planning tools for inventory procurement. The capacity limits for the upcoming month will be announced during the whole third week of each month via the Capacity Monitor in Seller Central, accompanied by email notifications. This feature helps sellers manage inventory costs based on practical data sharing facilitated by Amazon’s platform.

Estimated Capacity Limits for Long-Term Planning: In addition to the capacity limit for the upcoming month, Amazon also provides estimated limits for the subsequent two months. This estimate allows sellers to plan over a longer time horizon and opens up indirect and direct data monetization opportunities. Companies can leverage this data to run promotions and implement discount strategies more effectively, utilizing analytics and insights provided by Amazon.

Requesting Higher Capacity Limits: With tools like Capacity Manager, sellers can request additional capacity by paying a reservation fee. This feature is a deal improver for sellers as it unlocks cash and other resources. By utilizing predictive planning, sellers can save costs, streamline operations, and reduce waste. Sellers that are granted additional capacity can offset their reservation fees by earning performance credits tied to the sales generated using the extra capacity. This feature not only improves competitiveness but also provides sellers with a significant advantage over companies that don’t leverage similar analytic tools and data strategies.

Volume-Based FBA Capacity Limits: Capacity usage, capacity limits, and inventory usage will now be more accurately measured in cubic feet rather than the number of units. This shift to a more precise measurement system enables sellers to make better-informed decisions regarding their inventory management.

Amazon leverages its deep understanding of customer behavior through these direct data monetization techniques and enhances its sellers’ logistics chains and supply chain optimization. Amazon uses data as currency and may or may not charge for it, providing valuable insights and analytics to sellers on its platform. This data strategy gives Amazon a significant advantage over competitors like Target and Walmart.

Supply chain data: logistics, warehousing, shipping, etc.

Delivering parcels to the final destination can be costly due to poor routing and failed delivery attempts. Data can help manage this problem, but data monetization tools should be responsible for data strategy, data sharing, and process optimization.

Last mile costs can make up to 50% (according to aws.amazon.com ) of the overall fulfillment costs, which includes pickup, line haul, and sorting. The costs for the last mile can double if a delivery attempt fails and another shot is needed. Let’s find out how Amazon addresses these specific needs and brings measurable business performance improvements.

Amazon productizes everything, so the strategy is not different in this situation.

AWS has introduced a Last Mile Routing solution for its customers, incorporating machine learning for efficient last-mile delivery, data sharing, and monetization tools. Through reinforcement learning and neural graph networks, Amazon has analyzed real-life data, which includes package details, destination locations, customer preferences, delivery time windows, expected service times, and zone information. I call it a “data monetization comprehensive” play.

By leveraging the power of machine learning, AWS enables its customers to benefit from optimized route sequencing, real-time routing, and accurate delivery time windows. Additionally, AWS offers data-sharing capabilities, allowing customers to securely exchange relevant data with authorized stakeholders, such as logistics partners.

The solution is capable of real-time rerouting that handles constantly changing traffic conditions and customer time windows when drivers are already on the move. By developing two online policy improvement methods to satisfy time window constraints, Amazon extended state-of-the-art research on solving the traveling salesperson problem with policy gradient–based RL algorithms (Source: aws.amazon.com ).

The system learns to choose the best delivery routes based on historical data or decide the right route order within a set delivery schedule. This process could involve selecting the most efficient delivery path, sequencing multiple routes into a single delivery attempt, or planning multiple delivery attempts using an optimal combination of orders.

Last Mile Routing is, in my opinion, a data monetization tool set, and here are only a few ways it helps to bring new revenue streams and provide customers with even more value:

AWS’s Last Mile Routing solution enables customers to monetize their data by securely sharing relevant information with authorized stakeholders, such as logistics partners or local authorities.

Customers can optimize route sequencing and improve delivery efficiency through machine learning, leading to cost savings and increased customer satisfaction (an indirect method of data monetization).

By leveraging real-time routing and accurate delivery time windows, customers can offer premium delivery services to their end customers, potentially charging higher fees for expedited or time-sensitive deliveries. It’s also how companies can automate repetitive tasks that come into play and influence competitive advantage.

Analyzing historical data and making data-driven decisions on delivery routes allows customers to identify patterns and trends, which can be valuable insights for targeted marketing campaigns and customer manners analysis.

Customers can maximize their delivery capacity and revenue potential by planning multiple delivery attempts using an optimal combination of orders, resulting in increased profitability.

Data Collection and Processing

Insights into Amazon’s data collection methods

Amazon is known for its extensive data collection methods, which are crucial to its business model and customer experience. The company’s data monetization capability is almost endless as the data collection system is complex and comprehensive.

Figure 5: Data Monetization, Collection, and Utilization in Amazon’s Customer-Centric Ecosystem (Source: Arek Skuza research and insight)

Although Amazon’s data collection practices are not widely known, I have gathered some insights based on my research and personal experience as an Amazon customer.

Customer Interaction: Amazon collects data through various touchpoints, including its website, mobile applications, and devices like Kindle and Echo. This data includes customer preferences, search history, purchase history, and user-generated content like reviews and ratings.

Cookies and Tracking Technologies: Like many other online platforms, Amazon utilizes cookies and similar tracking technologies to gather information about users’ browsing behavior, session information, and interactions with the website. These technologies help personalize the shopping experience, provide relevant recommendations, and target advertising.

Amazon Web Services (AWS): As a provider of cloud services, Amazon collects data from companies and organizations that use AWS. This data could include website traffic, application usage, and performance metrics. However, it’s important to note that AWS data collection practices are separate from Amazon’s retail operations.

Alexa and Echo Devices: Amazon’s voice-controlled assistant, Alexa, powers Echo devices and collects data when users interact with these devices. Voice commands, inquiries, and other interactions are processed and stored to improve speech recognition, understand user preferences, and enhance the overall user experience.

Third-Party Sellers: Amazon hosts a vast marketplace with numerous third-party sellers. While Amazon may not directly control data collection by these sellers, it likely has access to aggregated sales data, inventory information, and performance metrics to optimize the marketplace and support seller services.

Advertising and Marketing: Amazon leverages customer data to deliver targeted advertising on its platform and through third-party websites and applications. Advertisers can use Amazon’s data to reach specific customer segments based on interests, browsing history, and purchase behavior.

Use of machine learning and AI algorithms to extract value from data

Amazon employs advanced machine learning algorithms and AI technologies to analyze the collected data. These technologies enable personalized recommendations, fraud detection, inventory management, supply chain optimization, and customer service improvements. I cannot describe all the possible ways the company uses to conduct data monetization, data sharing, and measurable business performance improvements, so I selected three cases that I found the most compelling.

Products, searches, and reviews as an indirect way of selling data

Fake review brokers utilize third-party platforms like social media and encrypted messaging services to buy, sell, and host fake reviews. These deceptive reviews can influence consumer purchasing decisions based on what appears to be genuine feedback from other shoppers.

Amazon has been leveraging AI and machine learning technologies to combat this issue for several years. The company employs sophisticated tools that analyze various factors, including the relationship between the reviewer and online accounts, sign-in activity, review history, and unusual behavior, to determine the likelihood that a review is fake. Amazon uses these advanced techniques to prevent customers from encountering fake reviews altogether.

Amazon’s efforts have shown some success, but fake reviews persist. Estimates suggest that approximately one in seven online consumer reviews in the UK are fake. Fake reviews significantly increase the chances of consumers selecting poor-quality products. Amazon has blocked over 200 million suspected counterfeit checks in the past year and intends to continue developing more sophisticated tools to protect its customers. Machine learning plays an essential role in the operation.

A different approach to data monetization is AWS Inferentia. AWS Inferentia accelerators are advanced tools and embedded analytics developed by AWS to enhance business operations by boosting performance, reducing costs, and unlocking new opportunities for data monetization. These accelerators enable companies to process and analyze vast amounts of data efficiently, leading to valuable insights and potential revenue streams. Additionally, they facilitate seamless data sharing and collaboration, fostering innovation and partnership opportunities.

By leveraging Inferentia accelerators, companies can achieve significant cost savings. Amazon saves up to 85% in infrastructure costs while maintaining robust throughput and latency performance (especially in search and large e-commerce onsite operations).

This substantial expense reduction allows businesses to allocate resources more effectively, invest in growth initiatives, and drive profitability (a fantastic variety of data monetization methods). Furthermore, by efficiently processing data using Inferentia accelerators, organizations can uncover valuable patterns, trends, and correlations that can be leveraged to create new revenue streams and business models.

Moreover, the Inferentia accelerators empower businesses to handle complex models and perform sophisticated tasks at scale, providing a solid foundation for data monetization. They enable companies to leverage their data assets, generate actionable insights, and develop innovative products and services. With the ability to process data rapidly and accurately, organizations can identify market trends, customer preferences, and emerging opportunities, enabling them to make data-driven decisions and capture new market segments.

In addition to data monetization, the Inferentia accelerators promote data sharing and collaboration within and across organizations. Businesses can securely share insights and collaborate with partners, customers, and stakeholders by efficiently processing and analyzing data. This analysis opens up avenues for new business partnerships, joint ventures, and data-driven collaborations that can drive innovation and fuel growth. I find these data monetization tools and opportunities exciting.

Video quality monetizes data too

Amazon Prime Video utilizes machine learning techniques to ensure video quality and enhance the customer viewing experience. Prime Video’s Video Quality Analysis (VQA) group has been using machine learning for the past three years to identify defects in captured content and validate new application releases or offline changes to encoding profiles. Computer vision models are trained to detect block corruption, audio artifacts, and audio-video synchronization problems. Prime Video can process hundreds of thousands of video streams by analyzing video at a large scale, including live events and catalog content.

Amazon developed a dataset to replicate defects in high-quality content to address the low occurrence of audiovisual defects in Prime Video. Machine learning models are trained using this dataset to develop detectors for different types of defects. For example, block corruption is detected using a residual neural network, audio artifacts are identified using a no-reference model based on a pre-trained audio neural network, and lip sync defects are detected using the LipSync pipeline based on the SyncNet architecture (read more at Amazon ).

I summarized all the efforts to explain briefly (the original source describes more details) how Prime Video teams maintain high video quality standards and provide customers with the best viewing experience.

Goods inspection increases revenues, but something else is more important.

Amazon uses artificial intelligence to check for damaged items in its warehouses before shipping them to customers. The company expects this technology to reduce the number of damaged goods and speed up the process of preparing orders for delivery. Amazon warehouse workers manually check items for damage, which can be time-consuming and mentally demanding. Since Amazon owns data from the whole supply chain, it can implement another element of data monetization strategy – waste reduction. AI technology and analytics tools take over this task and play a crucial role in automating more of Amazon’s fulfillment operations.

Amazon estimates that less than 0.1% of the items it handles are damaged, but due to the large volume of packages (around 8 billion annually), this still amounts to a significant number. By implementing AI, checking the internal quality process, and making data sharing the primary strategy, Amazon aims to improve the customer experience by minimizing the number of damaged goods that are shipped out.

AI in logistics operations is becoming increasingly common as companies seek ways to streamline workflows and manage complex supply chains more efficiently. Automating warehouses involves:

Developing technology to perform tasks that are typically simple for humans, such as picking,

Packing,

Checking items for damage.

Amazon has already introduced AI technology at two fulfillment centers and plans to expand to 10 more business sites in North America and Europe. The AI system checks items during the picking and packing process by comparing them to photos of undamaged items. If any damage is detected, the item is flagged for further inspection by a worker, and the internal process in other applications is triggered. The order will be packed and shipped to the business or consumer if everything looks fine.

Overall, Amazon’s use of AI in its warehouses aims to reduce the number of damaged goods, improve efficiency, and enhance the customer experience. Again, all is possible due to solid analytics, measurable business performance improvements, and well-planned data monetization actions.

Conclusion

Amazon’s remarkable success in the e-commerce industry can be attributed to its unwavering commitment to innovation, relentless pursuit of operational efficiency, and steadfast dedication to customer satisfaction.

By harnessing the power of internal data, Amazon consistently optimizes user experiences and pricing strategies and provides valuable insights to its vast network of sellers. Furthermore, Amazon’s strategic diversification across multiple business sectors and robust internal capabilities allow the enterprise to maintain a dominant market position.

Through leveraging data, Amazon stays ahead of the competition, expands into new areas such as advertising and logistics, and continually drives its growth. As the digital economy evolves, businesses worldwide must recognize the importance of effectively utilizing internal data to foster innovation and propel sustainable growth, just as Amazon’s phenomenal success story exemplifies.

Recap of Amazon’s data monetization strategies

Amazon harnesses data across its business to boost revenue, cut expenses, and enhance customer experiences. The company employs various data monetization techniques that revolve around leveraging the immense volume of data it collects from its e-commerce platform, customer transactions, seller performance, supply chain operations, and more.

By effectively analyzing this data, Amazon makes data-driven decisions and provides personalized recommendations, optimizes pricing through dynamic pricing strategies, offers valuable insights to sellers, and enhances logistics and delivery processes. Internal and external stakeholders can easily consider the enterprise a 100% data-driven business. There is no doubt.

One of Amazon’s key data monetization approaches is personalized recommendations and targeted advertising. Amazon can offer highly tailored product recommendations by carefully understanding customer behavior, purchase history, and wish lists, further improving the shopping experience. Moreover, the company leverages its profound understanding of customer manners to optimize its advertising strategies, enabling brands to target their desired customers precisely. This data-driven approach enhances customer satisfaction and generates substantial advertising revenue for Amazon.

Another crucial aspect of Amazon’s data monetization strategies lies in its dynamic pricing capabilities. With the assistance of advanced algorithms and data analysis, Amazon adjusts product prices in real-time, considering factors like demand, stock availability, and customer behavior. This adjustment allows the company to offer competitive prices while maximizing revenue.

Additionally, Amazon’s Fulfillment by Amazon (FBA) program provides sellers with invaluable data and insights, aiding them in optimizing inventory management, pricing strategies, and overall business performance. By offering data-driven solutions and analytics tools, Amazon empowers sellers to make more informed decisions, enhancing their competitiveness and driving revenue growth.

In summary, Amazon’s data monetization strategies encompass personalized recommendations, targeted advertising, dynamic pricing, and data-driven solutions for sellers. By effectively utilizing the vast data it collects, Amazon has enhanced customer experiences, driven revenue growth, and maintained its dominant position in the e-commerce industry.

The article was first published at www.arekskuza.com