From hidden gems to bestsellers: How machine learning is reshaping the online retail market

Traditionally, retailers have managed consumer models, inventory purchases, storage and fulfillment. This approach involves significant risks as retailers need to forecast demand and manage logistics efficiently and accurately. In contrast, the marketplace model allows retailers to offer a much wider range of products without assuming inventory risk. By enabling third-party sellers to list their products, retailers can expand their product assortment, diversify revenue streams, and increase customer engagement.

This shift created a complex, interconnected market ecosystem. Walmart.com currently operates as a hybrid marketplace where Walmart stocks some of its products, with other products coming from third-party sellers adding even more complexity to the market situation. The vast amount of products on these platforms creates important challenges. It's a way to express the right product to the right customers.

The market needs to solve the problem of connecting buyers, if not millions, of available listings, with the most relevant products. ML-driven algorithms analyze huge amounts of customer data to determine which products to display, optimizing relevance and conversion rates.

These algorithms rely on customer signals such as browsing history, purchase behavior, trending products, price sensitivity, time of day, device type, weather and other contextual factors. By processing these inputs, the ML model can create data-driven recommendations that improve both customer satisfaction and seller success.

For example, ML allows customers to identify emerging trends and surface new products that are likely to be interested in but not explicitly searched. This allows buyers to access curated selections of items that suit their tastes, but sellers can benefit from increased visibility and sales opportunities.

One of the key challenges in recommending ML-driven products is the “cold start issue.” There is very little historical data to inform you of recommendations when a new seller or product enters the market, but personalization is more important than ever for this day and age. According to a survey by McKinsey, 71% of consumers expect businesses to provide personalized interactions, and 76% will be frustrated if this doesn't happen. Without sufficient behavioral insight, it is difficult for market algorithms to determine which customers are most interested in these new products, and ultimately creates a hurdle for personalization.

Advertising plays an important role in overcoming this challenge. Sellers can invest in ML-driven sponsored lists through retail media networks (an advertising ecosystem within retailers' digital storefronts) to increase product visibility. By bidding on ad placement, new suppliers can generate initial engagement, which supplies organic recommended algorithms to the data. As customers begin to interact with the product, organic visibility improves as the algorithm improves understanding of which audience is most likely to transform. In this way, advertisements can act as tiebreakers in determining product exposure and compete effectively with newly established sellers.

The integration of ML-driven retail media solutions benefits stakeholders across all markets.

• Buyer: Improve your shopping experience by presenting personalized recommendation-related products, reducing the time spent searching for items and increasing overall satisfaction.
• seller: In particular, the improved visibility of new products drives higher conversion rates and sales, helping sellers reach their target audience more effectively.
• market: Discovering optimized products increases engagement, improve conversion rates, improve advertising revenue, and enhances overall platform success.

A compelling example of this dynamic can be seen in how ML-driven ads can help surface products that may be overlooked. Stockx, for example, has a historically prioritized list based on hype-driven demand, such as limited edition sneakers and streetwear. However, this approach can be challenging for sellers who offer high quality but trend-driven products (such as well-known branded running shoes) in order to reach potential buyers. One independent retailer who runs a specialized running shoe store in the Chicago suburbs faced this precise challenge. Despite carrying inventory tailored to the profits of customers performing performance, the platform's organic discovery algorithms tended to favor more high-profile releases. By leveraging ML-led advertising, retailers were able to introduce their products to the right audience, dramatically increasing sales.

When done correctly, machine learning and advertising work together to connect buyers with related products that have not been discovered otherwise, bringing profits to both sellers and shoppers alike. In the case of StockX, our partnership with Moloco, an AI-driven advertising platform with proven ML models, demonstrated distinctive results by combining its own dataset with Moloco's models to enhance its in-house ad technology stack.

As the market continues to grow in complexity, machine learning emerges as an important driver for an optimized, scalable, profitable ecosystem. By leveraging ML-driven solutions, retailers can create more dynamic shopping experiences that benefit both buyers, sellers and platforms. From solving cold start issues to improving personalized recommendations, machine learning ensures that the right products reach the right customers at the right time. In an increasingly interconnected market situation, this approach is necessary, not merely a competitive advantage.

Tim O'Malley is Vice President of Products and Engineering stockxhe leads a global team of 90 people across product, engineering, design, data science and AI. His work spans buyer experience, payments, advertising, driving market transformation, fintech innovation and monetization. Previously, O'Malley was Chief Product Officer at Delivery.com and acquired his startup, Brinkmat. He began his career as a software engineer at Goldman Sachs and found a passion for technology leadership.