|
One of the ways to survive on Amazon: the ranking of listings determines the traffic. The higher the product ranking, the higher the exposure rate, and the more traffic it captures. The underlying logic of its operation is based on Amazon's A9 algorithm, which selects the most relevant products from Amazon's huge product categories and displays them to customers based on relevance. Specifically, the factors that affect the A9 algorithm's ranking can be summarized into three indicators: conversion rate, relevance, and customer satisfaction and repurchase rate. In other words, the more a product meets the customer's search needs, stimulates the desire to buy, and provides a good consumer experience, the easier it is to be recommended to the top by the A9 algorithm. Therefore, Amazon sellers’ operational ideas and sales strategies are all based on the A9 algorithm. It can be said that the A9 algorithm is like a central control system that controls the fate of tens of millions of products. But the system needs constant maintenance and upgrades to run for a long time. Amazon has reformed the A9 algorithm many times. However, recently, Amazon has launched a new round of algorithm revolution - launching a new weapon, the Cosmo algorithm, based on the A9 algorithm. For sellers, once the new algorithm is put into large-scale application, what impact will it have on sales operations?
It is understood that the COSMO algorithm has not yet been fully put into use, but Amazon has secretly conducted small-scale trials in recent months. It is reported that Amazon extracted 10% of the keyword search results from the US site to test the new AI ranking mechanism, and the results showed that the conversion rate increased by 0.7% . Based on this, it is estimated that the optimization generated by the COSMO algorithm can increase sales by approximately US$4.9 billion a year. So compared with A9, what specific changes has the new COSMO algorithm brought to Amazon sellers?
Follow the [Cross-border e-commerce] public account and reply to the keyword [COSMO algorithm] in the background to obtain the official complete information The official document introduces it this way: Although the existing e-commerce knowledge graph integrates a large number of concepts or product attributes, it cannot accurately discover the user's intention. Based on the need to understand the personalized needs of consumers, COSMO came into being. By mining massive user behavior data, it builds a user-centric knowledge graph, thereby enabling diversified online e-commerce services. Users generate massive amounts of behavior logs on e-commerce platforms every day, so Amazon uses large language models (LLMs) to mine the intentions behind these behaviors, that is, search results not only match the search terms, but also meet the user's potential consumption needs. Amazon has now expanded COSMO to 18 major categories, generating millions of high-quality knowledge to capture user intentions. The consumer purchasing process is usually: demand is generated - product keywords are searched - platform algorithms select products and display them to customers based on relevance. However, the deep-seated needs of consumers in this process are often not intuitively presented. For example, when buying a pair of shoes, some people prefer practicality, while others focus on aesthetics. The role of COSMO is to continuously learn from user behavior, gain insight into potential needs that may arise, and make targeted product recommendations. In contrast, COSMO's workflow is as follows: - User behavior issues instructions
- LLM receives and generates knowledge graph
- Filter similar products to finally select the results that meet user needs
We can feel it more intuitively from specific cases. For example, if a pregnant woman wants to buy a pair of shoes, she enters and searches for "pregnant women's shoes" on the platform. Based on common sense analysis, COSMO-LM concludes that anti-slip is important for pregnant women, and ultimately recommends products such as anti-slip shoes that capture the pain points of pregnant women. Furthermore, COSMO's core function is multi-round navigation. That is, it provides multi-round search improvements through continuous recommendations. For example, searching for "camping" may lead to the selection of "inflatable mattress", and then recommend "camping inflatable mattress". COSMO will provide various types of camping air mattresses according to different needs, such as lakeside camping, mountain camping or 4-person camping. Such multi-round navigation can achieve deeper and more accurate demand positioning, and significantly enhance the user's search experience in such a natural demand discovery process. From the above, we can see that the launch of the COSMO algorithm does not mean that the A9 algorithm has completely withdrawn from the stage of history. In fact, it introduces a more intelligent recommendation system based on A9, from product-centric to user-centric, and achieves more accurate content push through insights into users' real needs. Many sellers believe that this is equivalent to Taobao's algorithm of "one thousand faces for one thousand people". However, there are still some differences between the two. The algorithm relies on big data and cloud computing capabilities to achieve intelligent recommendations through store tags and crowd tags, filtering out products that users are not interested in, so that everyone's page recommendations are their favorite products, thereby helping users make more accurate decisions in advance. In contrast, due to factors such as foreign privacy policies, it is difficult to achieve Taobao-style personalized pages on Amazon. The former pursues that each user can have an exclusive personalized page, helping sellers to lock in target customers and achieve precision marketing, while the latter focuses more on the matching of search results and the improvement of accuracy.
Although the COSMO algorithm is still in the small-scale testing stage, once it is put into large-scale application, it is foreseeable that it will have a profound impact on the operational strategies of Amazon sellers. Some sellers have already sensed the subtle changes brought about by the new algorithm: "This month, the ranking of a new product subcategory has been rising. Last year, the top products were all standard products, and various differentiated products were not in the top 100 before. The standard product discount seems to be useless. This way, users are not searching for the same products, and the user experience will be significantly better. I speculate that Amazon may have changed the algorithm." Based on the above, the core points of the COSMO algorithm can be simply summarized into the following aspects: 1. Analyze user consumption behavior and generate personalized product recommendations 2. Automatically filter similar products and provide differentiated product options 3. Focus on capturing potential consumer intentions and improving the relevance and accuracy of search results 4. Allow users to filter by specific attributes and adjust search results based on personal preferences Under the logic of the A9 algorithm, the conversion rate, relevance factors, customer satisfaction rate and retention rate of the product are analyzed to perform intelligent sorting and present it to customers. Therefore, in the process of formulating the operation strategy, keyword optimization is the key content. However, from the core points of the COSMO algorithm, sellers need to be more inclined to analyze and capture users' potential shopping intentions and gain insight into real consumer pain points on the one hand; on the other hand, they need to focus on creating differentiated advantages for products in order to meet users' personalized needs. Therefore, under this logic, Amazon sellers can make the following adjustments to their operations: - Deeply understand the target audience and create high-quality products that meet their needs
- Listing pictures, keywords, descriptions, etc. should be in line with the user's purchasing intention and accurately convey the product features and advantages
- AI-guided listing optimization strategy to enable effective AI recognition and avoid repeated filtering
- Avoid homogeneous competition, build product differentiation barriers, and provide diversified and personalized product choices
In a nutshell, the formulation of operational strategies should be based on anchoring target users and taking user needs and intentions as the first principle. It is not only necessary to let users understand what their needs are, but also to let AI understand what products you sell. Some sellers believe that if COSMO is widely used, it will change from the traditional operation logic of keyword optimization to studying and satisfying the needs of the crowd, fully reflecting the strong correlation between products and needs. In this way, the differentiated advantage of the product will be elevated, and the phenomenon of low-price involution may be alleviated to a certain extent. Of course, some sellers are not so optimistic about this. "On the contrary, this algorithm will increase PPC. No matter which country, low-priced goods have the largest sales volume, and this method is similar to the "one thousand faces for one thousand people" of domestic e-commerce. Just like Taobao's advertising costs are still the largest, merchants' bidding is still fierce."
While the launch of the COSMO algorithm has brought profound changes to the platform's search logic, traffic distribution, and sellers' operational ideas, we can also gain insight into a major development trend of Amazon: embracing AI e-commerce. Prior to this, Amazon's attempts to empower e-commerce with AI have been endless, and it has launched a number of AI tools: including AI review summary, AI generation of listing descriptions, AI production of pictures, etc. Of course, all of the above are generative applications of AI, and are also the most common AI trend among major e-commerce platforms in the cross-border industry. These production tools have more indirect impacts on the operations of sellers. The COSMO algorithm introduces an AI mechanism that continuously learns and analyzes user behavior through a large language model (LLM) to generate a huge knowledge graph, thereby quickly identifying consumers' potential intentions and presenting them in a more refined screening mode and interactive mode, guiding buyers step by step to purchase the products they really need. It is worth noting that Amazon recently launched an AI search tool - Rufus , which provides product consultation, recommendation of similar products and comparison of models in the form of chat questions and answers. Rufus is a traffic window independent of the search box. Buyers use this AI content interaction method to find product needs in application scenarios, which to a certain extent subverts the operating logic and traffic model of traditional search e-commerce. As a search e-commerce platform, Amazon used to rely mainly on keyword searches to match audiences, match products, and drive traffic conversion. But whether it is the launch of Rufus or the testing of the COSMO algorithm, they all send out the same signal: using AI to influence traffic layout. Some sellers bluntly stated that Amazon is trying to change the way buyers and listings connect. That is, Amazon not only wants to rely on "keywords" as a bridge to establish exposure and clicks between buyers and links, but also wants to further explore buyers' search needs and match more accurate links through "semantic interaction" and "finely filtered search recommendations". Under the search e-commerce thinking, the listing title, five points, and description all focus on popular words and long-tail words to serve the purpose of search word promotion. That is, during the optimization process, you need to think about "what words buyers will search for." Both Rufus and COSMO algorithms have overturned this way of thinking. The former requires thinking about the purpose and needs of buyers asking questions to Rufes, while the latter predicts the seller's possible purchasing intentions. In essence, both algorithms are about exploring consumers' potential needs, helping consumers clarify their real needs, and achieving accurate product recommendations. To a certain extent, both of them are aimed at people who "have purchasing needs but unclear needs", which is beneficial for sellers to tap into potential and marginal traffic. Buyers use AI to accurately locate products, while sellers use AI to anchor target groups. One seller commented: "Whether it is the Rufus portal or COSMO refined search traffic, will Amazon buyers eventually get used to and use these search functions more frequently? If so, it means that the traffic pattern is gradually evolving, and sellers' traffic diversion methods and link writing rules will also change significantly as a result." In short, whether at the platform level or the seller level, the entire cross-border e-commerce industry has embraced the AI wave. Now the platform algorithm changes are uncertain, and the traffic pattern is changing in a confusing way. For sellers, the most important thing right now is to think positively, assess the situation, take the initiative to change, and adapt to the times.
|
