This article is reproduced from: Blue Ocean Yiguan
Author: Yiguan Analysis Group
Amazon has launched a new algorithm, This new algorithm will subvert the keyword ranking mechanism under A9 algorithm.
Next, the original traffic fence will be broken. Even if the differentiated products of vertical categories rank lower, they may also get more traffic and orders, while the original search ranking advantages of homogeneous products will be weakened.
This means that the fate of a large number of sellers will be reshaped.
A senior seller "felt" the existence of this new algorithm. The seller recently found that the subcategory ranking of some new products is rising steadily, while some top ranking standard products seem to have no previous advantages and order volume, and even if they keep rolling low prices, it seems that the effect is not ideal. This means that the products searched by users are no longer the same standard products or popular products.
The seller speculated that this was because "Amazon changed its algorithm".
In fact, this new algorithm really exists.
Recently, Amazon's official website Science Channel published a paper on the algorithm. The algorithm is named COSMO "Is a new algorithm based on LLM (Large Language Model).
In other words, this is a new AI ranking algorithm, It is very different from Amazon's original traditional algorithm based on keyword search.
Figure/Introduction to COSMO by Amazon Science Channel
This means that there are two algorithms on Amazon platform in the future, one is COSMO algorithm and the other is A9 algorithm.
"Blue Ocean Yiguan" believes that with the birth of this algorithm, Amazon has entered the era of AI e-commerce, and sellers will face huge changes, The traditional operation logic surrounding "search e-commerce" will also undergo great changes.
A group of sellers welcomed the order boom due to the birth of the new algorithm, A group of sellers will lose the ranking advantage under the original A9 algorithm.
The reason why Amazon launched this algorithm is that it does not want to sleep on the vested interests of "search e-commerce", but to embrace "AI e-commerce" with open arms“ It is a response to Temu and other competitive products that take away a large number of markets with ultra low price standard products.
With the help of COSMO algorithm, it is easier for consumers to find products that meet their personalized needs, such as "copper buttons, double pockets, dragon embroidery, pink cotton jackets", etc, Because it can meet the personalized needs of customers, it will be mined by COSMO algorithm, Get more traffic and conversion rate, and some standard explosives can no longer lie on the original search rankings as before, sitting on the maximization of traffic and orders.
Amazon said after passing the test, After COSMO algorithm optimizes the top search navigation bar, the product sales increase 0.7%。
There is no doubt that a major change has come for sellers.
Amazon CEO Andy Jassy confirmed a year ago that he would use AI to transform Amazon's ecosystem. In his shareholder letter in April 2023, he said that, Amazon is investing heavily in big language models and generative AI, This will be the core of Amazon's innovation in various businesses and fields in the coming decades, and will change and improve almost all user experiences.
The COSMO algorithm and Rufus, the AI question and answer tool launched a while ago, are direct responses to Andy Jassy.
In this context, the operation logic under the traditional "search e-commerce" has changed. If the seller does not follow Amazon's AI algorithm and traffic portal for a new layout, then the traffic and orders they get may be significantly reduced and lag behind their peers.
01 .
What is the magic of Amazon COSMO algorithm?
Under the traditional A9 algorithm, Amazon Conversion rate, search relevance, satisfaction and repurchase rate determine the ranking order.
"Relevance" is only one of the dimensions to judge the ranking of commodities, This may lead to a situation that when customers want to buy some products with personalized functions or personalized styles, But I found that countless popular standard products flooded the search results page Because these standard products have been purchased by many people in the past, their conversion rate, satisfaction rate and repurchase rate are often higher than those of niche vertical products, so they rank first.
In this case, the first feeling of customers is that they can't find the goods they want, so they have to change the keywords and search again or more times. In this process, customers often have a sense of frustration, which is a bad customer experience 。
"Customer first" is the highest tenet of Amazon. Amazon wants to change this situation so that customers can find the goods they want at the first time.
Thanks to AI technology, Amazon finally took a substantial step forward and launched The algorithm of "mining user's intention" ——COSMO。 This algorithm is more intelligent and understands users better.
Under the framework of A9 algorithm, the purchase process of consumers is usually "generate demand - search product keywords - platform algorithm selects products and presents them to customers in order", which is a process of "search commodity".
In this case, users generally do not express very specific and detailed intentions. Therefore, traditional search algorithms It is possible to recommend products that are "not needed" or "not so needed" by customers.
Today, COSMO algorithm analyzes the potential intention of user behavior through LLM (big language model), and upgrades [Search Product] to [Query Product Intent].
For example, if a pregnant woman wants to buy a pair of shoes, search for "pregnant women's shoes" in Amazon's search box. A9 algorithm will recommend "pregnant women's shoes" based on conversion rate and repurchase rate, which may not meet the "specific needs of pregnant women", while COSMO algorithm will base on a large number of stored in big language models such as GPT3 or OPT "Human common sense" and asking "why users buy or buy together" , analysis results "The importance of antiskid and other functions for pregnant women" , thus recommending anti-skid shoes, etc Products that can catch the pain points of pregnant women.
Amazon mentioned the COSMO algorithm in its paper. The first purpose is It is applied to the search algorithm to improve the accuracy of the search algorithm, followed by It is applied in two scenarios:
I Improvement of search navigation bar function; 3、 Conversation shopping is generally the AI question and answer shopping tool Rufus, which was exposed some time ago. It helps customers search, find and buy goods in the way of "question and answer".
Here, we will detail two practical application scenarios:
1、 A very specific application scenario of COSMO algorithm is the improvement of Amazon's original search navigation bar—— Multi wheel navigation 。 The so-called multi round navigation is to continuously provide multi round search options, Step by step, and constantly "explore" the real intention of customers.
For example, a user searching for "Camping" is actually a vague search. At the beginning, there will be camping product options such as "inflatable mattress", "tent", "blanket" and "lantern". If the customer chooses "inflatable camping mattress", Then the COSMO algorithm will progress layer by layer , recommend camping air cushions in various scenarios according to different needs, Including lake camping, mountain camping or 4-person camping.
Multi wheel navigation reflects a natural discovery process , enhanced the user's search experience and greatly improved the conversion rate.
It is revealed that COSMO algorithm has improved the search navigation bar, using 10% of the traffic in the United States for A/B testing, Achieved 0.7% sales growth, At the same time, the participation rate of search navigation bar increased by 8%.
2、 Another application scenario of COSMO algorithm is "session based shopping" Rufus, the AI question and answer shopping tool exposed a while ago, helps customers search, discover, research and purchase goods in the way of "question and answer".
Rufus's entrance is very prominent, located in the search box at the top of Amazon's shopping app. At present, in the search box of some buyers, not only the prompt of "Search Amazon" is displayed, And "Ask a question" This is the new entrance of Rufus.
Through this entrance, buyers can "chat", Let Rufus help him find products, You can even ask vague or even detailed product questions, such as, "My mother is 5.5 feet tall, weighs 120KG, and likes red. Please recommend a dress and shoes for her for Christmas", etc.
Fig. Answer and Recommendation of Rufus
Rufus also has an entry in the product details page. When a buyer enters a specific product page and is too lazy to look at the product details page or the numerous comments left by past buyers, he/she can also send "Rufus entry in the product details page“ put questions to.
For example, if a customer sees a sweatshirt on the product details page and is interested, but does not know much about the product, he can ask Rufus: "Is this shirt made of thick material?" "Is this sweatshirt suitable for running?" or "Does a person who is 6 feet 7 inches tall fit after wearing it?"
Rufus will immediately access Amazon category nodes, past buyers' comments, Q&A network information (Internet information, which may include the storage "common sense" and "knowledge" of big language models such as GPT3), and give detailed answers and recommendations.
For example, Rufus would answer the question "Does a 6-foot-7 inch person fit after wearing it
Fig./Rufus's answers to various questions, including size, material, etc
In these scenarios, COSMO algorithm will play a huge role.
02 .
What is the actual impact of Amazon's COSMO algorithm on sellers?
After the birth of COSMO algorithm, there are at least several major impacts on sellers:
I A new traffic portal was born: In the past, the search box was Amazon's largest traffic portal. After the COSMO algorithm was born, the entrance of the search box was diverted. One was the "traditional search entrance" led by A9 algorithm, and the other was the "traditional search entrance" led by COSMO algorithm“ Multi wheel navigation ”(Pop up recommendation words for customers to filter layer by layer).
At the same time, after the launch of Rufus, an AI dialogue shopping tool, it has also become a new traffic portal. In addition to using the search box to search for products, a large number of users will also use Rufus to "ask" and "find" products, such as, "What earphones are suitable for outdoor sports?" "What is a good birthday gift for a 50 year old father?"
Since then, Rufus becomes a "private chat friend" who can provide shopping guides , built a new shopping scenario - Rufus Q&A shopping scenario.
Both "multi wheel navigation" and Rufus are new traffic portals generated based on COSMO algorithm.
II Segmented and niche personalized products will get new opportunities: In the past, buyers often entered very short information in the search box, such as "pink dress". This is because Amazon A9 algorithm is not intelligent enough and can only respond to very short words or sentences, which led to a large number of "high score", "high repurchase" and "high conversion rate" stereotyped blockbusters and blockbusters on the search results page.
Now, after Amazon introduced the COSMO algorithm, the system has become more intelligent and smarter, It can answer very complex and detailed questions, and can also recommend very detailed products 。
For example, buyers can find one by multiple rounds of searching or asking Rufus "The pink dress, with three quarter sleeves, bows, wrinkles and buttocks, is used to attend friends' birthday parties, and can make me the focus of the stage." 。
This is a product that takes into account very specific needs.
In this way, the flow of some traditional "high-quality and popular products" may decline, while products with subdivision characteristics and attributes will be mined by COSMO algorithm, recommended to consumers, and more orders will be obtained.
This is mainly because the COSMO algorithm has three functions: 1、 Analyze the intention behind user behavior and attach importance to capturing potential consumption intention , improve the relevance and accuracy of search results, Generate personalized product recommendation; 2、 Automatically filter similar products to provide differentiated product selection; 3、 Allow users to filter based on specific attributes and adjust search results based on personal preferences.
This means that the original commodity ranking and traffic under the A9 algorithm will be reallocated. Amazon's ecology will be more long tailed and healthier, changing the situation where a few top sellers receive most traffic and orders.
So, what should the seller do?
1、 Layout more subdivided products, avoid excessive homogenization competition, and change the thinking of low volume price: The seller should make full research and development for the vertical subdivision track, and develop more products to meet the personalized needs In line with COSMO algorithm recommendation and "Rufus Q&A shopping scenario", these products will get new traffic and sales;
2、 The listing and product detail pages need to be rearranged: In the era when the A9 algorithm search box is the largest traffic portal, most Amazon sellers will focus on the operation of "search terms", bury these "search terms" in the title and description of their product links, and at the same time, in advertising promotion, they will also collectively promote these words through competitive bidding. However, with the emergence of COSMO algorithm recommendation and "Rufus Q&A shopping scenario", the traffic entrance and pattern have changed. As a result, the seller's operators have to change the operation direction with "search words" as the core, Instead, the layout should focus on COSMO algorithm recommendation and "Rufus Q&A shopping scenario":
（1） The pictures, keywords and descriptions of the listing title and details page should be consistent with the user's purchase intention, and accurately convey the characteristics and advantages of product segmentation;
（2） Use AI tools to optimize AI oriented listing so that it can be effectively identified by AI. "Try to make it easier for COSMO and Rufus to find our products and recommend them to customers", so as to obtain more traffic and orders.
3、 Layout outside the station: In order to make the COSMO algorithm more intelligent, We will certainly capture more data from inside and outside the station to train it, so that it can better understand the needs of buyers.
Off site brand independent stations, social media pages, and forum posts are all good places to lay out their own brand and product information. These information may also be included and recommended by COSMO algorithm and Rufus.
"Blue Ocean Yiguan" once asked a senior independent website seller: "You spend a lot of energy to layout the SEO content of independent websites, and most of the traffic of independent websites comes from Google. But after the rise of Chat GPT, it will become a new traffic portal instead of Google. Will your independent website encounter a huge crisis?"
The seller replied: "Like Google, Chat GPT does not produce content. It also extracts content from thousands of websites and pages, analyzes and recommends it to users.
As long as we do a good job in content (including product content), Chat GPT will recommend our content to users, so that we still have traffic. Remember, The entrance can be changed, but the best ammunition for each entrance is always the best content. "
To be honest, COSMO algorithm and Rufus also want to capture a large amount of content from outside the site. Then, we will lay out a large amount of our own brand and product information on the independent brand stations, social media pages, and forums outside the site, which will greatly increase the possibility of entering their included content.
Of course, A9 algorithm is still an important foundation of Amazon. Next, the two algorithms will go hand in hand.
Conclusion: At home, Alibaba, which has been pressed by Pinduoduo, has high hopes for AI e-commerce and regards it as a "nuclear weapon" against Pinduoduo and even Tiaoyin e-commerce in the future. Therefore, Ma Yun believes that in the AI e-commerce market, it is still unknown who will win the last laugh.
This situation also applies to cross-border e-commerce. When Temu is sweeping the world with the low price of breaking through the floor, Amazon's AI e-commerce may also go another way. (Yiguan Analysis Group)
(Cover source: Tu Chong's creativity)
(Source: friends of Hugo. com)
The above content only represents the author's own opinion, not Hugo's cross-border position! If you have any questions about the content, copyright or other issues of the work, please contact Hugo cross-border within 30 days after the publication of the work.
