Unlike automatic ads, manual ads can be displayed directly in the top 3 positions of the search interface by adding a price. Therefore, when the competitor’s listing slowly rises, the selection of manual advertising keywords can be determined based on the “positioning” given to the competitor’s listing by the Amazon A9 algorithm.
Before deriving some basic logic of A9, we can first understand the next classic algorithm similar to search engines, the ant colony algorithm. Ant colony algorithm is a probabilistic algorithm used to find optimal paths. The theory was proposed in 1992 and was inspired by the behavior of ants in finding paths while searching for food.
The ant colony algorithm has the characteristics of distributed computing, positive information feedback and heuristic search. It is essentially a heuristic global optimization algorithm in evolutionary algorithms.
The ant colony algorithm is to find the shortest path to the destination. The ultimate purpose of search engine A is similar, which is to recommend the products that customers are most likely to buy based on the different keywords they search for. From this theoretical derivation, there is no difference in the logic level of the ant colony algorithm, but thousands of The search and transaction information of thousands of customers is like the crawling trajectories of countless ants in the ant colony algorithm, which will eventually converge to an optimal path or solution.
Take the traffic change charts of several products as an example and explain it in the form of Q&A.
Q: Why does product No. 3 have more intense traffic growth and changes than products No. 1 and 2?
A: Because product No. 3 is a hit, Amazon’s A9 algorithm It was found that the exposure click conversion rate and traffic order conversion rate of this product were much higher than other products, so it gave greater exposure and therefore greater traffic.
Q: Why does traffic not continue to grow? How to judge when a product reaches the peak exposure given by the A9 algorithm?
A: Refer to the ant colony algorithm, the so-called exposure in A9 The peak is just the shortest path, just like the straight line between two points is the shortest. All routes with bends or detours are not the best path. So on the Amazon platform, it can be understood as all the routes that are on the rise. Or the downtrending products are not on this optimal path. This can be reflected by the related product columns below the product listing. In Amazon’s product columns, they are generally divided into the following categories.
The above categories belong to “Sponsored products related to this item”. This category is mostly related to advertising product recommendations.
The above categories belong to “Customers also shopped for”. This category is mostly recommended for products in broad categories and is not very related to the original listing.
The above categories belong to “Customers who bought this item also bought”. This category is an order-related product. The order correlation degree in the clothing category is generally 2%~5%, that is, every 100 orders. 2~5 orders will purchase products from this listing and this recommended category at the same time.
The above categories belong to “Customers also considered”. This category is a recommendation for products in a broad category and has a certain correlation with the original listing.
Displays the previously mentioned “4 stars and above Sponsored” recommended product advertisements. The products displayed are all products with more than 100 reviews and a rating of 4 stars or above. According to its appearance position and recommended products, it can be found that most of its products are similar to the original product styles, and most of its products ranked within 50,000 are hot-selling products on the platform for similar purposes.
Let’s look back at the proposition just now. If a product reaches its peak, the Amazon platform will try its best to match it to the appropriate product category in order to obtain the maximum profit. Through the recommendation column below Give the most accurate combinations and recommendations to maximize commissions and advertising profits. Therefore, if you see that there is no “Customers who bought this item also bought” column or “4starsandabove Sponsored” does not appear below a product, the listing has not reached a stable position and will be in a serious fluctuation stage.
At the same time, even when a product has received 50 or even 100 orders per day, but the column below is still “Customers also considered” or “4 stars and above Sponsored” does not appear in the first recommendation position, then It is 100% certain that the products in this listing must be popular and their sales/traffic will continue to rise. The traffic/orders will not enter a stable period until Amazon finds a matching product category and relevant recommendations. This way, competition can be determined. Whether the opponent is in a single stable period, if isting itself is still in the rising stage, you can grab a certain market share through manual advertising.
Q: How to catch up with other sellers’ listings through manual advertising? Super or grab a certain market share?
A: Each product has its own corresponding attributes, and each corresponding attribute is similar to a value in a matrix in mathematical linear algebra. Take the T-shirt as an example.
The root attribute of the T-shirt is Blouse, and the other attributes are T-shirt, round neck, short-sleeved, casual, solid color, women, cotton…
Then you can first construct the following matrix (note that the content written in this section is not a strict mathematical calculation. It is just an attempt to draw relevant qualitative conclusions through the process of logical derivation to help us better engage in operational work. ):
Then we translate it into English:
Note that the above two steps are only for ease of understanding, and the value of each element in this matrix is In the Amazon search engine The search ranking generated by searching for the combination of any one of the attributes and the root attribute is the rank value.
For example, if you search for women blouse, the product ranking is 100, then the number in column 1 and row 2 of the above matrix is 100. At the same time, the search results of the entire keyword combination (how many products can be searched in total) are recorded as another matrix, and the Amazon platform is used to fill in each search result into the above matrix.
The search result matrix (more than X0000). is calculated as The words “short-sleeve Blouse” in the 70th place, “cotton Blouse” in the 80th place and “summer Blouse” in the 80th place represent a high degree of matching between the product and these keywords, but this is only the first layer of A9 attribute matching. , it is not accurate yet, and a second attribute matching of A9 is required.
Select a common attribute and a strong attribute to match as the new root attribute, and then search for the matched attributes. The search rankings are filled in, starting from a row and a column, and the new search result matrix is also recorded.
New search result matrix (calculated as X0000 if it exceeds X0000).
Other matrices can be deduced in the same way. In the second matching, there will be a total of 9 new matrices. One value in each matrix will remain unchanged, and the other values should become smaller to a certain extent. The ratio of the reduction is the ratio between the attribute and the new root. How well the attribute previously matched. Each two matrices of secondary matching can be understood as the former matrix is Y, which is the result, and the latter matrix is X, which is the environment.
Through the above secondary matching, we can already obtain the listing search rankings and the total number of search results corresponding to multiple keywords. Set the search ranking of a certain keyword to R and the total number of search results for this keyword to N, then the search visibility probability P(X) corresponding to the keyword matrix Y (the search visibility probability is a probability model, when the probability When the value is very close to 1, it means that when the user enters the keyword, almost 100% of them can browse to the product listing), so it can be calculated as follows:
P(X)=(N-R)/ N.
Taking the above primary matching matrix as an example, search for “T-shirt Blouse”. At this time, R is 80 and N is 20000. Then P(X)=(20000-80)/20 000=0.996 , that is 99.6%. We can think that when a user enters “T-shirt Blouse” on the Amazon platform, there is a 99.6% probability of seeing the product.
Take the above quadratic matching matrix as an example, search for “T-shirt Blouse round neck”, then R is 180N is 4000, then P(X)=(4000-180)/4000= 0.955, which is 95.5%. We can think that when a user enters “T-shirt Blouse round neck” on the Amazon platform, there is a 95.5% probability of seeing the product.
Of course, this probability only has a quantitative logical meaning, that is, the meaning of “big and small”, rather than a precise numerical meaning. You only need to calculate the search visibility corresponding to each element in the keyword matrix. Probability.
We need to count the high-valued items, and then avoid the combination of these keywords when listing the same product, and choose keywords with a larger primary matching probability but a smaller secondary matching probability. Combined as keywords, this way, when users search for certain high-traffic keywords, our products can be displayed first than those of our competitors.
Through the above analysis, we can understand that each product in the Amazon platform has its fixed search order. When your competitors have occupied a certain order, whether through advertising, FBA It is difficult to defeat already leading competitors by any means such as , review scores, etc. However, when the competitor’s listing is still in the rising stage, we can catch up through the top exposure of manual advertising and the keyword matrix analysis of the competitor’s listing.
When using this technique, please make sure that your competitor’s listing has the following characteristics:
The listing is still in the rising period, that is, it has not yet reached a plateau in rankings and sales;
p>
Products have multiple colors/sizes/sub-categories;
Products of different colors/sizes/sub-categories have distinct pictures.
Assume that the picture is the listing picture of our competitor. There are 10 sub-products of different colors under the listing. At the same time, we can also determine that the product is in the rising period and has not yet reached a plateau.
Assuming that the image is the title used by our competitors, then we can split the keyword/title combination of the product into the following categories:
(1) faux wrap long dress ;
( 2 ) short sleeve;
( 3 ) high low hem / asymmetric hem ;
( 4 ) loose / casual;
< p>(5)t shirt/tee/tops/blouse;
(6)Basic.
We arrange and combine the above 5 phrases with a total of 10 keywords, and then fill them in Amazon’s search engine column as keywork to search. You will find that sub-products of a certain color have a high probability of appearing in the search. results (for example, red color products appear in search results at a rate of 47%).
Through the above steps, you can find the color with the best sales or the highest click conversion rate in this listing, and then fill in the keywords corresponding to the remaining 53% of the search results into the keys of manual advertising. In the word column, select the best-selling color as the exposure object of manual advertising, and then choose a single click fee that is willing to increase by 50% so that the product link can appear in the top 3 or top 5 search results, so that you can get a lot of traffic and exposure. , thereby increasing the probability of catching up with competitor listings.
Through the above quantitative analysis, we can draw a conclusion:
The A9 algorithm does not understand at the beginning what is a hot item and what is not a hot item, it just gives After you are exposed, analyze the click conversion rate and order conversion rate of the product to determine whether you need to give you more traffic. Therefore, when the A9 algorithm finally finds the final positioning of a product, it can analyze the ranking search results and use manual advertising for exposure to catch up in sales.