We already know that “hot items” perform very well in terms of rankings and sales, that is, hot items in a short period of time will increase their rankings at an alarming rate. Therefore, we can use this logic to capture Amazon’s ranking information to learn which products have the potential to be popular.
Practitioners engaged in Amazon operations all know that although product ranking is a comprehensive ranking, it is mainly related to the size of the order. If your order volume for a certain style increases rapidly, its ranking will also rise rapidly.
Now that we know the process of generating a hot item, how do we find these data changes? First of all, the traffic and order volume of a certain style from other sellers are invisible. These are commercial items. Only changes in product rankings can be obtained and analyzed by us.
Through the above logical deduction, the crawler program (if the operator does not master programming skills, you can use third-party crawler software such as “Octopus” to complete the data capture) data capture and product selection process As follows:
(1) Select the major product categories involved on Amazon.
(2) Use a programming language crawler program to crawl 400 pages of product ranking information in this category.
(3) Set the data crawling update cycle, which is recommended to be 4~12 hours.
(4) Import the ranking change data into data analysis software or programs, and establish a model to evaluate whether the product is a “hot item”.
(5) Use all listing optimization methods to systematically optimize the product that is ready to be put on the shelves, and strive to defeat the opponent before the previous seller gains a large market share.
After using a crawler program to obtain the ranking of each product on Amazon, you need to do data analysis based on the ranking of each product. Generally speaking, it is divided into the following steps:
The first step : Confirm the product selection exit mechanism.
First, analyze the characteristics of “unpopular” styles based on a large amount of data statistics, so as to establish an exit mechanism to promptly kick out products whose rankings are falling rapidly. Generally speaking, the order of product listing ranking decline is: sales volume decreases→traffic decreases→sales volume finally reaches 0→ranking drops rapidly. These are the ranking data changes of 10 randomly selected products that have declined in ranking on the Amazon platform. It should be noted that although the literal expression is lower than the ranking, the numerical value shows an increase, because on Amazon, the smaller the ranking value, the greater the sales volume.
There is a certain pattern for the ranking to drop when there are no orders for the listing. The average value is around 30,000. That is, when there are no orders for the listing, the ranking value will increase at a rate of about 30,000/d. Therefore, we The following product selection exit mechanism can be established: if the ranking value of a certain listing increases by an average of 30,000 per day on x days (x can take any value from 3 to 30, depending on the category, the time can be relaxed for categories with high competition). When left or right, the product is judged to be an “unpopular” product and will not be considered for selection.
Step 2: Eliminate hot-selling models with stable rankings.
The reason why stable models need to be eliminated is because when the sales volume of a certain listing is stable, its ranking will also become stable. Amazon’s A9 algorithm will quickly help the product find its corresponding potential customers. Choosing it as a shelf style has missed the best time to put it on the shelf. It is difficult to overtake the former, so don’t get rid of these products.
Suppose I record data at the same time (for example, record data at 7:00~8:00 China time for 10 consecutive days). Taking the bodystocking category as an example, we can get the following data fluctuation conclusions:
p>
(1) Listings with ranking standard deviation less than 50,000 can almost all find stable search positions in the A9 algorithm. (That is, there will basically be a recommendation column for Customers who bought this item also bought below the listing).
(2) The ranking standard deviation is greater than 50,000 and less than 100,000. An unstable search position recommendation column will appear below most listings (that is, there will be a Customers who viewed this item also viewed recommendation column below most listings, which will appear. The proportion increases as the standard deviation of the rankings increases).
(3) Listings with a ranking standard deviation greater than 100,000 are almost unable to find a stable search position in the A9 algorithm (that is, the Customers who viewed this item also viewed recommended column will appear below the listing).
Based on the above conclusions, find those hot-selling models with stable rankings among many styles, and then eliminate them.
Step 3: Find potential styles.
There are two major categories of product selection plans for potential styles: one is the listing selection with rising rankings (that is, the ranking value shows a rapid downward trend).
The second is the selection of non-stable ranking listings. Select those listings with better ranking performance but higher standard deviation of ranking values. Large fluctuations in their rankings indicate that this listing has not found its suitability in the A9 search field. Search location. On the one hand, it may be due to insufficient operational level; on the other hand, it also means that the A9 algorithm has not accurately positioned the potential customers of the product itself, and there are still opportunities.
Step 4: Track the ranking of potential models in real time and finally determine the styles to be put on the shelves.
After completing the first three steps, we have selected some potential models through data analysis, and their proportion should be around 10%. However, given that the hit rate is generally less than 5%, a streamlining process still needs to be done, and only the ranking of potential models needs to be monitored in real time. Once it is found that its ranking has a stable trend and its sales volume is high, the product must be a hot seller and can be put on the shelves for sale immediately, as shown in Figure 4-17.
Of course, data-based product selection requires strong supply chain support. If your own supply chain is relatively weak, don’t worry. We can use the same steps to grab hot-selling products from wholesale websites and move them to Selling on Amazon.
It should be noted that if you choose data type product selection, the time for data capture and statistics is at least 1 week. At the same time, in order to reduce the risk, another 1 week of observation is needed to confirm whether the style is Potential money. However, if you wait until it is confirmed that it is a potential product and then integrate the supply chain for production, at least one month will have passed since the first data analysis by the time the FBA product arrives. Although the growth cycle of a product is generally more than one month, given that it is very difficult to track the data fluctuations of each new product on Amazon in real time through a crawler program, it is best to combine this “data analysis + test verification + real sales” The cycle is controlled within 21 days.
It is recommended to solve this problem through the following two solutions:
(1) First collect data in the early stage, find some potential styles, and then find out whether they are on 1688 (Alibaba) Or there are corresponding links on other domestic wholesale websites. If the seller with the link supports small order sales, you can directly purchase small batch orders.
If the quality is confirmed to be above average, within 10 to 15 days after the first data capture, the style will be confirmed to be a potential model, and the small batch of orders will be sent directly to FBA for sale. When the goods arrive, observe their traffic trends and decide whether to continue purchasing from 1688 or produce them yourself.
(2) Conduct data collection in the early stage, find some potential styles, and start making samples directly. Within 10 to 15 days after the first data capture, confirm that the style is a potential style, and confirm that the sample is sent. We strive to complete product production within 25 days, and strive to send self-produced products to FBA for sale within 1 month.
Considering that many small and medium-sized sellers on the Amazon platform purchase products on websites such as 1688 and then sell them on the Amazon website, we can analyze the sales data on these wholesale websites in real time and then convert the sales into Products with skyrocketing data are put on Amazon for sale.
Take skirts as an example. We first search for “Amazon skirts” on 1688, and then we can see the specific sales information.
Then use the same method to use a crawler program to capture its sales, and use the growth and fluctuation of sales to determine which styles have been hot-selling in the near future, thereby helping operators determine what are the hot products in the near future.
It should be noted that this method has advantages and disadvantages.
Pros: Real-time tracking of competitor data changes, not letting go of any product on the Amazon platform that has the potential to become a hit; the evaluation of “hot items” is objective and scientific, which can save a lot of product selection costs and product testing time .
Disadvantages: Ranking data still has a certain degree of lag; very high requirements for T capabilities and data mining capabilities of operators or industry organizations; efficient compilation of algorithms is required, otherwise the information processing cycle will be too long; for supply Chain requirements are higher.