Under the "hosted" and "self-operated" mode of e-commerce platform, the AI solution of one person and 10,000 stores
In 2023, the four tigers of cross-border e-commerce" Temu, AliExpress, Tiktok Shop, and Shein will make the "fully managed" model the hottest keyword in the e-commerce industry. The essence of the full custody model is that the merchant holds the right to the goods, and is responsible for the selection, delivery, and price verification, while the custody platform holds the business right of the merchant and is responsible for the pricing, operation, logistics and other related links of the goods. The fully managed model is more friendly to factory-based merchants who lack content production capacity and product operation capabilities, and can help merchants quickly clear their inventory in an environment where the overall market supply of goods exceeds demand.
Although the full custody model has the advantages of centralized traffic management and saving operating costs and labor costs for merchants, due to platform pricing, the profit margins of merchants are often not guaranteed, so the industry has given birth to the "semi-custody model", which returns the pricing power of goods under the full custody mode to merchants and provides more flexibility for merchants.
Whether it is the current popular "full custody" and "semi-custody" or the previous "self-operated model" of the e-commerce platform, it has brought a sudden increase in the pressure on the platform to operate goods, and also brought a new "operation model", that is, from the previous "one number of stores" to "one person and ten thousand stores" or even "one person hundreds of thousands of stores" The operation requirements This kind of problem related to production quantity and efficiency also needs to be solved under such a huge operation content, how to take the platform as a whole to improve the comprehensive operation effect and conversion, so as to improve traffic efficiency and increase the overall revenue of the platform. This is even more important in today's increasingly competitive environment of cross-platform e-commerce.
At present, the main means of e-commerce platforms for this kind of operational needs is to use AI tools to produce and process content in batches. This kind of AI batch tool is either independently developed by the platform, or provided by a third party to provide a tool-based interface call, and then a large number of designer teams are manually replaced by the mode of artificial + AI to increase the selling point of the product, and the product operator replaces the material to test the overall effect, but this kind of solution still has the following drawbacks:
The content output of the tool is semi-finished products, and there will be huge labor cost output
Taking the picture software of mainstream e-commerce as an example, the tool solves the problem of picture background generation and picture extension, and the output to the designer is mostly a product display map, and in the scene application of e-commerce, the main picture of the product and the key picture of the drainage position need to artificially increase the selling point, when it comes to a large number of goods, the synthesis of a large number of selling points needs to be completed manually, and the production capacity of a single designer peaks in dozens of pictures / day, in front of the number of millions of goods, which greatly affects the production efficiency of operating materials.
The separation of production and operation has a negative operational effect
Designers are producers of batch content, and each designer has a different understanding of the content, which leads to different quality and style of images produced by tools, which will lead to the dispersion of content strategy, the lack of a unified content strategy, and the difficulty of forming an effective linkage with the product strategy. After the operation team puts the materials jointly completed by the designer and the AI tool on the shelves, they will find that in most cases, the operation data is negative, and it is really difficult to define the reason.
The tool cannot iterate according to the operation effect to solve the industry know-how problem
Simple generation tools solve "technical problems", taking pictures as an example, the iteration direction of many instrumental software is how to make the combination of background and goods better, how to make the position of shadows and light and shadow more natural, and the core problem that the e-commerce platform as a whole needs to solve is that in addition to the above technical generation effects, it is also necessary to solve the specific scene problems, and accumulate and form the operation of the platform side "Know-How", such as "under what kind of background will the CTR conversion of the rice cooker be improved", " Is the angle of image generation better for heads-up effect or top-down effect?", "When the background is generated, is it better to use the original image to expand or use the white background to generate the effect", "Which categories are more suitable for generation and which are more suitable for extension"...... These are scenario-based requirements that need to be iterated one by one, and tooling is difficult to complete, and these scenarios will change according to the algorithm changes of the e-commerce platform itself, and the accumulation of industry know-how is the core of operational work.
RC Lightyear provides batch solutions for large-scale e-commerce platforms, different from traditional AI tools, which provides customers with a complete full-link solution in the true sense, which is created in the actual application of multiple customers, and after multiple rounds of verification, it can not only effectively improve the overall GMV of the platform, but also some projects are still positive when the platform calculates the input-output ROI, that is to say, the payment cost of the platform for batch generated content is less than the increase in revenue brought by its GMV growth, effectively forming a positive operation cycle, achieving a win-win situation for both parties, and also verifying such a basic factThat is, the hosted or self-operated e-commerce platform can effectively improve the operation effect of the platform as a whole through AI batch content production and operation (replacement of pictures, videos, selling points, etc.), including CTR, CVR, GMV and other operational data.
In essence, the batch full-link solution provided by Altron Lightyear is to deliver the final operation materials in batches through AI, and through "refined" operations, co-create content generation strategies with the platform, and then accumulate industry know-how according to the data results of the operation, iterate products in concrete scenarios and improve operational results.
For example, if a platform user wants to increase the CTR value of the public domain to improve the overall GMV of the platform, this is the core goal of the customer. Under this goal, the acceptance criteria for image generation have been changed to "comparison with the quality of the original image" rather than a single production.
Based on the goals, Ultron Lightyear starts with developing a content strategy with the client. According to the basic quality of the platform's pictures, Altron and its customers will divide the original pictures of merchants into three categories, which are divided into poor original picture quality (mostly white background images), medium to high picture quality, and several grades of excellent picture quality, and provide different technical means services such as background synthesis and background extension in different grades to improve the overall optimization effect. On the product strategy side, RC and customers will also divide the product contribution into different grades according to the GMV contribution value of the product, and also summarize the strategy of the product with different leaves.
Later, Altron then carried out mass production according to the product and content strategy, and finally submitted to the platform not only the product picture, but the main product image with a combination of selling points. However, the mass production content still needs to go through manual judgment of aesthetics (compared with the original picture) and data AB test before it can be launched, so Ultron will further iterate the strategy in the form of weekly meetings according to the results of the AB test, and after multiple rounds of generation, the total number of pictures + videos generated is 100,000, not only the established CTR improvement goal exceeds the original goal by 200%, but the absolute value of the positive rate increases by more than 20% It has created a positive GMV for platform merchants, and the ROI is also positive from the results of financial data, forming a good win-win pattern.
In the overall project, the e-commerce platform needs to invest no more than 3 people, and they only need to complete 4 things: provide original product materials and selling point information, jointly formulate product strategies, put product materials on the shelves, and regularly conduct data reviews in the form of weekly meetings.
It is worth mentioning that the early settlement method of this model is only based on the amount of material generated, which provides considerable cost flexibility for the e-commerce platform, without the need to invest a lot of manpower and annual tool purchase.
In this type of cooperation model, Altron Lightyear is not a single provider of AI content tools, but an AI in-depth partner under the customer's "one person, ten thousand stores" or "all-trust model". In the co-creation with customers, RC takes on the role of "content strategy" and "batch content production", and serves users through technical iteration and operational iteration.
Nowadays, after a large number of actual battles, Mogic AI can answer many specific operational questions, such as "what kind of background can be used to improve CTR in rice cookers?", "which categories should be converted with virtual backgrounds with high conversion efficiency and low conversion efficiency in actual scenes", etc.
On the basis of clear and unambiguous operational data, the technical team needs to focus not only on improving the light and shadow effects of various categories of products, but more importantly, on the effective integration of different categories of products and various excellent templates. In this way, we can continuously optimize the iterative effect and accumulate valuable industry operation experience, that is, "Know-How". This experience will be another competitive advantage for Ultron Lightyear in addition to its well-established technical capabilities. In the future, as RC cooperates with more new platform customers, this advantage will be more fully reflected and exerted.
Today, the hosting model has become the mainstream mode of factory e-commerce. The competition between e-commerce platforms mainly lies in the competition of traffic and the competition of traffic operation efficiency. Mogic AI has summed up a set of effective playing methods through a lot of practical experience, and will continue to iterate on playing methods in the future to explore more cost-effective solutions with e-commerce platform customers.