A live e-commerce virtual shelf intelligent building system

Abstract

The application discloses a kind of live broadcast electric business virtual shelf intelligent construction system, it is related to live broadcast electric business and intelligent shelf technical field, the system includes: acquisition module global acquisition and standardization processing multi-source data;Prediction module calculates inventory demand and early warning threshold;Monitoring module compares inventory and threshold in real time, triggers adjustment instruction;Adjustment module executes hierarchical adjustment, push alternative goods and pre-sale link;Visual module integrates supply chain data to realize dynamic display, forms closed-loop optimization;The application is integrated with intelligent algorithm by global data, inventory demand is calculated in real time and early warning threshold is set, accurately triggers shelf adjustment, hierarchical adjustment strategy matches sales scene, push alternative goods and pre-sale link, reduce customer loss, visual display supply chain information, improve the smoothness and controllability of operation, excavate live broadcast sales opportunity, help live broadcast electric business efficient operation.

Description

Technical Field

[0001] This invention relates to the field of live-streaming e-commerce and smart shelf technology, specifically to a smart virtual shelf building system for live-streaming e-commerce. Background Technology

[0002] Live-streaming e-commerce has become one of the mainstream consumption scenarios. As the core carrier for product display and transaction conversion, the virtual shelf's display strategy and inventory matching directly affect product exposure and user purchasing decisions. With the expansion of the live-streaming e-commerce industry, the increasing variety of product categories, and the diversification of live-streaming scenarios, real-time fluctuations in the sales process are becoming more and more obvious. This places higher demands on the dynamic adjustment capabilities of the virtual shelf and the accuracy of inventory management. Live-streaming operations involve multi-dimensional information such as sales data, inventory data, and scenario data. The dispersion and dynamism of this data make it difficult for traditional management methods to achieve efficient integration and utilization. At the same time, inventory demand is affected by multiple factors such as real-time sales rhythm and differences in live-streaming scenarios. Accurately predicting inventory demand and seizing replenishment opportunities to avoid customer loss due to stockouts or resource waste caused by inventory backlog has become a key requirement for live-streaming e-commerce operations. Against this backdrop, there is an urgent need for a virtual shelf building system that can integrate multi-source data and achieve intelligent prediction and dynamic adjustment to adapt to the dynamic operational characteristics of live-streaming e-commerce.

[0003] Traditional live-streaming e-commerce virtual shelf management relies heavily on manual experience and judgment, lacking systematic data support and intelligent processing mechanisms. This results in numerous limitations. In terms of data integration, it struggles to comprehensively cover multi-dimensional information such as sales, inventory, and scenarios. Data fragmentation leads to incomplete inventory demand forecasting, often relying solely on single historical sales data and ignoring the impact of real-time sales fluctuations and differences in live-streaming scenarios. This results in significant discrepancies between forecasts and actual demand. Regarding inventory monitoring and adjustment, replenishment trigger conditions are simplistically designed, lacking tiered response logic and failing to flexibly adapt to inventory tightness and sales rates. This leads to untimely or inappropriate shelf adjustments. Furthermore, the selection of alternative products lacks scientific basis, making it difficult to accurately match user needs. The lack of complete, traceable records after shelf adjustments hinders subsequent operational optimization. Additionally, information fragmentation across the supply chain prevents operators from real-time monitoring of the entire process, including stock preparation, allocation, and delivery. This results in a lack of effective supply chain-side linkage between inventory and shelf adjustments, further impacting live-streaming operational efficiency and user experience. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a smart virtual shelf building system for live-streaming e-commerce. This system comprises five modules: data collection, prediction, monitoring, adjustment, and visualization. It collects and preprocesses multi-source data across the entire domain and uses a dynamic inventory demand prediction algorithm to calculate inventory requirements. The monitoring module compares inventory with thresholds in real time and triggers adjustment commands. The adjustment module executes tiered adjustment strategies and pushes alternative products and pre-sale links. The visualization module displays supply chain information, enabling intelligent building and dynamic optimization of the virtual shelf, thereby improving the operational efficiency and user experience of live-streaming e-commerce.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a smart virtual shelf construction system for live-streaming e-commerce, the system comprising:

[0006] Data Acquisition Module: Collects multi-source data related to live e-commerce across the entire domain, and performs standardized preprocessing on the raw data to form a structured parameter set;

[0007] Prediction module: Receives the structured parameter set from the acquisition module, executes the dynamic inventory demand prediction algorithm and the replenishment trigger threshold algorithm in sequence, completes the calculation of commodity inventory demand and the setting of replenishment trigger warning threshold, and transmits the warning threshold to the monitoring module;

[0008] Monitoring module: Receives real-time multi-source data on goods from the acquisition module, retrieves replenishment trigger warning thresholds from the prediction module for real-time comparison, and generates standardized shelf adjustment instructions and corresponding product-related data when the real-time inventory of goods meets the trigger conditions.

[0009] Adjustment module: Receives shelf adjustment trigger instructions and product-related data from the monitoring module, performs hierarchical adjustment operations on the virtual shelf display priority, filters suitable alternative products and pushes their display links, generates and pushes pre-sale links for out-of-stock products, and generates a complete adjustment record after completing all adjustment operations.

[0010] Visualization module: Receives product-related data from the monitoring module, connects with the external supply chain to obtain full-chain data, combines virtual shelf adjustment records, completes the real-time display of product supply chain information on the virtual shelf, and feeds back data update records to the monitoring module.

[0011] Furthermore, the data collection module collects multi-source data related to live-streaming e-commerce, including real-time sales data such as clicks, add-to-cart volume, sales volume, and real-time growth rates of various indicators for products in the live-streaming room; omnichannel inventory data including actual inventory, occupied inventory, and available inventory in online and offline warehouses; historical sales trend data of daily and monthly sales and category-related sales data for products in the recent period; basic attribute data of product categories, specifications, and supply chain replenishment cycles; and live-streaming scene feature data configured on the live-streaming time slot and the anchor's operation end. Real-time data connections are established with the e-commerce platform's sales end, inventory management end, product information end, and live-streaming operation end through API interfaces. At the same time, historical data is batch-fetched from the platform's database. After collection, the data is classified and integrated according to data type, and then standardized preprocessing operations are performed on the raw data to form structured parameters.

[0012] Furthermore, in the prediction module, the mathematical expression for the prediction algorithm of the dynamic library is:

[0013] in, Dynamic inventory demand forecast; , This represents the average daily sales volume of the product in the same live streaming scenario over the past 30 days. This refers to the product's peak sales cycle coefficient. The basic inventory demand for goods based on historical sales trends; The real-time sales fluctuation coefficient has a value range of (0,2] and is determined by the ratio of the real-time growth rate of real-time clicks, add-to-cart, and transaction data in the live broadcast room to the historical average growth rate. It reflects the sudden fluctuations in real-time sales during the live broadcast. This is the potential best-selling coefficient for a product, with a value range of (0, 1.5], which quantifies the likelihood of a product becoming a best-seller in this live stream. This is a correction factor for live streaming scenarios, with a value range of [-0.3, 0.5]. It is set according to the live streaming time, the streamer's traffic level, and the promotional activities in the live streaming room to adapt to the sales differences in different live streaming scenarios. This refers to the amount of inventory that can be shared and deducted across all channels, and the amount of inventory that can be transferred to other sales channels.

[0014] Furthermore, in the prediction module, the mathematical expression for the replenishment trigger threshold algorithm is:

[0015] in, The alert threshold is set for when goods are replenished. The dynamic inventory demand forecast value for goods is obtained from the dynamic inventory demand forecasting algorithm; For supply chain replenishment cycle coefficient; Adjust the response latency factor for the virtual shelf; This is the minimum inventory level required to display a product. This is the inventory safety redundancy coefficient.

[0016] Furthermore, in the monitoring module, the triggering conditions are a main triggering condition superimposed with multiple auxiliary triggering conditions. The main triggering condition is that the real-time inventory value of the product is less than the replenishment triggering warning threshold, and this condition persists for a preset duration of 5 seconds. The first-level auxiliary triggering condition is that the ratio of the real-time inventory value to the replenishment triggering warning threshold is ≤0.8, triggering a regular adjustment command. The second-level auxiliary triggering condition is that the ratio of the real-time inventory value to the replenishment triggering warning threshold is ≤0.5, triggering an emergency adjustment command. Simultaneously, a sales rate triggering condition is added, which is triggered when the real-time sales volume per minute is ≥ × When the real-time inventory is below the replenishment trigger warning threshold, an adjustment instruction is directly triggered. When the monitoring module detects that the main trigger condition is met in combination with any auxiliary trigger condition, it generates a standardized shelf adjustment instruction containing the product SKU, real-time inventory, threshold value, and corresponding product-related data.

[0017] Furthermore, in the monitoring module, when generating standardized shelf adjustment instructions, the collected real-time multi-source data of goods and the retrieved replenishment trigger warning threshold are first double-verified. After confirming that there are no abnormalities in the data, instructions are generated according to the detected trigger condition type. The instruction generation follows a fixed process: first, the core identification information of the goods is entered, the goods SKU, real-time inventory value, replenishment trigger warning threshold and the ratio between the two are determined, then the trigger condition type is marked, distinguishing between regular and emergency trigger scenarios. Regular instructions are generated for the corresponding first-level auxiliary trigger conditions, and basic instruction elements are entered; emergency instructions are generated for the corresponding second-level auxiliary trigger conditions and sales rate trigger conditions, and additional information such as inventory urgency level and sales rate details is added. At the same time, a unique instruction identifier, generation time and execution time requirements are uniformly added, and all information is standardized.

[0018] Furthermore, the monitoring module generates standardized shelf adjustment instructions, which are divided into regular adjustment instructions and emergency adjustment instructions. Regular adjustment instructions are generated based on first-level auxiliary trigger conditions, while emergency adjustment instructions are generated based on second-level auxiliary trigger conditions and sales rate trigger conditions. Regular adjustment instructions include a unique instruction code, product SKU, real-time inventory, Ts threshold, execution action, and execution time. The execution action is set to lower the virtual shelf display order of the product, load three substitute products of the same category, and add a yellow inventory warning label to the display area. The execution time after the instruction is issued is preset to within 1 second. Emergency adjustment instructions add sales rate and inventory urgency level information. The execution requirement is to remove the product from the top 10 positions of the virtual shelf and push a red inventory urgency prompt in the product list pop-up window in the live broadcast room. The execution time after the instruction is issued is preset to within 0.5 seconds. Both types of instructions are accompanied by a unique identifier and generation time, and are simultaneously pushed to the virtual shelf control module and their execution traceability information is retained.

[0019] Furthermore, in the adjustment module, the tiered adjustment operation matches the regular adjustment instructions and emergency adjustment instructions issued by the monitoring module, and carries out adjustments in a tiered manner: for regular adjustments corresponding to the first-level auxiliary trigger conditions, the original display row is moved to the middle and back row areas, retaining the basic purchase entrance and not affecting the display order of other products; for emergency adjustments corresponding to the second-level auxiliary trigger conditions and sales rate trigger conditions, the delisted products are temporarily stored in the backend restocking area; at the same time, alternative products are selected according to product category and audience matching degree, and assigned corresponding display priorities according to suitability, and pushed to the original product display position and the live broadcast room recommendation bar, while the pre-sale links of the same out-of-stock products are pushed with medium display priority.

[0020] Furthermore, the complete adjustment record generated in the adjustment module is a structured and traceable record, associated with a unique identifier for the instruction issued by the monitoring module. The record content includes basic information, execution information, and result information. The basic information includes the instruction receiving time, product SKU, category, original display priority, specific conditions for triggering the adjustment and corresponding thresholds, and real-time inventory data. The execution information includes the adjustment range of the display priority, the original display position and the new display position, the selected alternative product SKUs and their compatibility ranking, the generation time of the pre-sale link for the same product, and the preset validity period. The result information includes the total execution time of the adjustment operation, the instruction execution status, and the link push channel.

[0021] Furthermore, the visualization module acquires end-to-end data including: supply chain inventory levels, inventory allocation progress, replenishment schedule, supplier capacity and raw material supply status; real-time inventory levels, picking progress, and packaging completion volume at the warehousing end; and pickup nodes, trunk transportation routes, last-mile delivery timeliness, and delivery success rate at the delivery end. It establishes real-time data interaction with each node in the supply chain and the warehousing management end through API interfaces, captures dynamic data of the entire delivery process nodes based on the logistics information platform, and simultaneously receives replenishment plans, capacity adjustments, and raw material supply warning data proactively uploaded by suppliers.

[0022] Compared with existing technologies, this intelligent virtual shelf construction system for live-streaming e-commerce has the following beneficial effects: First, this invention integrates multi-source data from live-streaming e-commerce across the entire domain, forms a structured parameter set through standardized preprocessing, relies on dynamic inventory demand prediction and replenishment trigger threshold related algorithms, and combines multi-dimensional factors such as live-streaming scenarios and sales trends to complete inventory demand calculation and early warning threshold setting. Then, through a monitoring module, it compares real-time product data with early warning thresholds in real time, constructs a detection mechanism that superimposes main trigger conditions and multiple auxiliary trigger conditions, and realizes the accurate generation of shelf adjustment instructions. This intelligent operation mode with multi-module collaboration can capture sales fluctuations and inventory changes in real time, trigger appropriate adjustment operations in a timely manner, effectively avoid sales losses caused by inventory backlog or stockouts, and at the same time ensure that the virtual shelf display is highly consistent with market demand, improve the accuracy of product exposure and conversion efficiency, provide efficient inventory and shelf management support for live-streaming e-commerce operations, and make shelf adjustments more in line with the dynamic changes of the live-streaming scenario, reducing the lag and error caused by manual intervention.

[0023] Second, this invention implements a tiered adjustment strategy based on different triggering conditions to accurately match the shelf optimization needs of both routine and emergency scenarios. It simultaneously selects highly suitable alternative products and pushes display links, generating pre-sale links for the same out-of-stock items. This ensures the continuity of shelf display while providing users with diverse choices and reducing potential customer churn. Furthermore, the visualization module connects to the external supply chain to obtain full-chain data, combining it with shelf adjustment records to achieve real-time display of supply chain information. This allows operators and users to clearly understand key information such as product inventory and replenishment progress. The structured and traceable adjustment records also provide data support for subsequent operational optimization. This design, which balances user experience and operational efficiency, promotes collaborative linkage among all links in the supply chain, improves the overall smoothness and controllability of live-streaming e-commerce operations, enhances the flexibility and targeting of shelf display, and helps maximize sales opportunities during live-streaming.

[0024] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0026] Figure 1 A flowchart of a smart virtual shelf construction system for live-streaming e-commerce;

[0027] Figure 2 A data transmission diagram of a smart virtual shelf building system for live-streaming e-commerce.

[0028] Figure 3 This is a schematic diagram of data transmission for the adjustment module of the present invention. Detailed Implementation

[0029] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0030] Example 1: Implementation of a Daily Beauty Makeup Livestream Scenario

[0031] Beauty brand livestreamers conduct regular evening beauty-focused livestreams. In this scenario, the livestreamer's traffic is stable, and the products mainly consist of popular lipsticks, foundations, and related beauty tools. It's crucial to balance regular sales with dynamic inventory management. This invention's system is deeply involved in the entire livestream operation, precisely implementing the following complete process:

[0032] The data collection module was the first to start operating, establishing a real-time and efficient data connection channel with the e-commerce platform's sales end, inventory management end, product information end, and live streaming operation end through a stable API interface. At the same time, it batch-pulled historical sales data of the live streamer's beauty special sessions in the same time period over the past 30 days from the platform's database, realizing multi-dimensional, full-scenario, and full-domain data collection. The specific data collected includes real-time clicks, add-to-cart counts, sales volume, and corresponding growth rates for various lipsticks, foundations, and beauty tools sold during the live stream. This data directly reflects consumers' current interest in and willingness to purchase different products, providing immediate data for assessing subsequent inventory needs. It also includes the actual inventory levels, occupied inventory, and available inventory in online and offline warehouses, providing a comprehensive understanding of the current inventory situation and preventing decision-making biases due to incomplete inventory data. Furthermore, it includes recent daily and monthly sales data, as well as category-related sales data, clearly presenting long-term sales patterns and inter-category relationships, helping to uncover potential consumer demand. Basic attribute data such as product category, specifications, and supply chain replenishment cycles provide crucial references for inventory demand calculations and replenishment planning. Finally, it includes scenario-specific data such as the live stream time, the streamer's traffic level, and small-amount discount promotions, accurately adapting to sales differences across various live stream scenarios. After data collection, the system categorizes and integrates data according to data type, removing redundant and invalid information. It then performs standardized preprocessing on the raw data, unifying data from different formats and sources into a structured parameter set. This ensures data consistency and usability, laying a solid foundation for accurate calculations in subsequent modules. Simultaneously, the structured parameter set is transmitted to the prediction module in real time. Figure 1 As shown.

[0033] After receiving the structured parameter set transmitted by the acquisition module, the prediction module immediately starts the core algorithm calculation to scientifically set inventory demand and early warning thresholds based on data. First, the dynamic library demand prediction algorithm is executed; its mathematical expression is:

[0034] in, Dynamic inventory demand forecast; The basic inventory demand for goods based on historical sales trends; This is the real-time sales fluctuation coefficient; The potential coefficient for a product to become a bestseller; Adjustment coefficients for live streaming scenarios; To achieve shared inventory deductions across all channels, the system fully integrates historical sales trends to form the basic inventory demand, incorporates real-time sales fluctuations in the live stream, accurately captures sudden consumer surges or demand drops, and considers the potential of each product to become a bestseller during the live stream. It then adjusts the scenario adaptation coefficient based on the live stream's time period characteristics, the streamer's traffic level, and small-amount discount promotions. Finally, it deducts the shared deduction amount formed by allocable inventory across all channels. Through a comprehensive balancing of multiple factors, the system calculates the dynamic inventory demand for each beauty product, ensuring that the inventory demand forecast not only aligns with historical patterns but also flexibly responds to real-time changes during the live stream, effectively avoiding prediction bias caused by relying on a single data point. Subsequently, the prediction module continues to execute the replenishment trigger threshold algorithm, the mathematical expression of which is:

[0035] in, The alert threshold is set for when goods are replenished. The dynamic inventory demand forecast value for goods is obtained from the dynamic inventory demand forecasting algorithm; For supply chain replenishment cycle coefficient; Adjust the response latency factor for the virtual shelf; This is the minimum inventory level required to display a product. The inventory safety redundancy coefficient takes into account the length of the supply chain replenishment cycle, ensuring that the set thresholds allow sufficient time for replenishment, while also taking into account potential delays in the virtual shelf adjustment response process to avoid impacting sales due to untimely adjustments. At the same time, it takes into account the minimum display inventory base of the products to ensure the basic needs of shelf display, and adds an inventory safety redundancy coefficient to cope with sudden demand fluctuations. Through the comprehensive integration of multiple factors, accurate and reasonable replenishment trigger warning thresholds are set for each product to ensure that inventory warnings neither cause unnecessary adjustments in advance nor lead to stockout risks due to delays. The set warning thresholds are then transmitted to the monitoring module in real time.

[0036] The monitoring module operates at high sensitivity, receiving real-time multi-source product data pushed by the acquisition module. Simultaneously, it frequently retrieves and compares the replenishment trigger warning thresholds set by the prediction module to ensure no inventory fluctuation signals are missed. When the real-time inventory of a popular lipstick shade falls below the warning threshold for a preset duration of 5 seconds, and the ratio of real-time inventory to the warning threshold is 0.7, fully meeting both the primary and secondary trigger conditions, the monitoring module does not directly generate an instruction. Instead, it first performs dual verification of the collected real-time multi-source product data and the retrieved replenishment trigger warning thresholds. This cross-checking eliminates interference factors such as data transmission errors and abnormal fluctuations, ensuring the authenticity and accuracy of the data. After confirming the data is normal, a standardized instruction generation process is initiated based on the detected trigger condition type. First, the core product identification information is accurately entered, clearly specifying the SKU of the popular lipstick shade, the real-time inventory value, the replenishment trigger warning threshold, and the ratio between the two. Then, the trigger condition type is clearly labeled, distinguishing it from regular trigger scenarios, and subsequently, a regular shelf adjustment instruction is generated. The instruction includes a unique code to ensure its uniqueness and traceability; a complete product SKU to ensure the accuracy of the adjustment target; real-time inventory and early warning threshold data to provide a reference for the adjustment operation; clear execution actions to ensure the direction of the adjustment; and a 1-second execution time requirement to ensure the timeliness of the adjustment and avoid a decline in user experience due to delays. After the instruction is generated, it is pushed to the adjustment module simultaneously and the execution traceability information is retained for subsequent verification.

[0037] Upon receiving a routine shelf adjustment instruction, the adjustment module immediately activates a tiered adjustment mechanism, precisely executing the adjustment operations according to the instruction requirements. For this popular lipstick shade, its virtual shelf display position is adjusted from the front row to the middle or back row. This reduces the exposure of out-of-stock items while retaining a basic purchase entry point, meeting the needs of users who still intend to buy, and maintaining the normal display order of other beauty products, thus preserving the overall rationality of the shelf layout. Simultaneously, based on core dimensions such as product category attributes and target audience matching, the system quickly selects three suitable alternative lipsticks of the same price range and color family from the product library. These are sorted by suitability from high to low and assigned corresponding display priorities, immediately pushed to the original product display position and the live stream recommendation section, allowing users to quickly find alternative options while browsing and reducing user loss due to unavailable desired products. Simultaneously, a pre-sale link for the same out-of-stock lipstick shade is generated and pushed to a prominent position in the live stream product list with medium display priority, providing a clear purchase channel for users waiting for restocking and locking in potential consumer demand. After all adjustments are completed, the system automatically generates a structured and traceable adjustment record. The record comprehensively covers basic information, execution information, and result information: Basic information includes the instruction receipt time, product SKU, category, original display priority, specific conditions triggering the adjustment and corresponding thresholds, and real-time inventory data, clearly reconstructing the adjustment background; Execution information covers the adjustment range of display priority, original and new display positions, selected alternative product SKUs and their suitability ranking, the generation time and preset validity period of the pre-sale link for the same product, recording the adjustment process in detail; Result information includes the total execution time of the adjustment operation, instruction execution status, and link push channel, intuitively presenting the adjustment effect and providing complete data support for subsequent operational optimization, such as... Figure 3 As shown.

[0038] The visualization module receives relevant data on the popular lipstick shade simultaneously from the monitoring module as it generates shelf adjustment instructions, and then initiates end-to-end data integration and display. Through a stable API interface, it establishes real-time data interaction with various nodes in the supply chain and warehouse management, accurately acquiring key supply chain data such as remaining stock, inventory allocation progress, replenishment schedule, and supplier capacity for the lipstick shade, clearly understanding the overall progress and feasibility of replenishment. Simultaneously, it collects picking progress data from the warehouse to understand the preparation status before goods are shipped. Relying on the logistics information platform, it captures dynamic data such as transportation trajectories, last-mile delivery timeliness, and delivery success rate from the delivery end, comprehensively controlling the entire process from warehouse to user. At the same time, it receives replenishment plans, capacity adjustments, and raw material supply warning data proactively uploaded by suppliers, anticipating potential replenishment delay risks in advance. By deeply integrating this end-to-end data with virtual shelf adjustment records, the corresponding product display areas on the virtual shelves are presented in a clear and intuitive way in real time. This allows operators to quickly grasp product inventory and replenishment dynamics, and adjust live-streaming promotion strategies in a timely manner. Consumers can clearly understand the replenishment progress and purchase channels for out-of-stock items, reducing the frequency of customer service inquiries and improving the shopping experience. Simultaneously, the visualization module feeds back data update records to the monitoring module in real time, forming a data loop that ensures information synchronization across modules and provides data support for any subsequent adjustments.

[0039] Throughout the entire daily beauty-themed live stream, all modules of the system work together, forming a complete operational loop from data collection to instruction execution and information feedback. This ensures dynamic inventory balance, optimizes the display effect of the virtual shelf, and improves user shopping experience and operational efficiency, perfectly meeting the operational needs of beauty live streams under stable daily traffic.

[0040] In this embodiment, in a typical beauty-focused live-streaming scenario, the system integrates and standardizes multi-source data across the entire domain through the data collection module, providing a reliable foundation for subsequent calculations. The prediction module uses a dynamic inventory demand prediction algorithm and a replenishment trigger threshold algorithm to accurately set inventory demand and early warning thresholds. The monitoring module compares data in real time and generates adaptation instructions after double verification. The adjustment module adjusts the product shelves in stages, pushes alternative products and pre-sale links, and retains traceable records. The visualization module integrates and displays real-time data from the entire supply chain and provides feedback. The collaborative efforts of all modules throughout the process achieve dynamic adaptation between inventory and product shelves while ensuring sales continuity and user experience, providing efficient and accurate operational support for beauty live-streaming with stable daily traffic.

[0041] Example 2: Implementation of a Live Streaming Scenarios for the 618 Shopping Festival Apparel Sales

[0042] Apparel brands participating in the 618 Grand Promotion's overnight livestreaming event are characterized by high-level streamer traffic and significant promotional activities such as large discounts and limited-time flash sales. The products primarily consist of new seasonal dresses and casual pants, resulting in rapid sales and volatile inventory. A quick response to inventory changes is crucial to prevent sales interruptions. This invention's system addresses the unique characteristics of this promotional scenario with the following refined process for efficient operation: The data collection module completes all preparatory work in advance, establishing a stable and high-speed real-time data transmission channel with the e-commerce platform's sales, inventory management, product information, and livestreaming operations ends via a high-speed API interface. This ensures uninterrupted and timely data transmission even under the high traffic and concurrency of the promotion. Simultaneously, it batch-fetches historical sales trend data for the brand's apparel products from the past three promotional periods. This proven historical data accurately reflects consumer purchasing habits, product sales patterns, and inventory fluctuation characteristics during the promotion, providing crucial reference for inventory management during this promotional livestream. After the live stream starts, the data collection module fully initiates comprehensive data collection, covering real-time clicks, add-to-cart volumes, sales volumes, and frequently updated real-time growth rates for various styles of dresses and casual pants within the live stream. This high-frequency data update allows for timely capture of sudden consumer hotspots during promotional periods, quickly responding to changes in consumer demand. The system also collects omnichannel inventory data, including actual inventory, occupied inventory, and available inventory across online and offline warehouses and various distribution channels, providing a comprehensive understanding of inventory distribution across different channels. This data supports cross-channel inventory allocation and sharing, preventing waste caused by inventory shortages in one channel while other channels remain idle. Recent daily and monthly sales data, as well as category-related sales data, help assess the sales potential of products during promotional periods, providing supplementary information for inventory demand forecasting. Basic attribute data such as product category, specifications, and supply chain replenishment cycles clarify the replenishment capacity and cycle limitations of products, making inventory planning more feasible. Finally, the system incorporates live stream scenario characteristics such as overnight periods during the 618 promotion, high-traffic levels of the streamer, and limited-time flash sales, accurately adapting to the high-traffic, high-conversion characteristics of promotional scenarios and providing a foundation for subsequent algorithmic calculations. After data collection, the system categorizes and integrates the data according to data type, performs standardized preprocessing operations on the raw data to remove abnormal data interference that may occur during peak sales periods, and unifies data from different sources and formats into a structured parameter set. This ensures the accuracy, consistency, and usability of the data, providing a solid guarantee for the efficient operation of subsequent modules. Simultaneously, the structured parameter set is synchronized to the prediction module in real time at millisecond speeds. Figure 2 As shown.

[0043] After receiving the structured parameter set, the prediction module immediately initiates algorithm calculations adapted to the promotional scenario, ensuring that inventory demand and warning thresholds accurately match the high-intensity sales pace during the promotion. First, the dynamic inventory demand prediction algorithm is activated, fully referencing the basic inventory demand formed by the sales trends of products during historical promotional cycles. Combined with real-time sales fluctuations during the current promotion's live stream, it quickly captures sudden sales peaks caused by activities such as flash sales, accurately assessing the potential of new dresses, casual pants, and other items to become bestsellers during this promotion. Then, based on scenario characteristics such as the 618 promotion's overnight duration, high-traffic levels of live streamers, large discounts, and flash sales, the module adjusts the scenario correction coefficients to fully adapt to sales differences under the promotional scenario. Finally, it deducts the shared discount amount formed by all-channel available inventory, comprehensively considering multiple factors to complete the dynamic inventory demand calculation for each apparel item. This ensures that the inventory demand prediction results fully consider both the sales surge of the promotion and the reasonable allocation of inventory, avoiding stockouts due to insufficient prediction or inventory backlog due to over-prediction. Subsequently, the prediction module executes a replenishment trigger threshold algorithm. Considering the potential extension of the supply chain replenishment cycle during peak sales periods, it reasonably sets a supply chain replenishment cycle coefficient to allow sufficient replenishment time. Taking into account the potential response delays caused by the heavy workload of virtual shelf adjustments during peak sales periods, it accurately sets a virtual shelf adjustment response delay coefficient. It references the minimum display inventory base for products during peak sales scenarios to ensure the richness of shelf displays. It also incorporates an inventory safety redundancy coefficient adapted to the high volatility of peak sales periods to cope with unexpected demand. Through the scientific allocation of multi-dimensional coefficients, it sets precise replenishment trigger warning thresholds for each apparel product, ensuring that inventory warnings are triggered promptly and accurately under the high-intensity sales rhythm of peak sales periods, buying time for subsequent adjustment operations. The warning thresholds are then immediately transmitted to the monitoring module.

[0044] The monitoring module, designed for the high traffic and high sales rate characteristics of major promotional events, employs a high-frequency data comparison mechanism. It captures real-time multi-source product data pushed by the acquisition module and compares it with the replenishment trigger warning threshold set by the prediction module at the millisecond level, ensuring no subtle inventory fluctuations are missed. When the minute-level real-time sales of a popular casual trousers reach the standard corresponding to the product of the basic inventory demand and the real-time sales fluctuation coefficient, and the real-time inventory value is less than the warning threshold for 5 seconds, while the ratio of real-time inventory to the warning threshold is 0.4, satisfying the primary trigger condition, the secondary auxiliary trigger condition, and the sales rate trigger condition, the monitoring module immediately initiates a dual data verification process. This process cross-checks the real-time sales data, inventory data, and warning threshold to eliminate abnormal data caused by data transmission congestion or instantaneous traffic surges during major promotional events, ensuring the accuracy of the trigger condition judgment. After confirming the data is anomaly-free, the emergency instruction generation process is initiated based on the trigger condition type. First, the core product identification information is accurately entered, specifying the SKU of the casual pants, real-time inventory value, replenishment trigger warning threshold, and the ratio between the two. The trigger condition type is marked as an emergency trigger scenario. Then, an emergency shelf adjustment instruction is generated. This instruction includes a unique code, product SKU, real-time inventory, warning threshold, execution action, and execution time. It also supplements sales rate details and inventory urgency level information, allowing the adjustment module to quickly grasp the urgency level and sales situation of the product. The execution time is set within 0.5 seconds to ensure rapid adjustment during high-conversion periods of major promotions, avoiding the loss of a large number of potential customers due to delayed adjustments. After the instruction is generated, it is simultaneously pushed to the adjustment module, and complete execution traceability information is retained, providing detailed data support for post-promotion review.

[0045] Upon receiving an emergency shelf adjustment instruction, the adjustment module immediately activates the emergency response mechanism and efficiently completes the adjustment operation as required. First, the best-selling casual pants are quickly removed from the top 10 positions of the virtual shelf and temporarily stored in the backend restocking area to prevent users from failing to place orders due to low inventory and remaining in a high-profile position, thus affecting the shopping experience. Simultaneously, a red low inventory warning is displayed in the product list pop-up in the live stream room, prominently informing users of the current inventory status and reducing invalid clicks and inquiries. Subsequently, the system quickly selected five matching casual pants of the same style and price range based on core dimensions such as product category, style, price, and audience relevance. These were then sorted from highest to lowest relevance and given high display priority, immediately pushed to the original product display position, the live stream recommendation bar, and the exclusive pop-up for flash sales. This leveraged the high popularity of the original product to drive the exposure and conversion of the alternative products, maximizing potential sales opportunities. Simultaneously, a pre-sale link for the same out-of-stock casual pants was generated and pushed to the top of the live stream product list with medium display priority, clearly indicating the pre-sale shipping time. This provided a clear purchase channel for users with strong purchasing intentions, locking in user demand and reducing customer churn. After all adjustments are completed, the system generates a structured and traceable adjustment record. The record is comprehensive and detailed: basic information includes the instruction receipt time, product SKU, category, original display priority, details of the specific conditions that triggered the adjustment and the corresponding threshold and real-time inventory data, clearly restoring the background and triggering reasons for the adjustment; execution information covers the adjustment range of display priority, original and new display positions, selected alternative product SKUs and their suitability ranking, the generation time and preset validity period of the pre-sale link for the same product, fully presenting the specific execution process of the adjustment; result information includes the total execution time of the adjustment operation, instruction execution status, and link push channel, intuitively reflecting the final effect of the adjustment, providing strong data support for real-time operation optimization and subsequent activity planning during the promotion period.

[0046] The visualization module operates continuously and efficiently during the live promotion. After receiving data related to the casual pants from the monitoring module, it establishes real-time data interaction with various nodes in the supply chain, warehouse management, and logistics information platform through a high-speed API interface. This allows for comprehensive acquisition of the casual pants' remaining stock, real-time progress of inventory allocation, replenishment schedule, supplier capacity, and raw material supply status, providing a clear understanding of the overall replenishment progress and potential risks. It also collects real-time inventory and picking progress data from the warehouse to understand the readiness for outbound shipment. Furthermore, it captures dynamic data from the delivery end, including pickup nodes, trunk transportation routes, last-mile delivery timeliness, and delivery success rate, providing comprehensive control over the product's logistics status. Simultaneously, it receives replenishment plans, capacity adjustments, and raw material supply warnings proactively uploaded by suppliers, allowing for early prediction of potential replenishment delays. By deeply integrating this end-to-end data with virtual shelf adjustment records, the information is displayed in a clear and intuitive real-time manner on the virtual shelves and the side information bar of the live stream. This allows operators to monitor product inventory, replenishment, and logistics dynamics in real time, enabling them to quickly adjust live stream promotion strategies and inventory allocation plans. Consumers can clearly understand the replenishment progress of out-of-stock items and pre-sale delivery times, reducing anxiety caused by information asymmetry and enhancing brand trust and the shopping experience. Simultaneously, the visualization module feeds back data updates to the monitoring module in real time, ensuring smooth information flow across all modules. This provides timely and accurate data support for any secondary adjustments that may occur during the promotion, guaranteeing the continuous and smooth operation of the promotional live stream.

[0047] Throughout the entire 618 promotional overnight live stream, the system successfully handled the drastic inventory fluctuations under high traffic and high sales rates thanks to its efficient module collaboration, rapid response capabilities, and precise adjustment strategies. It ensured the orderly display of the virtual shelves and the continuity of sales, maximized sales potential, and improved operational efficiency and user experience, perfectly adapting to the needs of apparel live stream operations during the promotional period.

[0048] In this embodiment, the system, designed for the high traffic and high sales rate of apparel live streaming during the 618 shopping festival, features a data acquisition module that rapidly connects to and preprocesses data from multiple platforms to meet the data demands of the festival. A prediction module, combining the characteristics of the festival scenario, uses two core algorithms to accurately calculate inventory demand and early warning thresholds. A monitoring module performs high-frequency comparisons and rapid verifications to generate timely emergency instructions. An adjustment module provides emergency response, efficiently completing shelf adjustments, pushing alternative products, and generating pre-sale links. A visualization module displays real-time data across the entire supply chain, aiding in real-time decision-making. The system's rapid response and precise adaptation throughout effectively address the dramatic fluctuations in inventory during the festival, maximizing sales potential and providing stable, efficient, and intelligent operational support for apparel live streaming during peak sales periods.

[0049] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A smart virtual shelf construction system for live-streaming e-commerce, characterized in that, The system includes: Data Acquisition Module: Collects multi-source data related to live e-commerce across the entire domain, and performs standardized preprocessing on the raw data to form a structured parameter set; Prediction module: Receives the structured parameter set from the acquisition module, executes the dynamic inventory demand prediction algorithm and the replenishment trigger threshold algorithm in sequence, completes the calculation of commodity inventory demand and the setting of replenishment trigger warning threshold, and transmits the warning threshold to the monitoring module; Monitoring module: Receives real-time multi-source data on goods from the acquisition module, retrieves replenishment trigger warning thresholds from the prediction module for real-time comparison, and generates standardized shelf adjustment instructions and corresponding product-related data when the real-time inventory of goods meets the trigger conditions. Adjustment module: Receives shelf adjustment trigger instructions and product-related data from the monitoring module, performs hierarchical adjustment operations on the virtual shelf display priority, filters suitable alternative products and pushes their display links, generates and pushes pre-sale links for out-of-stock products, and generates a complete adjustment record after completing all adjustment operations. Visualization module: Receives product-related data from the monitoring module, connects with the external supply chain to obtain full-chain data, combines virtual shelf adjustment records, completes the real-time display of product supply chain information on the virtual shelf, and feeds back data update records to the monitoring module.

2. The intelligent virtual shelf construction system for live-streaming e-commerce according to claim 1, characterized in that, The data collection module collects multi-source data related to live-streaming e-commerce, including real-time sales data such as clicks, add-to-cart volume, sales volume, and real-time growth rates of various indicators for products in the live-streaming room; omnichannel inventory data including actual inventory, occupied inventory, and available inventory in online and offline warehouses; historical sales trend data such as daily and monthly sales and category-related sales data for products in the recent period; basic attribute data of product categories, specifications, and supply chain replenishment cycles; and live-streaming scene feature data configured on the live-streaming time slot and the anchor's operation end. Real-time data connections are established with the e-commerce platform's sales end, inventory management end, product information end, and live-streaming operation end through API interfaces. Simultaneously, historical data is batch-fetched from the platform's database. After collection, the data is categorized and integrated according to data type, and then standardized preprocessing operations are performed on the raw data to form structured parameters.

3. The intelligent virtual shelf construction system for live-streaming e-commerce according to claim 1, characterized in that, In the prediction module, the mathematical expression for the prediction algorithm required by the dynamic library is: in, Dynamic inventory demand forecast; The basic inventory demand for goods based on historical sales trends; This is the real-time sales fluctuation coefficient; The potential coefficient for a product to become a bestseller; Adjustment coefficients for live streaming scenarios; This is a shared deduction amount for inventory across all channels.

4. The intelligent virtual shelf construction system for live-streaming e-commerce according to claim 1, characterized in that, In the prediction module, the mathematical expression for the replenishment trigger threshold algorithm is: in, The alert threshold is set for when goods are replenished. The dynamic inventory demand forecast value for goods is obtained from the dynamic inventory demand forecasting algorithm; For supply chain replenishment cycle coefficient; Adjust the response latency factor for the virtual shelf; This is the minimum inventory level required to display a product. This is the inventory safety redundancy coefficient.

5. The intelligent virtual shelf construction system for live-streaming e-commerce according to claim 1, characterized in that, In the monitoring module, the triggering conditions are a main triggering condition superimposed with multiple auxiliary triggering conditions. The main triggering condition is that the real-time inventory value of the product is less than the replenishment triggering warning threshold and remains so for a preset duration of 5 seconds. The first-level auxiliary triggering condition is that the ratio of the real-time inventory value to the replenishment triggering warning threshold is ≤0.8, triggering a regular adjustment command. The second-level auxiliary triggering condition is that the ratio of the real-time inventory value to the replenishment triggering warning threshold is ≤0.5, triggering an emergency adjustment command. Additionally, a sales rate triggering condition is added, which is triggered when the real-time sales volume per minute is ≥ × When the real-time inventory is below the replenishment trigger warning threshold, an adjustment instruction is directly triggered. When the monitoring module detects that the main trigger condition is met in combination with any auxiliary trigger condition, it generates a standardized shelf adjustment instruction containing the product SKU, real-time inventory, threshold value, and corresponding product-related data.

6. The intelligent virtual shelf construction system for live-streaming e-commerce according to claim 1, characterized in that, In the monitoring module, when generating standardized shelf adjustment instructions, the collected real-time multi-source data of goods and the retrieved replenishment trigger warning threshold are first double-verified. After confirming that there are no abnormalities in the data, instructions are generated according to the detected trigger condition type. The instruction generation follows a fixed process: first, the core identification information of the goods is entered, the goods SKU, real-time inventory value, replenishment trigger warning threshold and the ratio between the two are determined, then the trigger condition type is marked, distinguishing between regular and emergency trigger scenarios, generating regular instructions corresponding to the first-level auxiliary trigger conditions, and entering the basic instruction elements. Emergency instructions are generated based on the corresponding secondary auxiliary trigger conditions and sales rate trigger conditions. Additional information such as inventory urgency level and sales rate details are added. At the same time, a unique identifier for the instruction, the generation time, and the execution time requirements are added to standardize all information.

7. The intelligent virtual shelf construction system for live-streaming e-commerce according to claim 1, characterized in that, In the monitoring module, the generated standardized shelf adjustment instructions are divided into regular adjustment instructions and emergency adjustment instructions. Regular adjustment instructions are generated by matching first-level auxiliary trigger conditions, while emergency adjustment instructions are generated by matching second-level auxiliary trigger conditions and sales rate trigger conditions. Regular adjustment instructions include a unique instruction code, product SKU, real-time inventory, Ts threshold, execution action, and execution time. The execution action is set to lower the virtual shelf display order of the product, load 3 substitute products of the same category, and add a yellow inventory warning icon to the display area. The execution time after the instruction is issued is preset to within 1 second. Emergency adjustment instructions add sales rate and inventory urgency level information. The execution requirement is to remove the product from the top 10 positions of the virtual shelf and push a red inventory urgency prompt in the product list pop-up window in the live broadcast room. The execution time after the instruction is issued is preset to within 0.5 seconds. Both types of instructions are accompanied by a unique identifier and generation time.

8. The intelligent virtual shelf construction system for live-streaming e-commerce according to claim 1, characterized in that, In the adjustment module, the tiered adjustment operation matches the regular adjustment instructions and emergency adjustment instructions issued by the monitoring module, and carries out adjustments in a tiered manner: for regular adjustments corresponding to the first-level auxiliary trigger conditions, the original display row is moved to the middle and back row areas, retaining the basic purchase entrance and not affecting the display order of other products; for emergency adjustments corresponding to the second-level auxiliary trigger conditions and sales rate trigger conditions, the delisted products are temporarily stored in the back-end restocking area. At the same time, alternative products are selected based on product category and audience matching, and they are sorted by suitability and given corresponding display priorities. They are then pushed to the original product display position and the live broadcast room recommendation bar. Meanwhile, the pre-sale links of the same products that are out of stock are pushed with medium display priority.

9. The intelligent virtual shelf construction system for live-streaming e-commerce according to claim 1, characterized in that, The adjustment module generates a complete, structured, and traceable adjustment record, which is associated with a unique identifier for the instructions issued by the monitoring module. The record content includes basic information, execution information, and result information. The basic information includes the instruction receiving time, product SKU, category, original display priority, specific conditions for triggering the adjustment, corresponding thresholds, and real-time inventory data. The execution information includes the adjustment range of the display priority, the original and new display positions, the selected alternative product SKUs and their compatibility ranking, the generation time of the pre-sale link for the same product, and the preset validity period. The result information includes the total execution time of the adjustment operation, the instruction execution status, and the link push channel.

10. A smart virtual shelf building system for live-streaming e-commerce according to claim 1, characterized in that, The visualization module acquires end-to-end data including: supply chain inventory levels, inventory allocation progress, replenishment schedule, supplier capacity and raw material supply status; real-time inventory levels, picking progress, and packaging completion volume at the warehousing end; and pickup nodes, trunk transportation routes, last-mile delivery timeliness, and delivery success rate at the delivery end. It establishes real-time data interaction with each node in the supply chain and the warehousing management end through API interfaces, captures dynamic data of the entire delivery process nodes based on the logistics information platform, and simultaneously receives replenishment plans, capacity adjustments, and raw material supply warning data proactively uploaded by suppliers.

Images