Multi-channel third-party site promotion method and system with real-time effect feedback

Abstract

The present application relates to the technical field of website information analysis, and discloses a multi-channel third-party site promotion method and system with real-time effect feedback, which comprises collecting multi-channel promotion data and preprocessing, quantifying the dynamic contribution weight of each channel in the user conversion path, restoring the cross-channel user behavior time sequence path, jointly optimizing the keyword bidding, creative material, delivery time period and budget allocation, managing the third-party sites according to the conversion efficiency, and realizing the closed-loop optimization of multi-channel cooperation through execution verification, difference analysis and obstacle factor identification. Thus, the promotion efficiency and the return on investment can be improved.

Description

Technical Field

[0001] This invention relates to the field of Internet site information analysis technology, and more specifically, to a method and system for promoting multi-channel third-party sites with real-time effect feedback. Background Technology

[0002] With the widespread application of multi-channel promotion in the digital society, especially in complex marketing scenarios involving multiple third-party sites such as e-commerce platforms, content service platforms, and online education platforms, the demand for real-time feedback on promotion results and cross-channel collaborative optimization is becoming increasingly urgent.

[0003] While existing multi-channel promotion technologies have made some progress in multi-channel advertising, basic conversion tracking, and single-channel performance evaluation, they have not yet systematically solved core issues such as accurate restoration of cross-channel user behavior paths, dynamic collaborative optimization of multi-dimensional promotion elements (such as keyword performance, creative click-through rate, bidding efficiency, and ad placement time), and adaptive adjustment of strategies based on real-time feedback. They cannot fully meet the requirements for efficient resource allocation and stable improvement of conversion efficiency in complex marketing scenarios.

[0004] Therefore, how to build a promotion method and system that can integrate multi-channel data, quantify the contribution of each channel in the conversion path, achieve dynamic joint optimization of promotion strategies, and support the hierarchical management and collaborative execution of third-party sites has become an urgent technical problem to be solved. Summary of the Invention

[0005] This invention provides a method and system for promoting multi-channel third-party sites with real-time effect feedback, solving the technical problem in the prior art that it is difficult to build a system with real-time effect feedback, cross-channel data fusion and dynamic strategy optimization capabilities.

[0006] This invention provides a method and system for promoting third-party websites through multiple channels with real-time performance feedback, including: The first aspect involves a multi-channel third-party website promotion method with real-time performance feedback, including the following steps: Collect promotional data from third-party websites across multiple channels and perform preprocessing operations to obtain a promotional dataset in a unified format; Based on the aforementioned promotion dataset, the contribution of multiple channels in the user conversion path is quantitatively evaluated to obtain the effect evaluation results. Based on the promotion dataset and the effect evaluation results, the promotion behaviors, user click streams, conversion events and channel context information of different channels are correlated in time and restored through path reconstruction to obtain a cross-channel associated dataset. Based on the performance evaluation results and cross-channel related datasets, and combined with the preset optimization objective function, the keyword bidding strategy, creative material combination, ad placement time configuration and budget allocation ratio are jointly optimized to obtain the promotion execution strategy; The multiple third-party sites in the aforementioned promotion execution strategy are managed in layers according to their conversion efficiency characteristics, resulting in management outcomes. Based on the management results, obtain multi-channel collaborative promotion optimization strategies; The promotion and optimization strategy is a multi-process collaborative optimization and iteration strategy scheme that involves hierarchical management, site association and combination, actual execution verification, difference analysis and identification and elimination of obstacles.

[0007] Furthermore, based on the aforementioned promotion dataset, the contribution of multiple channels in the user conversion path is quantitatively evaluated, including: The promotional data includes exposure data, click data, conversion data, and user behavior trajectory data; Based on user behavior trajectory data, identify the complete user behavior path from first contact to final conversion; Arrange the user behavior paths in chronological order to obtain a user conversion path sequence; Based on the user conversion path sequence, the touchpoint information in each conversion path is extracted, and each touchpoint is weighted. According to the position of the touchpoint in the path, the time interval between the touchpoint and the final conversion, and the interaction depth of the touchpoint, the contribution weight value of each touchpoint to the final conversion is obtained. Based on the aforementioned contribution weight values, the contribution weight values ​​of each channel in all conversion paths are weighted and summarized. The contribution weight values ​​of all touchpoints belonging to the same channel are accumulated to generate the original performance score of each channel. Based on the original performance scores, the performance scores of each channel are mapped to a numerical range so that the performance scores fall within a preset standard range, resulting in standardized performance evaluation results.

[0008] Furthermore, based on the promotion dataset and the effect evaluation results, the promotion behaviors, user clickstreams, conversion events, and channel context information from different channels are subjected to time-series correlation and path reconstruction processing, including: Extract promotion behavior logs for each channel from the aforementioned promotion dataset; The promotion behavior logs are parsed to obtain promotion behavior feature data including ad placement time, placement location, bid amount and material type, and a unique promotion behavior identifier is assigned to each promotion behavior record; Extract user clickstream data associated with promotional behavior from the promotion dataset, parse the user clickstream data to obtain user interaction feature data including click time, click source, visited page and dwell time; Extract standardized user identifiers from the user clickstream data; and establish an association mapping between promotion behavior identifiers and user identifiers. Based on the aforementioned association mapping, extract the conversion event data associated with the user identifier; The conversion event data is parsed to obtain conversion result feature data including conversion time, conversion type, conversion value, and conversion path; Construct a three-dimensional data set based on promotion behavior characteristic data, user interaction characteristic data, and conversion event data; Based on the aforementioned three-data set, the promotion behavior logs, user clickstream data and conversion event data are sorted and matched in time sequence with timestamp as the key field to identify the behavior sequence of the same user at different time points and generate a time sequence association data table. Based on the aforementioned time-series association data table, combined with user identifiers and preset time window rules, user behavior records scattered across different channels are aggregated to reconstruct the complete behavioral path of users from ad exposure, click access to final conversion, resulting in a cross-channel association dataset containing cross-channel behavioral sequences and path features.

[0009] Furthermore, the promotion and implementation strategy includes: Based on the cross-channel associated dataset, historical promotion feature data such as keyword bidding, creative click-through rate, time period traffic density, and budget consumption rate are extracted to construct a multi-dimensional decision space containing multiple decision variables. The value range and change step size of each decision variable are determined to obtain the decision space model. Based on the decision space model, a promotion optimization objective function is set, with maximizing the return on investment or minimizing the conversion cost as the optimization objective. The mathematical expression and calculation rules of the objective function are defined to obtain the optimization objective function. Based on the objective function, the performance evaluation results are transformed into optimization constraints; Among them, for channels whose performance evaluation results are lower than the preset threshold, the upper limit of the budget allocation ratio is restricted; for channels whose performance evaluation results are higher than the preset threshold, the budget allocation ratio is increased first, resulting in a set of constraint conditions containing multiple constraint rules. Based on the decision space model, the objective function and the set of constraints, iterative optimization is performed in the multidimensional decision space to search for a combination of decision variables that satisfies the constraints and makes the objective function optimal, thereby obtaining the optimal decision solution. Based on the optimal decision solution, the decision variable values ​​are mapped to promotion operation instructions, and keyword bidding adjustment range, material priority sorting list, time period delivery switch status configuration and channel budget allocation scheme are generated by parsing, resulting in a promotion execution strategy containing multiple executable strategy parameters.

[0010] Furthermore, the multiple third-party sites in the aforementioned promotion execution strategy are managed in layers according to their conversion efficiency characteristics, including: Information about each third-party site is extracted from the promotion execution strategy, and conversion performance characteristics of each site, including conversion rate, cost per click, return on investment and user quality score, are calculated based on the cross-channel associated dataset. The conversion efficiency features are converted into multi-dimensional feature vectors to obtain the feature dataset of site efficiency; Based on the feature dataset, a performance similarity analysis is performed on each third-party site to obtain the feature distance between each site. Sites with performance feature similarity higher than a preset threshold are grouped into the same level, generating a site hierarchical structure containing multiple levels, and obtaining the site hierarchical result. Based on the site stratification results, differentiated budget allocation strategies and delivery control strategies are set for different strata. For sites in the high-performance strata, the budget allocation weight and delivery frequency are increased, while for sites in the low-performance strata, the budget allocation weight is reduced and the monitoring frequency is increased, resulting in a stratified differentiated management strategy. Based on the site hierarchical results and the hierarchical differentiated management strategy, a management data structure is constructed that includes site hierarchical affiliation, hierarchical performance indicator ranges, and hierarchical management strategy parameters to obtain management results.

[0011] Furthermore, based on the management results, a multi-channel collaborative promotion optimization strategy is obtained, including: The management results will associate and combine multiple third-party sites to obtain multiple site association groups; The promotion execution strategy is sent to the corresponding channel interface to drive the related groups of each site to execute promotion tasks, and the actual execution effect data of each related group of the site is collected at the same time to obtain verification data of the promotion effect; Based on the verification data, a difference analysis was conducted on the actual conversion performance of each site's associated group and the expected target to obtain the results of the performance differences. Based on the results of the performance differences, identify and eliminate obstacles that affect promotion conversion, and use these obstacles as optimization feedback inputs to output promotion optimization strategies.

[0012] Furthermore, the management results will associate and combine multiple third-party sites separately to obtain multiple site association groups, including: Based on the management results, the hierarchical affiliation information and performance indicator data of each third-party site are extracted, and the historical promotion behavior data and user conversion data of each site are extracted from the cross-channel association dataset to construct a site feature database containing site identifiers, hierarchical identifiers, performance characteristics and behavioral characteristics. Based on the site feature database, obtain the site attribute features of each third-party site, including business type, user group, promotion target, geographical distribution and channel affiliation, construct site attribute feature vector, and calculate the attribute similarity between any two sites to obtain the site attribute similarity matrix. Based on the site attribute similarity matrix and the site feature database, the promotion synergy effect between sites is analyzed. By calculating the user overlap, conversion path intersection and promotion time complementarity between sites, site pairs with synergistic promotion effects are identified, and a synergistic effect score is calculated for each site pair to obtain the site synergy effect matrix. Based on the site synergy effect matrix, a synergy effect score threshold is set, and site combinations with synergy effect scores higher than the threshold are selected. The high-scoring site combinations are then clustered to associate the sites within the same group, thus obtaining preliminary site association groups. Based on the initial site association groups, the sites within each group are verified to ensure consistency of promotion goals and feasibility of resource allocation. Sites that do not meet the association conditions are removed, forming stable site collaborative promotion units. Each site collaborative promotion unit is assigned a unique unit identifier and a site ranking sequence within the group, resulting in multiple site association groups.

[0013] Furthermore, the promotion execution strategy is distributed to the corresponding channel interfaces to drive the promotion tasks of each site's associated groups, and the actual execution effect data of each site's associated groups is collected synchronously to obtain verification data of the promotion effect, including: Based on the aforementioned promotion optimization strategy, extract the set of strategy parameters that need to be issued; The strategy parameter set includes keyword bidding adjustment instructions, creative material optimization configuration, ad placement start and stop rules, and channel budget dynamic allocation parameters. The strategy parameters are converted according to the interface protocols of each channel to obtain a channel-compatible strategy parameter package. Based on the strategy parameter package, the strategy parameters are pushed to the advertising delivery system of each channel through the parameter interface, and the parameter reception confirmation signal returned by each channel is received to establish a confirmation record of successful strategy delivery. Based on the confirmation records, the advertising delivery systems of each channel dynamically adjust keyword bids, material display order, delivery time periods and budget allocation according to the received strategy parameters, and generate strategy execution status data to obtain strategy execution feedback. Based on the strategy execution feedback, during the execution of the promotion task, exposure data, click data, conversion data and cost data of each site-related group are collected in real time through the data collection interface, and each collected data is marked with a site-related group identifier and a collection timestamp to obtain the original execution effect data. Based on the original execution effect data, data is aggregated according to the site association group identifier, and the data of the same site association group is summarized and statistically analyzed. The total exposure, total clicks, total conversions and total cost of each site association group are calculated to generate the actual execution effect data of each site association group and obtain the verification data of the promotion effect.

[0014] Furthermore, based on the verification data, the actual conversion performance data of each site-related group is extracted, including the actual conversion rate, actual conversion cost, and actual return on investment. The expected conversion targets corresponding to each site-related group are extracted from the promotion optimization strategy, including the target conversion rate, target conversion cost, and target return on investment, to obtain a comparison dataset. Based on the comparison dataset, the difference value and difference rate between the actual conversion performance and the expected target are calculated. The absolute values ​​of the difference rates are sorted, and site association groups with difference rates exceeding a preset threshold are identified to obtain a list of abnormal site association groups. Based on the list of abnormal site association groups, a multi-dimensional cause analysis is conducted on each site association group in the list. Key influencing factors leading to the differences are identified from dimensions such as keyword matching degree, material attractiveness, accuracy of delivery time, rationality of budget allocation and channel competition intensity, and a difference cause analysis report is obtained. Based on the aforementioned analysis report of the reasons for the differences, potential obstacles affecting promotion conversion are identified, including low keyword matching, insufficient attractiveness of creative materials, inaccurate timing of ad placement, unreasonable budget allocation, and intense channel competition. The importance of these obstacles is assessed, the weight of each obstacle on the conversion effect is calculated, and the key obstacles with the greatest impact weight are selected to obtain a priority list of obstacles. Based on the priority list of obstacles, the key obstacles ranked at the top are transformed into quantifiable optimization feedback inputs. Keyword optimization suggestions are generated for the problem of low keyword matching, material replacement suggestions are generated for the problem of insufficient material attractiveness, time period adjustment suggestions are generated for the problem of inaccurate delivery time, and budget reallocation suggestions are generated for the problem of unreasonable budget allocation, thus obtaining an optimization feedback instruction set. Based on the optimized feedback instruction set, the keyword bidding strategy, creative material combination, ad placement time configuration and budget allocation ratio are adjusted, the corresponding parameters in the promotion optimization strategy are updated, and a new round of optimized promotion optimization strategy is generated.

[0015] Secondly, a multi-channel third-party website promotion system with real-time performance feedback includes: Data acquisition module: Used to collect promotional data from third-party websites across multiple channels and perform preprocessing operations to obtain a promotional dataset in a unified format; Performance evaluation module: Based on the promotion dataset, this module quantifies the contribution of multiple channels in the user conversion path to obtain performance evaluation results. Association module: Based on the promotion dataset and the effect evaluation results, it performs time-series association and path reconstruction processing on the promotion behavior, user click flow, conversion events and channel context information of different channels to obtain a cross-channel association dataset; Strategy generation module: Based on the performance evaluation results and cross-channel related datasets, combined with a preset optimization objective function, it jointly optimizes the keyword bidding strategy, creative material combination, ad placement time configuration and budget allocation ratio to obtain the promotion execution strategy; Strategy optimization module: used to manage multiple third-party sites in the promotion execution strategy in a hierarchical manner according to conversion efficiency characteristics, and obtain management results; based on the management results, obtain a multi-channel collaborative promotion optimization strategy.

[0016] The beneficial effects of this invention are as follows: By collecting multi-channel promotional data and processing it in a unified format, this invention achieves end-to-end data fusion, solving the problems of data silos and inconsistent formats among channels in existing technologies; by performing time-series correlation and path reconstruction processing on promotional behavior, user clickstream, and conversion events, it can completely track the cross-channel behavioral path of users from initial exposure to final conversion, significantly improving the accuracy of path reconstruction, and quantitatively evaluating the contribution of each touchpoint based on the complete path, overcoming the defect of traditional attribution models that over-attribute to the last touchpoint; by jointly optimizing keyword bidding strategies, creative material combinations, ad placement time configurations, and budget allocation ratios, it changes the shortcomings of independent optimization of strategies in various dimensions and lack of coordination in existing technologies, achieving the optimal overall effect of the promotional strategy; by managing multiple third-party sites in layers according to conversion efficiency characteristics, increasing the budget weight and ad placement frequency for high-performance sites, and reducing the budget and increasing monitoring for low-performance sites, it improves resource allocation efficiency. 20%-30%; By identifying the positive synergy and negative conflict effects between channels, the linkage strength of positive synergy channels is enhanced, effectively avoiding mutual interference and budget waste between channels; By associating and combining multiple sites to form site collaborative promotion units, the collaborative conversion rate is increased by 15%-25% by utilizing the synergy between sites; By issuing the promotion execution strategy to the channel interface and simultaneously collecting actual execution effect data, a response time of seconds or minutes from strategy issuance to effect feedback is achieved, solving the problem of lagging strategy adjustment and feedback cycle of several days in the existing technology; By comparing the actual conversion performance with the expected target for difference analysis, obstacles are identified and screened from multiple dimensions such as keyword matching degree, material attractiveness, accuracy of placement time, rationality of budget allocation and channel competition intensity. Key obstacles are transformed into specific optimization suggestions and automatically fed back to the next round of strategy adjustment, forming an adaptive closed loop of "execution → monitoring → analysis → optimization", shortening the strategy iteration cycle to minutes. Through the combined effect of the above technical solutions, this invention can improve the overall return on investment (ROI), reduce conversion costs, increase conversion rates, improve the efficiency of marketing budget utilization, and reduce labor operating costs. At the same time, it enables the promotion strategy to automatically adapt to changes in different time periods, different user groups, and different competitive environments, maintaining the stability and continuous improvement of conversion efficiency. It systematically solves the shortcomings of existing technologies in cross-channel behavior path restoration, multi-dimensional element collaborative optimization, real-time feedback and adaptive adjustment, etc., and provides a complete, efficient and intelligent solution for multi-channel promotion in complex marketing scenarios. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the multi-channel third-party site promotion method with real-time effect feedback provided in the embodiments of the present invention; Figure 2This is a schematic diagram of a multi-channel third-party site promotion system module providing real-time effect feedback, as provided in an embodiment of the present invention. Detailed Implementation

[0018] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, some features described in the examples may be combined in other examples.

[0019] At least one embodiment of the present invention discloses a multi-channel third-party site promotion method and system with real-time effect feedback, including: like Figure 1 As shown, the multi-channel third-party site promotion method with real-time performance feedback includes the following steps: Step 1: Collect promotional data from third-party websites across multiple channels and perform preprocessing operations to obtain a promotional dataset in a unified format; Step 2: Based on the aforementioned promotion dataset, quantitatively evaluate the contribution of multiple channels in the user conversion path to obtain the effect evaluation results; Step 3: Based on the promotion dataset and the effect evaluation results, perform time-series correlation and path reconstruction processing on the promotion behaviors, user click streams, conversion events and channel context information of different channels to obtain a cross-channel related dataset; Step 4: Based on the performance evaluation results and cross-channel related datasets, and combined with the preset optimization objective function, jointly optimize the keyword bidding strategy, creative material combination, ad placement time configuration, and budget allocation ratio to obtain the promotion execution strategy; Step 5: Manage the multiple third-party sites in the promotion execution strategy in a hierarchical manner according to their conversion efficiency characteristics to obtain management results; based on the management results, obtain a multi-channel collaborative promotion optimization strategy.

[0020] In a specific implementation of this invention, a multi-channel third-party website promotion system with real-time effect feedback is provided. This system includes a data collection module, an effect evaluation module, an association module, a strategy generation module, and a strategy optimization module. These modules work collaboratively to dynamically optimize the promotion strategy. The technical solution of this invention is described in detail below with reference to specific implementation processes.

[0021] First, the data acquisition module starts running, performing promotional data collection and preprocessing operations. This module receives raw log data in real-time or near real-time from multiple third-party promotional channels (such as Baidu search ads, Douyin feed ads, Xiaohongshu product recommendations, WeChat Moments ads, etc.) through standardized API interfaces or log retrieval protocols. This raw log data includes exposure data (recording the time, location, device type, and user identifier of the ad being displayed), click data (recording the time, source URL, target page, and dwell time of the user clicking the ad), conversion data (recording the time, conversion type, and conversion value of the user completing target behaviors such as registration, order placement, and payment), and user behavior trajectory data (recording the user's page jump sequence, interaction events, session duration, and user identifier within the site).

[0022] Upon receiving the heterogeneous data, the data acquisition module immediately performs preprocessing operations. First, the data is cleaned to remove duplicate, abnormal, and invalid data. For example, abnormal records with a single click duration exceeding 24 hours are removed according to preset rules. Second, the cleaned data is standardized: the timestamp format is unified to the ISO8601 standard format; user identifiers provided by different channels (such as Baidu's baiduid, Douyin's deviceid, WeChat's openid, etc.) are converted into unified desensitized user IDs (such as UIDHASHxxx) using the SHA-256 hash algorithm, and a globally unified user identifier system is established to ensure that cross-channel user identities can be linked without compromising privacy; event type tags (such as exposure, click, add-to-cart, payment success) are mapped to a predefined set of enumerated values ​​{EXPOSURE, CLICK, DEEPINTERACTION, CONVERSION}; and numerical fields (such as bid amount, conversion value) are uniformly converted to RMB yuan with two decimal places. For missing fields, historical average values ​​are used for imputation. For example, for click records lacking dwell time, the system completes the data based on the user's historical average dwell time. Finally, the standardized data is stored in a structured format, outputting a promotional dataset in a unified format.

[0023] The performance evaluation module reads the promotion dataset and performs a quantified evaluation of channel contribution. Based on user behavior trajectory data within the promotion dataset, this module identifies the complete user behavior path for each user from initial contact to final conversion. For example, a user might sequentially experience search engine ad exposure → click to enter the landing page → browse product details → close the browser → click again via social media ads the next day → complete registration.

[0024] The module strictly sorts the user behavior path according to timestamps to obtain a user conversion path sequence. In this sequence, each touchpoint carries a channel identifier (e.g., Baidu SEM, Douyin feed), interaction type (exposure / click / deep interaction), interaction time, and interaction depth (e.g., page views, video completion rate). Based on the user conversion path sequence, the module extracts touchpoint information from each conversion path and assigns weights to each touchpoint. According to the touchpoint's position in the path, the time interval between the touchpoint and the final conversion, and the touchpoint's interaction depth, a time-decay weighted model is used to calculate the contribution weight of each touchpoint to the final conversion. Specifically, an exponential decay function is used: Where T is the conversion time, ti is the interaction time of the i-th touchpoint, di is the normalized value of the interaction depth, and λ is the decay coefficient. Touchpoints closer to the final conversion time have higher weights; touchpoints with deeper interaction depths also have higher weights.

[0025] Based on the aforementioned contribution weight values, the module performs a weighted summary of the contribution weights of each channel across all conversion paths, accumulating the contribution weight values ​​of all touchpoints belonging to the same channel to generate the original performance score for each channel. To facilitate horizontal comparison, based on the original performance scores, a linear transformation s'=(s-smin) / (smax-smin)×100 is used to map the performance scores of each channel within a standard range of 0-100, resulting in a standardized performance evaluation result.

[0026] The association module receives the promotion dataset and performance evaluation results, and performs time-series association and reconstruction processing of user paths across channels. This module first extracts the promotion behavior logs of each channel from the promotion dataset, parses the fields of the promotion behavior logs, obtains promotion behavior feature data including ad placement time, placement location, bid amount, and creative type, and assigns a unique promotion behavior identifier (such as PB20240520001234) to each promotion behavior record.

[0027] Simultaneously, user clickstream data associated with promotional activities is extracted from the promotional dataset. This clickstream data is then parsed to obtain user interaction feature data, including click time, click source, visited URL, and dwell time. Standardized user identifier fields (such as UIDHASHabc123) are extracted from the user clickstream data. A one-to-many association mapping between promotional activity identifiers and user identifiers is established through the implicit association between them (e.g., the tracking ID passed via the landing page URL parameter pid=PBxxx).

[0028] Based on the aforementioned association mapping, the module further extracts conversion event data associated with the user identifier, parses the conversion event data, and obtains conversion result feature data including conversion time, conversion type (such as new customer registration and repeat purchase orders), conversion value (such as order amount), and conversion path ID. A three-dimensional data set is constructed based on the promotion behavior feature data, user interaction feature data, and conversion result feature data.

[0029] Based on a three-dimensional data set, using timestamps as the key field, promotional behavior logs, user clickstream data, and conversion event data are sorted and matched temporally to identify the behavioral sequences of the same user at different points in time, generating a time-series correlation data table. Finally, based on the time-series correlation data table, combined with user identifiers and preset time window rules (such as a 7-day attribution window), user behavior records scattered across different channels are aggregated to reconstruct the complete behavioral path of users from ad exposure, click access to final conversion, resulting in a cross-channel correlation dataset containing cross-channel behavioral sequences and path features (such as device type, geographical location, and network environment).

[0030] The strategy generation module performs joint optimization of promotion execution strategies based on performance evaluation results and cross-channel associated datasets. This module first extracts historical promotion characteristics data such as keyword bidding, click-through rate (CTR) of creative materials, time-period traffic density (e.g., hourly UV), and budget consumption rate from the cross-channel associated dataset to construct a multi-dimensional decision space containing multiple decision variables. For example, the keyword bidding variable can be set to change in increments of 0.1 yuan within ±10% of the current bid; the creative material combination variable can be represented as the priority ranking of different creative material IDs; the campaign time variable can be configured as the on / off status (on / off) every hour; and the budget allocation ratio variable can be adjusted in increments of 1% between 0% and 100%. The value range and change step size of each decision variable constitute the decision space model.

[0031] Based on the decision space model, the module sets an optimization objective function, taking maximizing return on investment (ROI) or minimizing conversion costs as the optimization objective, and defines the mathematical expression of the objective function as: maxΣ(Vi / Ci), where Vi is the total conversion value brought by the i-th channel, and Ci is its cost. The calculation rules for the objective function are also defined to obtain the optimized objective function.

[0032] Based on the objective function, the performance evaluation results are transformed into optimization constraints. For channels with performance evaluation results below a preset threshold (e.g., 60 points), the upper limit of their budget allocation ratio is limited to 105% of the current value; for channels with performance evaluation results above a preset threshold (e.g., 85 points), their budget allocation ratio is preferentially increased to 120% of the current value, resulting in a constraint set containing multiple constraint rules.

[0033] Based on the decision space model, the objective function, and the set of constraints, the module uses a genetic algorithm to iteratively search for optimization in a multidimensional decision space: the population is initialized with a random strategy combination, new individuals are generated through crossover and mutation, the objective function value is used as the fitness for selection, and the evolution converges after several generations. The module searches for the combination of decision variables that satisfies the constraints and makes the objective function optimal, thus obtaining the optimal decision solution.

[0034] Based on the optimal decision solution, the numerical values ​​of decision variables are mapped to specific promotion operation instructions. The algorithm analyzes and generates keyword bidding adjustment range (e.g., increase the bid for the keyword "smartwatch" by 0.3 yuan), material priority sorting list (e.g., material A > material C > material B), time period on / off status configuration (e.g., enable advertising from 20:00 to 22:00), and channel budget allocation scheme (e.g., increase the budget share of Douyin channel to 35%), resulting in a promotion execution strategy containing multiple executable strategy parameters.

[0035] The strategy optimization module receives the promotion execution strategy and performs site-level management operations. First, it extracts all third-party site information (such as landing page domains, app store links, and H5 activity page URLs) from the promotion execution strategy. Based on the cross-channel associated dataset, it calculates the conversion performance characteristics of each site, including conversion rate (CVR), cost per click (CPC), return on investment (ROI), and user quality score (calculated based on a comprehensive evaluation of user retention rate, average order value, and other indicators).

[0036] The conversion efficiency features are converted into multi-dimensional feature vectors. For example, the vector for site S1 is [CVR=0.08, CPC=1.2, ROI=3.5, user quality=0.72], thus obtaining the feature dataset of site efficiency.

[0037] Based on the feature dataset, a clustering algorithm is used to perform performance similarity analysis on each third-party site. The Euclidean distance between the feature vectors of any two sites is calculated to obtain the feature distance between each site. Sites with performance feature similarity (i.e., feature distance less than a preset threshold, such as 0.15) higher than a preset threshold are grouped into the same performance level, generating a site hierarchical structure containing multiple levels. Typically, it can be divided into a high-performance layer (e.g., ROI > 4.0), a medium-performance layer (2.0 ≤ ROI ≤ 4.0), and a low-performance layer (ROI < 2.0), resulting in the site hierarchical results.

[0038] Based on the site stratification results, differentiated budget allocation and delivery control strategies are set for different strata. For high-performance sites, the system automatically increases the budget allocation weight (e.g., 1.2 times the baseline value) and delivery frequency (e.g., triggering a delivery check every 15 minutes); for low-performance sites, the budget allocation weight is reduced (e.g., 0.6 times) and the monitoring frequency is increased (e.g., collecting execution data hourly), resulting in a stratified management strategy.

[0039] Based on the site hierarchical results and hierarchical differentiated management strategies, a management data structure is constructed that includes site hierarchical affiliation, hierarchical performance indicator ranges (such as ROI > 4.0 for high-performance layers), and hierarchical management strategy parameters (such as budget weight and monitoring frequency) to obtain management results.

[0040] Based on the management results, the strategy optimization module further performs site association and combination operations to obtain multiple site association groups. First, based on the management results, it extracts the hierarchical affiliation information and performance indicator data of each third-party site, and extracts the historical promotion behavior data and user conversion data of each site from the cross-channel association dataset to construct a site feature database containing site identifiers, hierarchical identifiers, performance characteristics, and behavioral characteristics.

[0041] Based on the site feature database, we obtain site attribute features such as business type, user group, promotion target, geographical distribution and channel affiliation of each third-party site, construct site attribute feature vectors, and calculate the attribute similarity between any two sites to obtain the site attribute similarity matrix.

[0042] Based on a site attribute similarity matrix and a site feature database, the synergistic effect of promotion between sites is analyzed. By calculating user overlap (the proportion of a single user's visits to multiple sites), conversion path crossover (the proportion of users who browse on site A and then convert on site B), and promotion time complementarity (the difference in peak traffic times between sites), site pairs with synergistic promotion effects are identified, and a synergistic effect score is calculated for each site pair, resulting in a site synergistic effect matrix. For example, the analysis found that 35% of users first browsed reviews on site C (Xiaohongshu) and then jumped to site B (Douyin H5) to place an order. The synergistic conversion rate between the two sites was 23% higher than that of individual sites, indicating a higher synergistic effect score between sites C and B.

[0043] Based on the site synergy effect matrix, a synergy effect score threshold is set, and site combinations with synergy effect scores higher than the threshold are selected. The high-scoring site combinations are then clustered to associate the sites within the same group, thus obtaining preliminary site association groups.

[0044] Based on the initial site association groups, the sites within each group are verified to ensure consistency in promotion goals (e.g., whether they all aim to acquire new customers) and feasibility of resource allocation (e.g., the rationality of budget allocation). Sites that do not meet the association conditions are removed to form stable site collaborative promotion units. Each site collaborative promotion unit is assigned a unique unit identifier (e.g., G1, G2) and a ranking sequence of sites within the group, resulting in multiple site association groups.

[0045] The system distributes promotion optimization strategies to the corresponding channel interfaces, driving the associated groups on each site to execute promotion tasks and simultaneously collecting actual execution performance data for each associated group. First, based on the promotion optimization strategies, the system extracts the set of strategy parameters to be distributed. This set of parameters includes keyword bidding adjustment instructions, optimal configuration of creative materials, start / stop rules for ad placement periods, and dynamic allocation parameters for channel budgets. These strategy parameters are then converted to the appropriate format according to the interface protocols of each channel to obtain a channel-compatible strategy parameter package.

[0046] Based on the strategy parameter package, the strategy parameters are pushed to the advertising delivery system of each channel through the parameter interface, and the parameter reception confirmation signal returned by each channel is received to establish a confirmation record of successful strategy delivery.

[0047] Based on the confirmation records, the advertising delivery systems of each channel dynamically adjust keyword bids, creative display order, start / stop status of delivery time periods, and budget allocation according to the received strategy parameters, and generate strategy execution status data to obtain strategy execution feedback.

[0048] Based on strategy execution feedback, during the execution of promotion tasks, exposure data, click data, conversion data and cost data of each site's associated group are collected in real time through the data collection interface. Each piece of collected data is marked with a site associated group identifier (such as G1) and a collection timestamp to obtain the raw execution effect data.

[0049] Based on the original execution effect data, the data is aggregated according to the site association group identifier. The data of the same site association group is summarized and statistically analyzed to calculate the total exposure, total clicks, total conversions and total cost of each site association group. The actual execution effect data of each site association group is generated to obtain the verification data of the promotion effect.

[0050] Based on the validation data, the system analyzes the differences between the actual conversion performance and expected goals of each site's associated groups, and identifies hindering factors for optimization feedback. First, based on the validation data, the system extracts the actual conversion performance data of each site's associated groups, including actual conversion rate, actual conversion cost, and actual return on investment. Then, it extracts the expected conversion goals corresponding to each site's associated groups from the promotion optimization strategy, including target conversion rate, target conversion cost, and target return on investment, resulting in a comparison dataset.

[0051] Based on the comparison dataset, the difference value and difference rate between the actual conversion performance and the expected target are calculated. The absolute values ​​of the difference rates are sorted, and site association groups with difference rates exceeding a preset threshold (such as ±15%) are identified to obtain a list of abnormal site association groups.

[0052] Based on the list of abnormal site association groups, a multi-dimensional causal analysis was conducted on each site association group in the list. Key influencing factors leading to the differences were identified from dimensions such as keyword matching degree (relevance score between query terms and ad keywords), creative attractiveness (based on historical CTR prediction or eye-tracking simulation), accuracy of delivery time (coverage of target user's active time periods), rationality of budget allocation (budget consumption rate and conversion rate), and channel competition intensity (median bid of advertisers in the same industry), resulting in a report analyzing the reasons for the differences.

[0053] Based on the discrepancy analysis report, potential obstacles affecting promotion conversion were identified, including low keyword matching, insufficient creative appeal, inaccurate timing of ad placement, unreasonable budget allocation, and intense channel competition. The importance of these obstacles was assessed, and the weight of each obstacle on conversion performance was calculated using SHAP or LIME methods. The key obstacles with the highest weights were then identified, resulting in a priority list of obstacles.

[0054] Based on the priority list of obstacles, the top-ranked key obstacles are transformed into quantifiable optimization feedback inputs. For issues like low keyword relevance, keyword optimization suggestions are generated (e.g., adjusting keywords to more precise long-tail keywords); for issues like insufficient creative appeal, creative replacement suggestions are generated (e.g., replacing the current creative with the high CTR template #T2024); for issues like inaccurate ad placement times, time slot adjustment suggestions are generated (e.g., focusing on ad placements during the peak user activity period of 20:00-22:00); and for issues like unreasonable budget allocation, budget reallocation suggestions are generated (e.g., reducing the budget for inefficient channels by 10%), resulting in a set of optimization feedback instructions.

[0055] Based on the optimized feedback instruction set, the system adjusts keyword bidding strategies, creative material combinations, ad placement time configurations, and budget allocation ratios, updates the corresponding parameters in the promotion optimization strategy, and generates an optimized promotion optimization strategy. The system writes the new round of promotion optimization strategy back to the execution queue, repeating the steps of strategy issuance, promotion task execution, performance data collection, difference analysis, and obstacle factor identification, forming a data-driven continuous optimization closed-loop process.

[0056] Throughout the system's operation, modules transmit data and synchronize their states via message queues or shared databases. The promotion dataset output by the data acquisition module is written to a distributed data lake; the effect evaluation and correlation modules read this dataset in parallel, outputting effect evaluation results and cross-channel correlation datasets respectively; the strategy generation module relies on these two outputs to generate promotion execution strategies; and the strategy optimization module, based on these, completes hierarchical management, site association combinations, and collaborative optimization, writing the new round of strategies back to the execution queue. All processing steps are guaranteed for transaction consistency, ensuring a rollback to a stable state in the event of partial failures. Through this rigorous module connection, data flow, and feedback mechanism, this invention achieves real-time perception, accurate attribution, dynamic optimization, and collaborative execution of multi-channel third-party site promotion activities, solving core problems in existing technologies such as channel fragmentation, attribution distortion, and strategy lag.

[0057] To enable those skilled in the art to fully understand and implement this invention, the specific implementation principle of this invention is further explained below in conjunction with a specific application scenario.

[0058] During the 618 shopping festival on a certain e-commerce platform, the new smartwatch was promoted through four third-party channels: Baidu search ads, Douyin feed ads, Xiaohongshu product recommendations, and WeChat Moments ads.

[0059] The system's data acquisition module retrieves raw log data generated between 00:00 and 24:00 daily through API interfaces provided by various platforms. This includes keyword exposure and click logs returned by Baidu SEM, video ad completion rates and conversion events returned by Douyin, note interaction behaviors (likes, favorites, and redirects) returned by Xiaohongshu, and click and landing page dwell time records returned by WeChat Ads. All log timestamps are uniformly converted to ISO8601 format under UTC+8 timezone. User identifiers such as Baidu's baiduid, Douyin's deviceid, and WeChat's openid are mapped to a unified desensitized IDUIDHASHxxx using the SHA-256 hash function. Event types such as playback completion, add-to-cart, and successful payment are mapped to predefined enumeration values ​​{EXPOSURE, CLICK, DEEPINTERACTION, CONVERSION}. Order amount fields are uniformly expressed in RMB and retained to two decimal places. For a click record with a missing dwell time, the system imputes it based on the user's historical average dwell time of 38 seconds. An abnormal record marked as having a 72-hour dwell time after clicking is automatically removed because it exceeds a preset threshold (maximum effective dwell time is 2 hours). After this processing, a unified promotion dataset containing 1.2 million structured events is output.

[0060] The performance evaluation module reads the dataset and extracts the complete behavioral path of users who successfully completed payment conversions. For example, user UIDHASH7a3f9b's path is as follows: June 1st, 10:15 AM, triggered keyword ad exposure through a Baidu search for smartwatch recommendations (channel: Baidu SEM, interaction type: EXPOSURE, depth d=0.1); 10:17 AM, clicked to enter the H5 landing page (interaction type: CLICK, d=0.5, dwell time 92 seconds); left without conversion; the next day, June 2nd, 7:30 PM, saw a video ad in the Douyin feed (channel: Douyin, EXPOSURE, d=0.2); 7:32 PM, clicked and watched the full 30-second video (DEEPINTERACTION, d=0.9); 7:35 PM, completed the order (CONVERSION, value V=1299 yuan). This user behavioral path is sorted by timestamp to obtain the user conversion path sequence. Using the conversion time T = June 2nd, 19:35 as the baseline, the contribution weight values ​​of each touchpoint were calculated as follows: Baidu exposure w1 = e(-0.05×28.3)×0.1; Baidu clicks w2 = e(-0.05×28.3)×0.5; Douyin exposure w3 = e(-0.05×0.08)×0.2; Douyin deep interaction w4 = e(-0.05×0.08)×0.9. The contribution weights of each channel were summarized and statistically analyzed: the total weight of the Baidu channel was 0.024 + 0.121 = 0.145, and the total weight of the Douyin channel was 0.199 + 0.896 = 1.095, generating the original performance score. After normalization, the total contribution of the Baidu channel was (0.024 + 0.121) / 1.24, and that of Douyin was (0.199 + 0.896) / 1.24. After aggregating all users, Baidu's standardized score was 58 points, and Douyin's score was 92 points, resulting in the effectiveness evaluation results.

[0061] Based on the above data, the association module generates unique identifiers PB20240601BD001 to PB20240602DY892 for each promotional activity and parses the tracking parameter pid from the landing page URL, thereby binding user click behavior with promotional behavior and establishing an association mapping between promotional behavior identifiers and user identifiers. Simultaneously, it links UIDHASH7a3f9b to the conversion record orderid=O202406021935 in the order system, obtaining the conversion value of 1299 yuan and conversion path ID=CP7a3f9b. Promotional features (such as keywords like smartwatch, bid of 1.8 yuan, and material ID=MAT204), interaction features (device iOS, geographical location Shanghai, and network Wi-Fi), and conversion features (new customer, average order value of 1299 yuan, and 7-day retention rate of 85%) are combined into a three-element data set and globally sorted by timestamp to form a time-series association data table. Based on this, using a 7-day attribution window, and combining user identifiers and time window rules, all cross-channel events under the same UIDHASH are aggregated to reconstruct the complete behavioral path including channel sequence, device switching, and time period jump, and output a cross-channel associated dataset.

[0062] Based on the cross-channel associated dataset and performance evaluation results, the strategy generation module constructs a decision space model: the current bid for the Baidu keyword "smartwatch" is 1.8 yuan, with an allowable adjustment range of 1.62-1.98 yuan (±10%) and a step size of 0.1 yuan; the Douyin material library contains 10 materials from MAT201 to MAT210, and their priority order needs to be determined; the campaign period variable covers 24 hours, and each hour can be turned on and off independently; the initial budget allocation ratio is 30% for Baidu, 50% for Douyin, 15% for Xiaohongshu, and 5% for WeChat.

[0063] The objective function is set as maximizing ROI = Σ(Vi / Ci), where Vi is the total conversion value brought by each channel, and Ci is the actual cost. The performance evaluation results are transformed into constraints: Baidu score 58 < 60, its budget ratio cannot exceed 31.5% (i.e., 30% × 105%); Douyin score 92 > 85, can be increased to 60% (50% × 120%), thus obtaining the constraint set.

[0064] A genetic algorithm was used to initialize 100 strategy individuals. After 50 generations of evolution, the optimal decision solution was converged: Baidu bid was lowered to 1.7 yuan, Douyin material priority was set to [MAT204, MAT207, MAT201], and the campaign was launched from 18:00 to 23:00. The budget allocation was adjusted to Baidu 28%, Douyin 58%, Xiaohongshu 10%, and WeChat 4%. This optimal solution was mapped to specific execution instructions: keyword bid adjustment: Baidu keyword bid lowered by 0.1 yuan, material priority sorting list: MAT204 > MAT207 > MAT201, time slot campaign on / off status configured to be on from 18:00 to 23:00, channel budget allocation scheme: Douyin 58%, Baidu 28%, Xiaohongshu 10%, and WeChat 4%, thus obtaining the promotion execution strategy.

[0065] After receiving the promotion execution strategy, the strategy optimization module extracts the relevant third-party sites: Baidu landing page urla.com / watch, Douyin H5 page urlb.cn / smartwatch, Xiaohongshu redirect page urlc.me / product, and WeChat mini-program page urld.app / order. Based on the historical 7-day cross-channel correlation dataset, the conversion efficiency characteristics of each site are calculated, forming a four-dimensional feature vector: urla=[CVR=0.05, CPC=1.9, ROI=2.1, User Quality=0.65]; urlb=[0.12, 1.1, 4.8, 0.82]; urlc=[0.07, 1.5, 3.2, 0.70]; urld=[0.04, 2.0, 1.8, 0.60], obtaining the site efficiency feature dataset. The Euclidean distance between each site is calculated, and the distance between urlb and urlc is found to be... Sites with a ROI greater than 0.15 are not classified into the same layer; and the distance between urla and urld is 0.38, so they are also not merged. Based on this, the system divides sites into a high-performance layer (urlb only, ROI > 4.0), a medium-performance layer (urlc, 2.0 ≤ ROI ≤ 4.0), and a low-performance layer (urla, urld, ROI < 2.0), resulting in site stratification. The high-performance layer site urlb is assigned a 1.2x budget weight and is set to trigger a campaign check every 15 minutes; the low-performance layer sites urla and urld are included in hourly monitoring with a budget weight of 0.6x, resulting in a tiered differentiated management strategy. A management data structure is constructed, including site tier affiliation (urlb → high-performance layer), tier performance indicator range (high-performance layer ROI > 4.0), and tier management strategy parameters (budget weight 1.2, monitoring frequency 15 minutes), to obtain management results. Based on the management results, the strategy optimization module further analyzes the promotional synergy between sites. Historical promotional behavior data and user conversion data for each site are extracted from cross-channel associated datasets to construct a site feature database. Site attribute features such as business type, user group, and promotion goals are obtained for each site, and a site attribute similarity matrix is ​​calculated.

[0066] Analysis of user paths revealed that 35% of users first browsed reviews on urlc (Xiaohongshu) before redirecting to urlb (Douyin H5) to place an order. The user overlap between the two was 35%, and the conversion path crossover was 23%, resulting in a high synergy score. A synergy score threshold was set, and the high-scoring site combination of urlb and urlc was selected, forming a preliminary site association group.

[0067] The initial site association group is verified to ensure consistency in the promotion goals of urlb and urlc (both aiming at acquiring new customers) and feasibility of resource allocation (sufficient budget), confirming that they meet the association conditions. A stable site collaborative promotion unit is formed and assigned a unique unit identifier G1, resulting in the site association group {G1: [urlc, urlb]}.

[0068] The promotion and optimization strategies are distributed to the interfaces of each channel. First, the strategy parameter set is extracted and converted to the correct format according to the interface protocols of each channel to obtain a channel-compatible strategy parameter package. The strategy parameters are then pushed to the advertising systems of Baidu, Douyin, Xiaohongshu, and WeChat through the parameter interfaces, and confirmation signals are received from each channel to establish confirmation records.

[0069] The advertising delivery systems across all channels dynamically adjust based on the received strategy parameters: Baidu keyword bids are adjusted from 1.8 yuan to 1.7 yuan, the display order of Douyin creative materials is adjusted to prioritize MAT204, the delivery time is adjusted to 18:00-23:00, and budget allocation is executed according to the new plan. Strategy execution status data is generated, and strategy execution feedback is obtained.

[0070] During the promotion task execution, the system collects exposure, click, conversion, and cost data for each site's associated group in real time. For site associated group G1, the following data was collected: total exposures of 500,000, total clicks of 25,000, total conversions of 300, and total cost of 33,000 yuan. Each piece of collected data is labeled with the site associated group identifier G1 and the collection timestamp to obtain the raw execution effect data. Data is aggregated according to the site associated group identifier to generate the actual execution effect data for G1, thus obtaining the promotion effect verification data.

[0071] Based on verification data, the system extracts the actual conversion performance of site-related group G1: actual conversion rate = 300 / 25000 = 1.2%, actual conversion cost = 33000 / 300 = 110 yuan, actual return on investment = 389700 / 33000 = 11.8 (assuming an average order value of 1299 yuan). The system extracts the expected conversion targets for G1 from the promotion optimization strategy: target conversion rate = 1.5%, target conversion cost = 100 yuan, target return on investment = 13.0, resulting in a comparison dataset. The system calculates the difference rates: conversion rate difference rate = (1.2% - 1.5%) / 1.5% = -20%; conversion cost difference rate = (110 - 100) / 100 = +10%; return on investment difference rate = (11.8 - 13.0) / 13.0 = -9.2%. Since the conversion rate difference rate of -20% exceeds the preset threshold ±15%, G1 is added to the list of abnormal site-related groups.

[0072] A multi-dimensional analysis of the reasons for G1's performance was conducted: Regarding keyword matching, the relevance between the Xiaohongshu search term "watches with long battery life" and the ad term "smartwatches" was only 0.41, indicating a low matching degree. In terms of creative appeal, the current Xiaohongshu image CTR prediction value is 2.1%, lower than the 3.8% of the #T2024 template. Regarding the accuracy of the ad placement time, the peak activity time for the target users is 20:00-22:00, but Xiaohongshu's ad placement covers the entire time period from 18:00-23:00, resulting in insufficient accuracy. A report analyzing the reasons for these differences was obtained.

[0073] The impact weight of each obstacle factor was calculated using SHAP values: Creative attractiveness weight 0.47, Keyword relevance weight 0.28, Targeting time accuracy weight 0.18, and Others 0.07. Creative attractiveness was identified as the key obstacle factor with the highest impact weight, resulting in a priority list of obstacle factors.

[0074] Transform key obstacles into optimization feedback inputs: generate material replacement suggestions for Xiaohongshu channel to replace materials with high CTR template #T2024, generate keyword optimization suggestions to adjust Xiaohongshu keywords to "long battery life smartwatch", and generate time period adjustment suggestions to adjust Xiaohongshu ad time to 20:00-22:00, thus obtaining an optimization feedback instruction set.

[0075] Based on the optimized feedback instruction set, the promotion optimization strategy was adjusted: the Xiaohongshu creative materials were replaced with the template #T2024, the keyword was adjusted to "long-lasting smartwatch," and the ad placement period was shortened to 20:00-22:00, generating a new round of optimized promotion strategies. The system writes the new strategy back to the execution queue, repeatedly executing steps such as strategy distribution, task execution, data collection, and difference analysis, forming a closed-loop process of continuous optimization.

[0076] Through the above-described complete implementation process, this invention achieves fully automated operation of the entire process of promotional activities on four channels—Baidu, Douyin, Xiaohongshu, and WeChat—during the 618 promotional event on the e-commerce platform. This includes real-time data collection, precise performance evaluation, cross-channel path restoration, dynamic strategy optimization, site-level hierarchical management, collaborative execution, and differential analysis and feedback. This effectively improves the promotion conversion rate and return on investment, verifying the effectiveness and practicality of the technical solution of this invention.

[0077] like Figure 2 As shown, the multi-channel third-party website promotion system with real-time performance feedback includes: Data acquisition module: Used to collect promotional data from third-party websites across multiple channels and perform preprocessing operations to obtain a promotional dataset in a unified format; Performance evaluation module: Based on the promotion dataset, this module quantifies the contribution of multiple channels in the user conversion path to obtain performance evaluation results. Association module: Based on the promotion dataset and the effect evaluation results, it performs time-series association and path reconstruction processing on the promotion behavior, user click flow, conversion events and channel context information of different channels to obtain a cross-channel association dataset; Strategy generation module: Based on the performance evaluation results and cross-channel related datasets, combined with a preset optimization objective function, it jointly optimizes the keyword bidding strategy, creative material combination, ad placement time configuration and budget allocation ratio to obtain the promotion execution strategy; Strategy optimization module: used to manage multiple third-party sites in the promotion execution strategy in a hierarchical manner according to conversion efficiency characteristics, and obtain management results; based on the management results, obtain a multi-channel collaborative promotion optimization strategy.

[0078] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A multi-channel third-party site promotion method with real-time performance feedback, characterized in that... include: Collect promotional data from third-party websites across multiple channels and perform preprocessing operations to obtain a promotional dataset in a unified format; Based on the aforementioned promotion dataset, the contribution of multiple channels in the user conversion path is quantitatively evaluated to obtain the effect evaluation results. Based on the promotion dataset and the effect evaluation results, the promotion behaviors, user click streams, conversion events and channel context information of different channels are correlated in time and restored through path reconstruction to obtain a cross-channel associated dataset. Based on the performance evaluation results and cross-channel related datasets, and combined with the preset optimization objective function, the keyword bidding strategy, creative material combination, ad placement time configuration and budget allocation ratio are jointly optimized to obtain the promotion execution strategy; The multiple third-party sites in the aforementioned promotion execution strategy are managed in layers according to their conversion efficiency characteristics, resulting in management outcomes. Based on the management results, obtain multi-channel collaborative promotion optimization strategies; The promotion and optimization strategy is a multi-process collaborative optimization and iteration strategy scheme that involves hierarchical management, site association and combination, actual execution verification, difference analysis and identification and elimination of obstacles.

2. The multi-channel third-party site promotion method with real-time effect feedback according to claim 1, characterized in that, Based on the aforementioned promotion dataset, the contribution of multiple channels in the user conversion path is quantitatively evaluated, including: The promotional data includes exposure data, click data, conversion data, and user behavior trajectory data; Based on user behavior trajectory data, identify the complete user behavior path from first contact to final conversion; Arrange the user behavior paths in chronological order to obtain a user conversion path sequence; Based on the user conversion path sequence, the touchpoint information in each conversion path is extracted, and each touchpoint is weighted. According to the position of the touchpoint in the path, the time interval between the touchpoint and the final conversion, and the interaction depth of the touchpoint, the contribution weight value of each touchpoint to the final conversion is obtained. Based on the aforementioned contribution weight values, the contribution weight values ​​of each channel in all conversion paths are weighted and summarized. The contribution weight values ​​of all touchpoints belonging to the same channel are accumulated to generate the original performance score of each channel. Based on the original performance scores, the performance scores of each channel are mapped to a numerical range so that the performance scores fall within a preset standard range, resulting in standardized performance evaluation results.

3. The multi-channel third-party site promotion method with real-time effect feedback according to claim 1, characterized in that, Based on the aforementioned promotion dataset and the aforementioned performance evaluation results, the promotional behaviors, user clickstreams, conversion events, and channel context information from different channels are subjected to time-series correlation and path reconstruction processing, including: Extract promotion behavior logs for each channel from the aforementioned promotion dataset; The promotion behavior logs are parsed to obtain promotion behavior feature data including ad placement time, placement location, bid amount and material type, and a unique promotion behavior identifier is assigned to each promotion behavior record; Extract user clickstream data associated with promotional behavior from the promotion dataset, parse the user clickstream data to obtain user interaction feature data including click time, click source, visited page and dwell time; Extract standardized user identifiers from the user clickstream data; and establish an association mapping between promotion behavior identifiers and user identifiers. Based on the aforementioned association mapping, extract the conversion event data associated with the user identifier; The conversion event data is parsed to obtain conversion result feature data including conversion time, conversion type, conversion value, and conversion path; Construct a three-dimensional data set based on promotion behavior characteristic data, user interaction characteristic data, and conversion event data; Based on the aforementioned three-data set, the promotion behavior logs, user clickstream data and conversion event data are sorted and matched in time sequence with timestamp as the key field to identify the behavior sequence of the same user at different time points and generate a time sequence association data table. Based on the aforementioned time-series association data table, combined with user identifiers and preset time window rules, user behavior records scattered across different channels are aggregated to reconstruct the complete behavioral path of users from ad exposure, click access to final conversion, resulting in a cross-channel association dataset containing cross-channel behavioral sequences and path features.

4. The multi-channel third-party site promotion method with real-time effect feedback according to claim 3, characterized in that, The promotion and implementation strategy includes: Based on the cross-channel associated dataset, historical promotion feature data such as keyword bidding, creative click-through rate, time period traffic density, and budget consumption rate are extracted to construct a multi-dimensional decision space containing multiple decision variables. The value range and change step size of each decision variable are determined to obtain the decision space model. Based on the decision space model, a promotion optimization objective function is set, with maximizing the return on investment or minimizing the conversion cost as the optimization objective. The mathematical expression and calculation rules of the objective function are defined to obtain the optimization objective function. Based on the objective function, the performance evaluation results are transformed into optimization constraints; Among them, for channels whose performance evaluation results are lower than the preset threshold, the upper limit of the budget allocation ratio is restricted; for channels whose performance evaluation results are higher than the preset threshold, the budget allocation ratio is increased first, resulting in a set of constraint conditions containing multiple constraint rules. Based on the decision space model, the objective function and the set of constraints, iterative optimization is performed in the multidimensional decision space to search for a combination of decision variables that satisfies the constraints and makes the objective function optimal, thereby obtaining the optimal decision solution. Based on the optimal decision solution, the decision variable values ​​are mapped to promotion operation instructions, and keyword bidding adjustment range, material priority sorting list, time period delivery switch status configuration and channel budget allocation scheme are generated by parsing, resulting in a promotion execution strategy containing multiple executable strategy parameters.

5. The multi-channel third-party site promotion method with real-time effect feedback according to claim 4, characterized in that, The multiple third-party sites in the aforementioned promotion execution strategy are managed in layers according to their conversion efficiency characteristics, including: Information about each third-party site is extracted from the promotion execution strategy, and conversion performance characteristics of each site, including conversion rate, cost per click, return on investment and user quality score, are calculated based on the cross-channel associated dataset. The conversion efficiency features are converted into multi-dimensional feature vectors to obtain the feature dataset of site efficiency; Based on the feature dataset, a performance similarity analysis is performed on each third-party site to obtain the feature distance between each site. Sites with performance feature similarity higher than a preset threshold are grouped into the same level, generating a site hierarchical structure containing multiple levels, and obtaining the site hierarchical result. Based on the site stratification results, differentiated budget allocation strategies and delivery control strategies are set for different strata. For sites in the high-performance strata, the budget allocation weight and delivery frequency are increased, while for sites in the low-performance strata, the budget allocation weight is reduced and the monitoring frequency is increased, resulting in a stratified differentiated management strategy. Based on the site hierarchical results and the hierarchical differentiated management strategy, a management data structure is constructed that includes site hierarchical affiliation, hierarchical performance indicator ranges, and hierarchical management strategy parameters to obtain management results.

6. The multi-channel third-party site promotion method with real-time effect feedback according to claim 5, characterized in that, Based on the management results, a multi-channel collaborative promotion optimization strategy is obtained, including: The management results will associate and combine multiple third-party sites to obtain multiple site association groups; The promotion execution strategy is sent to the corresponding channel interface to drive the related groups of each site to execute promotion tasks, and the actual execution effect data of each related group of the site is collected at the same time to obtain verification data of the promotion effect; Based on the verification data, a difference analysis was conducted on the actual conversion performance of each site's associated group and the expected target to obtain the results of the performance differences. Based on the results of the performance differences, identify and eliminate obstacles that affect promotion conversion, and use these obstacles as optimization feedback inputs to output promotion optimization strategies.

7. The multi-channel third-party site promotion method with real-time effect feedback according to claim 6, characterized in that, The management results associate and combine multiple third-party sites to obtain multiple site association groups, including: Based on the management results, the hierarchical affiliation information and performance indicator data of each third-party site are extracted, and the historical promotion behavior data and user conversion data of each site are extracted from the cross-channel association dataset to construct a site feature database containing site identifiers, hierarchical identifiers, performance characteristics and behavioral characteristics. Based on the site feature database, obtain the site attribute features of each third-party site, including business type, user group, promotion target, geographical distribution and channel affiliation, construct site attribute feature vector, and calculate the attribute similarity between any two sites to obtain the site attribute similarity matrix. Based on the site attribute similarity matrix and the site feature database, the promotion synergy effect between sites is analyzed. By calculating the user overlap, conversion path intersection and promotion time complementarity between sites, site pairs with synergistic promotion effects are identified, and a synergistic effect score is calculated for each site pair to obtain the site synergy effect matrix. Based on the site synergy effect matrix, a synergy effect score threshold is set, and site combinations with synergy effect scores higher than the threshold are selected. The high-scoring site combinations are then clustered to associate the sites within the same group, thus obtaining preliminary site association groups. Based on the initial site association groups, the sites within each group are verified to ensure consistency of promotion goals and feasibility of resource allocation. Sites that do not meet the association conditions are removed, forming stable site collaborative promotion units. Each site collaborative promotion unit is assigned a unique unit identifier and a site ranking sequence within the group, resulting in multiple site association groups.

8. The multi-channel third-party site promotion method with real-time effect feedback according to claim 6, characterized in that, The promotion execution strategy is distributed to the corresponding channel interfaces to drive the associated groups of each site to execute promotion tasks, and the actual execution effect data of each associated group is collected synchronously to obtain verification data of the promotion effect, including: Based on the aforementioned promotion optimization strategy, extract the set of strategy parameters that need to be issued; The strategy parameter set includes keyword bidding adjustment instructions, creative material optimization configuration, ad placement start and stop rules, and channel budget dynamic allocation parameters. The strategy parameters are converted according to the interface protocols of each channel to obtain a channel-compatible strategy parameter package. Based on the strategy parameter package, the strategy parameters are pushed to the advertising delivery system of each channel through the parameter interface, and the parameter reception confirmation signal returned by each channel is received to establish a confirmation record of successful strategy delivery. Based on the confirmation records, the advertising delivery systems of each channel dynamically adjust keyword bids, material display order, delivery time periods and budget allocation according to the received strategy parameters, and generate strategy execution status data to obtain strategy execution feedback. Based on the strategy execution feedback, during the execution of the promotion task, exposure data, click data, conversion data and cost data of each site-related group are collected in real time through the data collection interface, and each collected data is marked with a site-related group identifier and a collection timestamp to obtain the original execution effect data. Based on the original execution effect data, data is aggregated according to the site association group identifier, and the data of the same site association group is summarized and statistically analyzed. The total exposure, total clicks, total conversions and total cost of each site association group are calculated to generate the actual execution effect data of each site association group and obtain the verification data of the promotion effect.

9. The multi-channel third-party site promotion method with real-time effect feedback according to claim 6, characterized in that, Based on the verification data, the actual conversion performance data of each site's associated group is extracted, including the actual conversion rate, actual conversion cost, and actual return on investment. The expected conversion targets corresponding to each site's associated group are extracted from the promotion optimization strategy, including the target conversion rate, target conversion cost, and target return on investment, to obtain a comparison dataset. Based on the comparison dataset, the difference value and difference rate between the actual conversion performance and the expected target are calculated. The absolute values ​​of the difference rates are sorted, and site association groups with difference rates exceeding a preset threshold are identified to obtain a list of abnormal site association groups. Based on the list of abnormal site association groups, a multi-dimensional cause analysis is conducted on each site association group in the list. Key influencing factors leading to the differences are identified from dimensions such as keyword matching degree, material attractiveness, accuracy of delivery time, rationality of budget allocation and channel competition intensity, and a difference cause analysis report is obtained. Based on the aforementioned analysis report of the reasons for the differences, potential obstacles affecting promotion conversion are identified, including low keyword matching, insufficient attractiveness of creative materials, inaccurate timing of ad placement, unreasonable budget allocation, and intense channel competition. The importance of these obstacles is assessed, the weight of each obstacle on the conversion effect is calculated, and the key obstacles with the greatest impact weight are selected to obtain a priority list of obstacles. Based on the priority list of obstacles, the key obstacles ranked at the top are transformed into quantifiable optimization feedback inputs. Keyword optimization suggestions are generated for the problem of low keyword matching, material replacement suggestions are generated for the problem of insufficient material attractiveness, time period adjustment suggestions are generated for the problem of inaccurate delivery time, and budget reallocation suggestions are generated for the problem of unreasonable budget allocation, thus obtaining an optimization feedback instruction set. Based on the optimized feedback instruction set, the keyword bidding strategy, creative material combination, ad placement time configuration and budget allocation ratio are adjusted, the promotion optimization strategy is updated, and the optimized promotion optimization strategy is generated.

10. A multi-channel third-party site promotion system with real-time performance feedback, used to execute the multi-channel third-party site promotion method with real-time performance feedback as described in any one of claims 1-9, characterized in that, include: Data acquisition module: Used to collect promotional data from third-party websites across multiple channels and perform preprocessing operations to obtain a promotional dataset in a unified format; Performance evaluation module: Based on the promotion dataset, this module quantifies the contribution of multiple channels in the user conversion path to obtain performance evaluation results. Association module: Based on the promotion dataset and the effect evaluation results, it performs time-series association and path reconstruction processing on the promotion behavior, user click flow, conversion events and channel context information of different channels to obtain a cross-channel association dataset; Strategy generation module: Based on the performance evaluation results and cross-channel related datasets, combined with a preset optimization objective function, it jointly optimizes the keyword bidding strategy, creative material combination, ad placement time configuration and budget allocation ratio to obtain the promotion execution strategy; Strategy optimization module: used to manage multiple third-party sites in the promotion execution strategy in a hierarchical manner according to conversion efficiency characteristics, and obtain management results; based on the management results, obtain a multi-channel collaborative promotion optimization strategy.

Images