Policy evaluation and optimization method, apparatus and device, computer program product

Abstract

This application discloses a strategy evaluation and optimization method, apparatus, and computer program product. The method includes: acquiring and binding structured configuration parameters and business execution result data of the current strategy scenario; using a multi-agent collaborative evaluation and attribution mechanism to quantitatively evaluate and logically attribute the bound data, obtaining quantitative evaluation results and logical attribution results; generating and displaying a visual performance dashboard based on the quantitative evaluation results and logical attribution results; and, based on the logical attribution results, using a preset optimization agent to output structured optimization suggestions, and optimizing the structured configuration parameters of the current strategy scenario based on the visual performance dashboard and the structured optimization suggestions. This application achieves full-link automated processing of the strategy's "evaluation-attribution-optimization" through a multi-agent collaborative mechanism, improving the evaluation accuracy, attribution efficiency, and iteration speed of enterprise and micro-enterprise operation strategies, and realizing a business closed loop of cost reduction and efficiency improvement.

Description

Technical Field

[0001] This application relates to the fields of data processing and artificial intelligence technology, and in particular to a strategy evaluation and optimization method, apparatus and equipment, and computer program product. Background Technology

[0002] In current private domain operations, WeChat Work (hereinafter referred to as "WeChat Work") has become a core platform for enterprises to conduct refined communication with customers. Operations personnel typically use SOP (Standard Operating Procedure) systems to set automated operational rules, such as, "Automatically send a poster for financial product A three days after a customer is tagged as 'high net worth'." After the strategy is executed, data analysts will build a data dashboard through the underlying data warehouse to collect statistics such as "distribution rate," "click-through rate," and "final conversion rate" for quarterly or monthly reviews.

[0003] However, in existing private domain operation systems, strategy configuration and execution are disconnected from data performance feedback. Traditional data dashboards can only display lagging results indicators and cannot perform automated attribution and inference. Furthermore, the lack of a scientific incremental evaluation model leads to low efficiency and high subjectivity in manual strategy optimization. Summary of the Invention

[0004] This application provides a strategy evaluation and optimization method, apparatus and equipment, and computer program product, which achieves a business closed loop of cost reduction and efficiency improvement by enhancing the accuracy and efficiency of strategy evaluation.

[0005] The embodiments of this application adopt the following technical solutions:

[0006] In a first aspect, embodiments of this application provide a strategy evaluation and optimization method, the strategy evaluation and optimization method comprising:

[0007] Obtain the structured configuration parameters and corresponding business execution result data of the current policy scenario and bind the data to obtain the bound current policy scenario data;

[0008] The data of the current policy scenario after binding is quantitatively evaluated and logically attributed using an evaluation and attribution mechanism based on multi-agent collaboration, so as to obtain the quantitative evaluation results and logical attribution results of the current policy scenario.

[0009] A visual performance dashboard is generated and displayed based on the quantitative evaluation results and logical attribution results of the current strategy scenario.

[0010] Based on the logical attribution results of the current strategy scenario, a pre-defined optimization agent outputs structured optimization suggestions, and based on the visualized results dashboard, the structured configuration parameters of the current strategy scenario are optimized and updated using the structured optimization suggestions.

[0011] Optionally, the step of obtaining the structured configuration parameters and corresponding business execution result data of the current policy scenario and binding them to obtain the bound current policy scenario data includes:

[0012] Real-time monitoring of the running status of policy scenarios in the SOP rule engine;

[0013] If the running status of the policy scenario is detected as running or completing a cycle, the structured configuration parameters of the policy scenario are parsed.

[0014] Retrieve the business execution result data corresponding to the strategy scenario from the pre-set data platform wide table;

[0015] The structured configuration parameters of the strategy scenario are bound to the corresponding business execution result data to obtain the current strategy scenario data after binding.

[0016] Optionally, the intelligent agent includes a quantitative analysis intelligent agent, and the quantitative evaluation result includes the pure incremental value corresponding to the current policy scenario and the scores of each funnel stage. The step of using a multi-agent collaborative evaluation and attribution mechanism to perform quantitative evaluation and logical attribution on the bound current policy scenario data to obtain the quantitative evaluation result and logical attribution result of the current policy scenario includes:

[0017] Using the aforementioned quantitative analysis agent, virtual control group data is extracted from the unreached silent customer group, and the bound current strategy scenario data is used as experimental group data.

[0018] By comparing the experimental group data and the virtual control group data, the pure incremental value corresponding to the current strategy scenario is calculated;

[0019] Each funnel stage in the current strategy scenario is scored to obtain a score for each funnel stage in the current strategy scenario.

[0020] Optionally, the intelligent agent includes a logical attribution intelligent agent, and the step of using a multi-agent collaborative evaluation and attribution mechanism to quantitatively evaluate and logically attribute the bound current policy scenario data to obtain the quantitative evaluation result and logical attribution result of the current policy scenario includes:

[0021] The overall score of the current strategy scenario is calculated based on the scores of each funnel stage under the current strategy scenario.

[0022] If the overall score of the current strategy scenario is lower than a preset score threshold, the logical attribution agent is used to perform logical attribution on the bound current strategy scenario data to obtain the logical attribution result of the current strategy scenario.

[0023] Optionally, the step of using the logical attribution agent to perform logical attribution on the bound current policy scenario data to obtain the logical attribution result of the current policy scenario includes:

[0024] The low-scoring funnel stage is determined based on the scores of each funnel stage in the current strategy scenario.

[0025] Using the logical attribution agent, cross-logical diagnosis is performed on the structured configuration parameters of the low-score funnel and the policy scenario to obtain the logical attribution result of the current policy scenario.

[0026] Optionally, the visualization results dashboard includes at least one of a dynamic incremental funnel, an interpretable attribution module, and a strategy health radar chart; the dynamic incremental funnel is used to compare and display the experimental group funnel and the virtual control group funnel in a parallel manner; the interpretable attribution module is used to visualize the logical attribution results; and the strategy health radar chart is used to comprehensively display the health scores of multi-dimensional indicators.

[0027] Optionally, the step of outputting structured optimization suggestions using a preset optimization agent based on the logical attribution results of the current strategy scenario, and optimizing and updating the structured configuration parameters of the current strategy scenario using the structured optimization suggestions based on the visualized results dashboard, includes:

[0028] Based on the logical attribution results of the current strategy scenario, the preset optimization agent is used to search the strategy knowledge base to obtain historically high-efficiency strategies.

[0029] Output the structured optimization suggestions based on the historically high-efficiency strategies;

[0030] Based on the structured configuration parameters of the current strategy scenario and the structured optimization suggestions, parameter tuning instruction cards are generated and displayed in the visualization results dashboard, so that users can optimize and update the structured configuration parameters of the current strategy scenario through the parameter tuning instruction cards in the visualization results dashboard.

[0031] Secondly, embodiments of this application also provide a strategy evaluation and optimization apparatus, the strategy evaluation and optimization apparatus comprising:

[0032] The acquisition unit is used to acquire the structured configuration parameters and corresponding business execution result data of the current policy scenario and bind the data to obtain the bound current policy scenario data;

[0033] The evaluation and attribution unit is used to perform quantitative evaluation and logical attribution on the bound current policy scenario data using a multi-agent collaborative evaluation and attribution mechanism, so as to obtain the quantitative evaluation result and logical attribution result of the current policy scenario.

[0034] The visualization unit is used to generate and display a visualization performance dashboard based on the quantitative evaluation results and logical attribution results of the current strategy scenario.

[0035] The optimization unit is used to output structured optimization suggestions using a preset optimization agent based on the logical attribution results of the current strategy scenario, and to optimize and update the structured configuration parameters of the current strategy scenario based on the structured optimization suggestions using the visualization results dashboard.

[0036] Thirdly, embodiments of this application also provide an apparatus, comprising:

[0037] A processor; and a memory arranged to store computer-executable instructions, which, when executed, cause the processor to perform any of the aforementioned strategy evaluation and optimization methods.

[0038] Fourthly, embodiments of this application also provide a computer program product, including a computer program / instruction, which, when executed by a processor, implements any of the aforementioned strategy evaluation and optimization methods.

[0039] The at least one technical solution adopted in this application embodiment can achieve the following beneficial effects: The strategy evaluation and optimization method of this application embodiment first obtains the structured configuration parameters and corresponding business execution result data of the current strategy scenario and binds the data to obtain the bound current strategy scenario data; then, it uses an evaluation and attribution mechanism based on multi-agent collaboration to perform quantitative evaluation and logical attribution on the bound current strategy scenario data to obtain the quantitative evaluation result and logical attribution result of the current strategy scenario; then, it generates and displays a visual performance dashboard based on the quantitative evaluation result and logical attribution result of the current strategy scenario; finally, based on the logical attribution result of the current strategy scenario, it uses a preset optimization agent to output structured optimization suggestions, and uses the structured optimization suggestions to optimize and update the structured configuration parameters of the current strategy scenario based on the visual performance dashboard. The strategy evaluation and optimization method of this application realizes the full-link automated processing of strategy "evaluation-attribution-optimization" through a multi-agent collaborative mechanism. It breaks through the blind spot of traditional reports that only rely on absolute conversion rates for evaluation, and uses scientific causal inference to truly measure the pure incremental business value of the strategy. At the same time, it upgrades the inefficient attribution mode that used manual SQL writing for data analysis to automated root cause location through intelligent agent collaboration, and upgrades the data dashboard from a one-way display tool to a two-way interactive console that can directly issue optimization instructions. This significantly improves the evaluation accuracy, attribution efficiency and iteration speed of enterprise and micro-business operation strategies, and realizes a business closed loop of cost reduction and efficiency improvement. Attached Figure Description

[0040] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0041] Figure 1 This is a flowchart illustrating a strategy evaluation and optimization method according to an embodiment of this application;

[0042] Figure 2 This is a schematic diagram of the structure of a strategy evaluation and optimization device according to an embodiment of this application;

[0043] Figure 3 This is a schematic diagram of the structure of a device according to an embodiment of this application. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0045] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0046] Existing technologies have the following significant shortcomings in evaluating and optimizing private domain operation strategies:

[0047] First, the effectiveness evaluation is unscientific and lacks pure incremental measurement. Existing dashboards usually only count the "percentage of customers who convert after being reached," which includes "natural conversion" (i.e., customers who would buy even without being messaged), and cannot assess the real incremental contribution (uplift) brought by the SOP strategy.

[0048] Second, the attribution chain is broken, heavily reliant on manual investigation. When the dashboard shows a decline in the conversion rate of a certain strategy, the system cannot explain "why the decline occurred." Data personnel need to manually compare historical reports to guess whether it is due to "incorrect timing of outreach (T+3)" or "narrowing of audience tags," which is time-consuming and extremely dependent on experience.

[0049] Third, the optimization process lacks a closed loop. Currently, the data dashboard merely provides a visual display of results and lacks reverse control capabilities. After analyzing the results, operations personnel still need to exit the dashboard and manually modify rule parameters in the SOP configuration backend, failing to achieve an intelligent iteration loop for the strategy.

[0050] In summary, the existing private domain operation system is disconnected from the strategy configuration and execution and data performance feedback. Traditional data dashboards can only display lagging results indicators and cannot perform automated attribution analysis. Furthermore, the lack of a scientific incremental evaluation model leads to low efficiency and high subjectivity in manual strategy optimization. There is an urgent need for a technical solution that can realize automated strategy evaluation, intelligent attribution analysis, and closed-loop parameter iteration.

[0051] Based on this, embodiments of this application provide a strategy evaluation and optimization method, such as... Figure 1 The diagram shows a flowchart of a strategy evaluation and optimization method according to an embodiment of this application. The strategy evaluation and optimization method includes the following steps S110 to S140:

[0052] Step S110: Obtain the structured configuration parameters and corresponding business execution result data of the current strategy scenario and bind the data to obtain the bound current strategy scenario data.

[0053] First, the structured configuration parameters for the current strategy scenario are read. These parameters are in the form of, for example, {Target Customer Group: No login in the last 30 days, Trigger Condition: T+3 days, Push Material: Resurrection Poster}. These parameters together constitute a complete strategy configuration profile. Simultaneously, business execution result data corresponding to this strategy scenario is obtained, including business metrics such as delivery rate (the percentage of users successfully reached by the message), click-through rate (the percentage of users who click on the pushed content), and daily active user retention rate (the percentage of reached users who log in again within the next N days).

[0054] Subsequently, using the unique identifier of the strategy scenario (such as the strategy ID) as the association key, the above structured configuration parameters are bound one by one with the corresponding business execution result data to form a bound dataset with the strategy scenario as the dimension, which serves as the input basis for subsequent evaluation and attribution.

[0055] Step S120: Quantitatively evaluate and logically attribute the bound current policy scenario data using a multi-agent collaborative evaluation and attribution mechanism to obtain the quantitative evaluation results and logical attribution results of the current policy scenario.

[0056] By utilizing a multi-agent collaborative evaluation and attribution mechanism, the current policy scenario data after binding is first quantitatively evaluated to calculate the pure incremental value (Uplift Score) brought by the policy. Then, when the overall policy score is lower than a set threshold, logical attribution analysis such as parameter diagnosis is performed. Finally, the quantitative evaluation results and logical attribution results of the current policy scenario are output.

[0057] Step S130: Generate and display a visual performance dashboard based on the quantitative evaluation results and logical attribution results of the current strategy scenario.

[0058] Based on the quantitative evaluation results and logical attribution results, a customized visual performance dashboard for the specific strategy scenario is dynamically generated on the web or client side. This dashboard dynamically renders performance and attribution information, providing a visual presentation. The dashboard is dynamically generated; that is, whenever the strategy scenario data is updated, the dashboard content is refreshed in real time, thus achieving an intuitive and immediate presentation of the strategy's effectiveness.

[0059] Step S140: Based on the logical attribution results of the current strategy scenario, a pre-defined optimization agent outputs structured optimization suggestions, and based on the visualized results dashboard, the structured configuration parameters of the current strategy scenario are optimized and updated using the structured optimization suggestions.

[0060] Based on the above logical attribution results, the pre-defined optimization agent automatically outputs structured optimization suggestions based on the problem parameters diagnosed by attribution, combined with a pre-defined optimization rule base and historical strategy knowledge base. For example, adjusting the "trigger condition" from "T+3" days to "T+1" days. Simultaneously, the visual results dashboard supports interactive operation, allowing relevant personnel to directly adopt, modify, or reject optimization suggestions on the dashboard interface. Once an optimization suggestion is confirmed, based on the reverse deduction logic of the strategy configuration parameters, the optimized structured configuration parameters are automatically written back to the strategy management system, completing the parameter optimization update for the strategy scenario, thus forming a closed-loop business process of "evaluation-attribution-optimization".

[0061] The strategy evaluation and optimization method of this application realizes the full-link automated processing of strategy "evaluation-attribution-optimization" through a multi-agent collaborative mechanism. It breaks through the blind spot of traditional reports that only rely on absolute conversion rates for evaluation, and uses scientific causal inference to truly measure the pure incremental business value of the strategy. At the same time, it upgrades the inefficient attribution mode that used manual SQL writing for data analysis to automated root cause location through intelligent agent collaboration, and upgrades the data dashboard from a one-way display tool to a two-way interactive console that can directly issue optimization instructions. This significantly improves the evaluation accuracy, attribution efficiency and iteration speed of enterprise and micro-business operation strategies, and realizes a business closed loop of cost reduction and efficiency improvement.

[0062] In some embodiments of this application, obtaining the structured configuration parameters of the current policy scenario and the corresponding business execution result data and binding them to obtain the bound current policy scenario data includes: real-time monitoring of the running status of the policy scenario in the SOP rule engine; when the running status of the policy scenario is detected as running or completing a cycle, parsing the structured configuration parameters of the policy scenario; pulling the business execution result data corresponding to the policy scenario from a preset data platform wide table; and binding the structured configuration parameters of the policy scenario with the corresponding business execution result data to obtain the bound current policy scenario data.

[0063] First, the SOP (Standard Operating Procedure) rule engine is monitored in real time through a pre-defined system interface to continuously track changes in the operational status of each policy scenario. When the operational status of a policy scenario changes to "running" or "one execution cycle completed," the subsequent data acquisition and binding process is triggered. Here, "one cycle" can be understood as the complete business cycle from triggering to execution of the policy. For example, a 7-day user wake-up policy is considered to have completed one cycle after execution.

[0064] Secondly, when the above triggering conditions are detected, the structured configuration parameters of the strategy scenario are automatically parsed. Specifically, the complete configuration information of the strategy, organized in key-value pairs, is read from the SOP rule engine, including but not limited to target customer group fields such as "not logged in for the past 30 days", triggering condition fields such as "T+3 days", and push material fields such as "wake-up poster". These parameters are extracted as a structured feature profile of the strategy scenario.

[0065] Then, based on the unique identifier of the strategy scenario (such as the strategy ID), the system retrieves the business execution result data corresponding to that strategy scenario from the pre-set data platform wide table. The data platform wide table pre-integrates execution data from multiple business systems, specifically including multi-dimensional business metrics such as distribution rate, click-through rate, and daily active user retention rate. During retrieval, the strategy ID and target customer group range can be used as filtering conditions to ensure that the obtained business execution result data accurately corresponds to the current strategy scenario. This application's embodiment, based on the pre-set wide table's strategy-data automated dynamic binding mechanism, can automatically select and retrieve performance observation metrics strongly correlated with the strategy from the underlying massive pre-set business metric wide table according to the currently executed strategy rules.

[0066] Finally, the parsed structured configuration parameters and the retrieved business execution result data are bound one-to-one using the unique identifier of the policy scenario as the association key, forming a bound dataset with a single policy scenario as the dimension, i.e., the bound current policy scenario data. This bound dataset contains both the "cause" (configuration parameters) and the "effect" (execution result) of the policy, providing a complete data input foundation for the quantitative evaluation and logical attribution of multi-agent collaboration in subsequent steps.

[0067] This application embodiment monitors the running status of the SOP rule engine in real time through the system interface, automatically triggers the parsing of configuration parameters and the retrieval of business execution result data, and accurately binds the two based on the unique policy identifier. This realizes the dynamic and automated mapping of policy feature data and performance data, and completes the entire link from policy execution to data readiness without manual intervention. It effectively avoids the problems of poor timeliness and high mismatch rate caused by manual data export and matching in the traditional mode. It provides a timely, accurate and structured data foundation for subsequent intelligent evaluation and attribution analysis, thereby ensuring the real-time performance and reliability of the entire policy evaluation and optimization process.

[0068] In some embodiments of this application, the intelligent agent includes a quantitative analysis intelligent agent, and the quantitative evaluation result includes the pure incremental value corresponding to the current strategy scenario and the scores of each funnel stage. The step of using a multi-agent collaborative evaluation and attribution mechanism to perform quantitative evaluation and logical attribution on the bound current strategy scenario data to obtain the quantitative evaluation result and logical attribution result of the current strategy scenario includes: using the quantitative analysis intelligent agent to extract virtual control group data from an unreached silent customer group, and using the bound current strategy scenario data as experimental group data; comparing the experimental group data and the virtual control group data to calculate the pure incremental value corresponding to the current strategy scenario; and scoring each funnel stage under the current strategy scenario to obtain the score of each funnel stage under the current strategy scenario.

[0069] First, after receiving the current strategy scenario data after binding, the quantitative analysis agent marks the corresponding reached user group in the data as experimental group data. The experimental group data includes complete business execution results such as the delivery rate, click-through rate, and daily active user retention rate of the target customer group after receiving the strategy.

[0070] Then, the quantitative analysis agent extracts virtual control group data from the silent customer group that was not reached by the strategy at the bottom layer using a propensity score matching (PSM) algorithm. Specifically, the quantitative analysis agent uses the multidimensional characteristics of the experimental group users (such as user activity, historical consumption level, login frequency, etc.) as matching variables to calculate the propensity score of each silent customer group user to be reached, and performs one-to-one or stratified matching with the experimental group users according to the propensity score. In this way, a group of users who are highly similar to the experimental group users in terms of feature distribution but have not been reached by the strategy are selected from the silent customer group. The business performance data of this group is used as virtual control group data. The construction of the virtual control group simulates "the natural business state that the user group should have exhibited if the strategy had not been implemented", providing a reliable counterfactual benchmark for causal inference.

[0071] Subsequently, the quantitative analysis agent compared the experimental group data with the virtual control group data one by one across various business metrics, such as delivery rate, click-through rate, and daily active user retention rate. By calculating the difference between the experimental group's metric values ​​and the virtual control group's metric values, the uplift score corresponding to the current strategy scenario was obtained. This uplift score eliminates the basic conversion brought about by users' natural behavior and truly reflects the net incremental contribution brought about by the strategy itself.

[0072] Finally, the quantitative analysis agent independently scores each funnel stage in the current strategy scenario. Specifically, the strategy execution path is divided into multiple funnel stages such as "reach-click-retention," and the incremental performance of each stage relative to the virtual control group is calculated. Combined with preset scoring rules such as incremental magnitude and statistical significance, each funnel stage is assigned a quantitative score, thus obtaining the score for each funnel stage in the current strategy scenario. The aforementioned pure incremental value and the scores of each funnel stage together constitute the quantitative evaluation result of the current strategy scenario, serving as the input basis for the subsequent logical attribution agent to perform parameter-level root cause diagnosis.

[0073] In addition to using the PSM matching algorithm based on the experimental group-virtual control group mentioned above, if the business system supports it, the Uplift incremental calculation can also be performed directly using the A / B Test experimental group / control group data randomly distributed from the front end.

[0074] This application embodiment uses a propensity matching algorithm to construct a virtual control group with highly similar characteristics to the experimental group from an unreached, silent customer base through quantitative analysis of the intelligent agent. Based on the comparison between the experimental group and the virtual control group, the pure incremental value brought by the strategy is calculated. Simultaneously, each funnel stage in the strategy execution path is independently scored, thus overcoming the blind spot of traditional reports that rely solely on absolute conversion rates to evaluate strategy effectiveness. It uses a scientific causal inference method to truly eliminate the interference of natural user behavior, accurately measuring the actual net incremental business value of the strategy. Furthermore, the refined scoring of the funnel stages further identifies the strength of the strategy's performance at each conversion node, providing a precise and quantifiable evaluation basis for subsequent logical attribution and parameter optimization.

[0075] In some embodiments of this application, the intelligent agent includes a logical attribution intelligent agent. The step of using a multi-agent collaborative evaluation and attribution mechanism to quantitatively evaluate and logically attribute the bound current strategy scenario data to obtain the quantitative evaluation result and logical attribution result of the current strategy scenario includes: calculating the comprehensive score of the current strategy scenario based on the scores of each funnel stage under the current strategy scenario; if the comprehensive score of the current strategy scenario is lower than a preset score threshold, using the logical attribution intelligent agent to perform logical attribution on the bound current strategy scenario data to obtain the logical attribution result of the current strategy scenario.

[0076] The logical attribution agent receives the scores of each funnel stage in the current strategy scenario from the quantitative analysis agent, and calculates the comprehensive score of the current strategy scenario by weighting and aggregating the scores of each funnel stage according to a preset comprehensive scoring rule. Specifically, the comprehensive scoring rule can assign different weight coefficients to different funnel stages according to business priorities. For example, the funnel stage corresponding to "daily active user retention rate" can be given a higher weight, while the funnel stage corresponding to "distribution rate" can be given a lower weight. The comprehensive score is obtained by weighted summation, thereby realizing the quantification of a single score for the overall performance of the strategy. Of course, those skilled in the art can flexibly set the weight coefficients of each stage according to the needs of the actual business scenario, and no specific limitation is made here.

[0077] The logical attribution agent compares the calculated comprehensive score with a preset score threshold. When the comprehensive score is lower than the preset score threshold, it indicates that the overall performance of the strategy scenario has not met expectations, and the deep diagnostic process of the logical attribution agent is triggered. When the comprehensive score is higher than or equal to the preset threshold, the strategy performance is considered qualified, and no further logical attribution is required; the current quantitative evaluation result is directly output.

[0078] This application embodiment automatically intervenes when the overall strategy score is lower than a preset threshold by using a logical attribution agent to perform cross-logical diagnosis. This changes the inefficient mode of relying on data analysts to manually write SQL and check data one by one to find the cause. It realizes millisecond-level automatic location from abnormal overall strategy score to root cause at the business parameter level, which greatly improves the diagnostic efficiency and location accuracy of strategy problems. It provides clear and executable attribution basis for generating targeted structured optimization suggestions.

[0079] In some embodiments of this application, the step of using the logical attribution agent to perform logical attribution on the bound current policy scenario data to obtain the logical attribution result of the current policy scenario includes: determining the low-scoring funnel links based on the scores of each funnel link in the current policy scenario; and using the logical attribution agent to perform cross-logical diagnosis on the low-scoring funnel links and the structured configuration parameters of the policy scenario to obtain the logical attribution result of the current policy scenario.

[0080] The logical attribution agent receives the scores of each funnel stage in the current strategy scenario from the quantitative analysis agent, and compares the score of each funnel stage with a preset stage score threshold. It then selects one or more funnel stages with scores below the threshold and identifies them as low-scoring funnel stages. For example, if the score of the "click-to-conversion" stage is below the stage's score threshold, that stage is identified as a low-scoring funnel stage, indicating a significant user churn problem at that conversion node.

[0081] The logical attribution agent uses the identified low-scoring funnel stage as the diagnostic anchor point. It retrieves the structured configuration parameters of the strategy scenario extracted and bound in the aforementioned embodiments, including various parameter fields such as target customer group, triggering conditions, and push materials. It then performs cross-logical diagnosis between the business problems reflected in the low-scoring funnel stage and each configuration parameter. Specifically, the logical attribution agent, based on the business behavior characteristics corresponding to the low-scoring funnel stage and the business meaning of each configuration parameter, uses logical methods such as time series analysis and feature correlation reasoning to match and investigate each one. For example, when the low-scoring funnel stage is "click-to-conversion drop-off," the logical attribution agent first analyzes the time dimension characteristics of this drop-off. Through time series analysis, it finds that the reached user is already in a low-activity peak period on T+3 days after receiving the push notification, and the probability of the user opening the application is significantly lower than on T+1 day. Therefore, the problem is pointed to the "triggering condition" configuration parameter, which is determined to be unreasonable and the core reason for the low score of this funnel stage.

[0082] Based on the entire reasoning process of the above cross-logic diagnosis, the logical attribution agent performs logical back-reasoning on the low-scoring strategy in conjunction with the original SOP configuration parameters, converges layer by layer, accurately identifies one or more core parameters that cause the overall effectiveness of the strategy to decline, and generates logical attribution results containing root cause localization conclusions. It clearly points out whether the problem lies in time parameters, frequency parameters, or customer group parameters, providing accurate diagnostic input for the subsequent optimization agent to output targeted and structured optimization suggestions.

[0083] This application embodiment uses a logical attribution agent to take the low-score funnel as the diagnostic anchor point and performs cross-logical diagnosis with the structured configuration parameters of the strategy scenario. By using methods such as time series analysis and logical back-inference, it traces back from the abnormal link in the strategy execution result to the specific configuration parameter level, accurately locating the core parameters that cause the strategy effectiveness to decline. This changes the inefficient mode of relying on data analysts to manually write SQL and check massive amounts of data one by one to find the cause. It realizes the automated and accurate location of the root cause from the abnormal comprehensive score of the strategy to the business parameter level, which greatly improves the diagnostic efficiency and location accuracy of strategy problems. This enables operations personnel to quickly identify the optimization direction and provides a clear and executable attribution basis for generating structured optimization suggestions and driving parameter iteration.

[0084] In some embodiments of this application, the visualization performance dashboard includes at least one of a dynamic incremental funnel, an interpretable attribution module, and a strategy health radar chart; the dynamic incremental funnel is used to compare and display the experimental group funnel and the virtual control group funnel in a parallel manner; the interpretable attribution module is used to visualize the logical attribution results; and the strategy health radar chart is used to comprehensively display the health scores of multi-dimensional indicators.

[0085] The visualization performance dashboard mainly includes a dynamic incremental funnel, an interpretable attribution module, and a strategy health radar chart.

[0086] Regarding the dynamic incremental funnel, based on the experimental group data (i.e., conversion data at each stage of the funnel where the strategy actually reaches users) and the virtual control group data (i.e., data at each stage of the control group constructed using the PSM algorithm from unreached silent customers), two independent funnel views are rendered simultaneously in the dashboard in a parallel manner. The experimental group funnel displays the actual conversion rate at each stage from "distribution-click-retention" after the strategy is implemented, while the virtual control group funnel displays the natural conversion rate of the same type of users without receiving strategy outreach. The two funnels are presented in parallel, either side-by-side or stacked vertically, allowing operations personnel to intuitively compare the incremental difference between the experimental group and the control group at each stage. This allows them to immediately identify at which stage the strategy brings positive incremental growth and at which stage it even produces negative effects, intuitively highlighting the pure incremental value of the strategy.

[0087] Regarding the explainable attribution module, based on the logical attribution results output by the logical attribution agent, the root cause localization conclusions are presented in an explainable and visual form. Specifically, this module presents the diagnostic logic of the attribution agent in a structured manner using highlighted text, decision trees, or causal path diagrams. For example, when the attribution result determines that "the unreasonable setting of the trigger condition T+3 days leads to a sharp increase in bounce rate," this module marks the parameter field "Trigger condition: T+3 days" with a red highlighted box or warning color block in the dashboard, and displays the complete attribution reasoning chain from "low-scoring funnel stage (click-to-conversion churn)" to "time series analysis found that T+3 days is the peak of inactivity" to "unreasonable trigger condition parameter" using the branch path of a decision tree. This allows operations personnel not only to see which parameter is problematic, but also to clearly understand the reason why this parameter is problematic, achieving transparency and explainability in the attribution process.

[0088] The strategy health radar chart comprehensively displays the scores of each funnel stage in the quantitative evaluation results, as well as the multi-dimensional indicators involved in the overall strategy score (such as launch rate, click-through rate, retention rate, and uplift score). Each axis of the radar chart corresponds to an evaluation dimension, and the value on the axis reflects the health score of that dimension. The area of ​​the polygon formed by connecting the axes visually reflects the overall health of the strategy. When the score of a certain dimension is low, the radar chart will be noticeably concave in the direction of that axis, allowing operations personnel to easily identify the weaker dimensions of the strategy and quickly grasp its overall health status.

[0089] Finally, the content of the aforementioned visualization dashboard is combined and integrated into a dashboard page on the web or client side as needed, forming a customized visualization dashboard for the current strategy scenario. When the strategy scenario data is updated, the content of each module of the dashboard is refreshed in real time to ensure that the displayed content remains synchronized with the latest evaluation and attribution results.

[0090] This application's embodiments visually demonstrate the transformation differences between the experimental group and the virtual control group at each funnel stage through a dynamic incremental funnel in a parallel comparison manner, enabling operators to instantly identify the pure incremental value of the strategy. The interpretable attribution module presents the logical attribution reasoning chain transparently in the form of highlighted text or decision trees, making the root cause localization conclusion not only known but also understandable. A strategy health radar chart comprehensively displays the health scores of multiple dimensions, making the overall shortcomings of the strategy readily apparent. The collaboration of these three elements transforms the processing results of multiple agents into an intuitive and interactive visual language, upgrading traditional one-way data reports into a comprehensive decision dashboard with incremental comparison, attribution explanation, and health diagnosis capabilities. This significantly lowers the cognitive threshold for operators to understand the effectiveness and problems of the strategy, providing a clear decision-making basis for subsequent parameter optimization operations at the dashboard level.

[0091] In some embodiments of this application, the step of outputting structured optimization suggestions using a preset optimization agent based on the logical attribution results of the current strategy scenario, and optimizing and updating the structured configuration parameters of the current strategy scenario using the structured optimization suggestions based on the visualized performance dashboard, includes: retrieving historical high-performance strategies from a strategy knowledge base using the preset optimization agent based on the logical attribution results of the current strategy scenario; outputting the structured optimization suggestions based on the historical high-performance strategies; generating parameter tuning instruction cards based on the structured configuration parameters of the current strategy scenario and the structured optimization suggestions, and displaying them on the visualized performance dashboard, so that users can optimize and update the structured configuration parameters of the current strategy scenario through the parameter tuning instruction cards in the visualized performance dashboard.

[0092] The optimization agent receives the logical attribution results output by the logical attribution agent. These results clearly identify the core parameter issues causing the current strategy's effectiveness to decline, such as "the trigger condition T+3 day setting is unreasonable." The optimization agent uses these as search criteria to perform a matching search in a pre-built strategy knowledge base. This knowledge base pre-stores historically executed high-performance strategies with high effectiveness evaluation scores. Each record contains a complete set of structured configuration parameters and their corresponding quantitative evaluation results. The optimization agent uses a feature matching algorithm to filter out historical high-performance strategies from the knowledge base that are similar to the current strategy in dimensions such as target customer type and pushed material type, and whose overall score is significantly higher than the current strategy.

[0093] The optimization agent compares the configuration parameters of historically high-performing strategies with the structured configuration parameters of the current strategy item by item, identifies the parameters that differ between the two, and combines this with the core problem parameters located in the logical attribution results to generate structured optimization suggestions. These suggestions can be explicitly output in key-value pairs, including the name of the parameter field to be modified, the current parameter value, the suggested adjusted parameter value, and the expected improvement in performance calculated based on historical data from the knowledge base. For example, the optimization suggestion output might be: parameter field "Trigger Condition", current value "T+3 days", suggested value "T+1 days", expected click-through rate increase of approximately 15%.

[0094] Based on the structured configuration parameters of the current strategy scenario and the aforementioned structured optimization suggestions, a "Parameter Tuning Instruction Card" is dynamically generated and pops up in the visualization dashboard. This card displays several core elements: first, the original parameter configuration of the current strategy is displayed in read-only format; second, the target parameters for optimizing the agent are highlighted or contrasted; and third, the expected performance improvement data after applying the optimization suggestion is displayed. The parameter tuning instruction card can be presented in the dashboard interface as a floating pop-up or sidebar, allowing users to interact directly at the dashboard level.

[0095] After viewing the parameter tuning instruction card in the visualization dashboard, users can click the "Confirm Application" button on the card. In response to this confirmation, the system sends the parameter tuning instruction containing the optimized parameter values ​​back to the SOP rule engine's configuration backend via a preset API interface. This automatically overwrites the original strategy configuration rules, completing the optimization and update of the structured configuration parameters for the current strategy scenario, thus achieving a complete business loop from data evaluation to display to optimization.

[0096] In addition to the above-mentioned interaction methods, for strategy parameter modifications that the system has pre-determined to be low-risk (such as minor adjustments to pushed materials), it can be configured to automatically pass the preset threshold acceptance logic and be directly applied after the optimization suggestion is generated, without requiring the user to manually click to confirm on the Kanban board, thereby achieving fully unattended adaptive strategy optimization (Auto-Tuning).

[0097] This application embodiment optimizes the intelligent agent by retrieving historical high-efficiency strategies from the strategy knowledge base based on logical attribution results. It automatically generates structured optimization suggestions containing current parameters, suggested parameters, and expected performance improvement. The intelligent agent then interactively displays these suggestions in the form of parameter tuning instruction cards on a visual performance dashboard. Users can confirm the optimization instructions with a single click at the dashboard level, automatically updating the parameters by sending them back to the SOP configuration backend. It also supports automatic acceptance and unmanned application of low-risk parameters. This upgrades the traditional data dashboard from a one-way display tool to a two-way interactive console that can directly issue optimization instructions, realizing a complete business closed loop of "evaluation-display-optimization" and greatly shortening the strategy iteration cycle.

[0098] In some alternative implementations, the aforementioned analytical agent, attribution agent, and optimization agent can be decoupled into three independently deployed agent modules, which work together sequentially to complete the entire process; alternatively, they can be merged into a large-scale integrated agent with Chain of Thought (CoT) capabilities, which completes the entire reasoning process from data extraction, incremental analysis, logical attribution to outputting optimization suggestions in one go, reducing communication overhead between agents and improving overall processing efficiency.

[0099] This application embodiment also provides a strategy evaluation and optimization device 200, such as Figure 2 The diagram shows a structural schematic of a strategy evaluation and optimization device according to an embodiment of this application. The strategy evaluation and optimization device 200 includes:

[0100] The acquisition unit 210 is used to acquire the structured configuration parameters and corresponding business execution result data of the current strategy scenario and bind the data to obtain the bound current strategy scenario data;

[0101] The evaluation and attribution unit 220 is used to perform quantitative evaluation and logical attribution on the bound current policy scenario data using a multi-agent collaborative evaluation and attribution mechanism, so as to obtain the quantitative evaluation result and logical attribution result of the current policy scenario.

[0102] The visualization unit 230 is used to generate and display a visualization performance dashboard based on the quantitative evaluation results and logical attribution results of the current strategy scenario.

[0103] The optimization unit 240 is used to output structured optimization suggestions using a preset optimization agent based on the logical attribution results of the current strategy scenario, and to optimize and update the structured configuration parameters of the current strategy scenario based on the structured optimization suggestions using the visualization results dashboard.

[0104] In some embodiments of this application, the acquisition unit 210 is specifically used for: real-time monitoring of the running status of the strategy scenario in the SOP rule engine; when the running status of the strategy scenario is detected as running or completing a cycle, parsing the structured configuration parameters of the strategy scenario; pulling the business execution result data corresponding to the strategy scenario from the preset data platform wide table; binding the structured configuration parameters of the strategy scenario with the corresponding business execution result data to obtain the bound current strategy scenario data.

[0105] In some embodiments of this application, the intelligent agent includes a quantitative analysis intelligent agent, and the quantitative evaluation result includes the pure incremental value corresponding to the current strategy scenario and the scores of each funnel stage. The evaluation and attribution unit 220 is specifically used to: use the quantitative analysis intelligent agent to extract virtual control group data from an unreached silent customer group, and use the bound current strategy scenario data as experimental group data; compare the experimental group data and the virtual control group data to calculate the pure incremental value corresponding to the current strategy scenario; and score each funnel stage under the current strategy scenario to obtain the score of each funnel stage under the current strategy scenario.

[0106] In some embodiments of this application, the intelligent agent includes a logical attribution intelligent agent, and the evaluation and attribution unit 220 is specifically used to: calculate the comprehensive score of the current strategy scenario based on the scores of each funnel stage under the current strategy scenario; when the comprehensive score of the current strategy scenario is lower than a preset score threshold, use the logical attribution intelligent agent to perform logical attribution on the bound current strategy scenario data to obtain the logical attribution result of the current strategy scenario.

[0107] In some embodiments of this application, the evaluation and attribution unit 220 is specifically used to: determine the low-scoring funnel links based on the scores of each funnel link in the current strategy scenario; and use the logical attribution agent to perform cross-logical diagnosis on the low-scoring funnel links and the structured configuration parameters of the strategy scenario to obtain the logical attribution result of the current strategy scenario.

[0108] In some embodiments of this application, the visualization performance dashboard includes at least one of a dynamic incremental funnel, an interpretable attribution module, and a strategy health radar chart; the dynamic incremental funnel is used to compare and display the experimental group funnel and the virtual control group funnel in a parallel manner; the interpretable attribution module is used to visualize the logical attribution results; and the strategy health radar chart is used to comprehensively display the health scores of multi-dimensional indicators.

[0109] In some embodiments of this application, the optimization unit 240 is specifically used to: retrieve historically high-efficiency strategies from the strategy knowledge base using the preset optimization agent based on the logical attribution results of the current strategy scenario; output the structured optimization suggestions based on the historically high-efficiency strategies; generate parameter tuning instruction cards based on the structured configuration parameters of the current strategy scenario and the structured optimization suggestions, and display them on the visualization results dashboard, so that users can optimize and update the structured configuration parameters of the current strategy scenario through the parameter tuning instruction cards in the visualization results dashboard.

[0110] It is understood that the above-described strategy evaluation and optimization apparatus can implement each step of the strategy evaluation and optimization method provided in the foregoing embodiments. The relevant explanations of the strategy evaluation and optimization method are applicable to the strategy evaluation and optimization apparatus, and will not be repeated here.

[0111] Figure 3 This is a schematic diagram of the structure of a device according to an embodiment of this application. For example... Figure 3 As shown, the device includes one or more processors (or processing units), and may also include one or more memories coupled to the processors, and may also include a communication module coupled to the processors.

[0112] A communication module can be used to communicate with other devices or apparatuses, such as sending or receiving data and / or signals. A communication module may have at least one communication module for communication. A communication module may include any interface necessary for communicating with other devices. Exemplarily, a communication module may be a transceiver, circuit, bus, module, or other type of communication module.

[0113] The processor may include, but is not limited to, one or more of the following: a general-purpose computer, a special-purpose computer, a microcontroller, a digital signal processor (DSP), or a controller-based multi-core controller architecture. The device may have multiple processors, such as application-specific integrated circuit (ASIC) chips, which are time-dependent on a clock synchronized with the main processor.

[0114] The memory may include one or more non-volatile memories and one or more volatile memories. Examples of non-volatile memories include, but are not limited to, at least one of the following: read-only memory (ROM), electrically programmable read-only memory (EPROM), flash memory, hard disk, compact disc (CD), digital video disc (DVD), or other magnetic and / or optical storage. Examples of volatile memories include, but are not limited to, at least one of the following: random access memory (RAM), or other volatile memories that do not persist during the duration of a power outage.

[0115] A computer program consists of computer-executable instructions that are executed by an associated processor. Programs can be stored in ROM. A processor can perform any appropriate action and processing by loading the program into RAM.

[0116] Possible implementations of this application can be achieved through a program, enabling the communication device to execute any of the processes discussed in the foregoing embodiments. Possible implementations of this application can also be achieved through hardware or a combination of software and hardware.

[0117] In some implementations, the program may be tangibly contained in a computer-readable storage medium, which may include in a device (such as in memory) or other storage device accessible by the device. The program may be loaded from the computer-readable storage medium into RAM for execution. The computer-readable storage medium may include any type of tangible non-volatile memory, such as ROM, EPROM, flash memory, hard disk, CD, DVD, etc.

[0118] This application also provides a computer-readable storage medium storing computer instructions or program code thereon, which, when executed by a processor, causes the processor to perform the methods and functions involved in any of the above embodiments. A computer-readable medium can be any tangible medium that contains or stores a program for or relating to an instruction execution system, apparatus, or device. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. More detailed examples of computer-readable storage media include electrical connections with one or more wires, magnetic media (e.g., disks, floppy disks, hard disks, magnetic tapes, magnetic storage devices), optical media (e.g., optical storage devices, DVDs), semiconductor media (e.g., solid-state drives), random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), or any suitable combination thereof.

[0119] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. Embodiments of this application also provide at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. This computer program product includes one or more computer-executable instructions, such as instructions included in a program module, which execute in a device on a target real or virtual processor to perform the processes, methods, and functions involved in any of the above embodiments. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0120] This application also proposes a computer program product, including a computer program or instructions that, when run on a computer, cause the computer to perform the processes, methods, and functions described in the above embodiments. Typically, program modules include routines, programs, libraries, objects, classes, components, data structures, etc., that perform specific tasks or implement specific abstract data types. In various embodiments, the functionality of program modules can be combined or divided as needed. The machine-executable instructions for the program modules can be executed locally or in a distributed device. In a distributed device, the program modules can reside in both local and remote storage media.

[0121] Generally, the various embodiments of this application can be implemented in hardware or dedicated circuitry, software, logic, or any combination thereof. Some aspects can be implemented in hardware, while others can be implemented in firmware or software, which can be executed by a controller, microprocessor, or other computing device. Although various aspects of the embodiments of this disclosure are shown and described as block diagrams, flowcharts, or represented using some other illustration, it should be understood that the blocks, apparatuses, systems, techniques, or methods described herein can be implemented as, as non-limiting examples, in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or controllers or other computing devices, or some combination thereof.

[0122] It should be noted that although embodiments of this application have been described above with reference to the accompanying drawings, these embodiments are not independent of each other, and they can be combined to obtain other embodiments. The methods, situations, categories, and classifications of embodiments in this application are only for the convenience of description and should not constitute a special limitation. Various methods, categories, situations, and features in embodiments can be combined with each other if logically consistent. The various embodiments of this application can be arbitrarily combined to achieve different technical effects. The embodiments of this application will not list various combinations.

[0123] Furthermore, although the operation of the methods of this disclosure is described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all of the operations shown must be performed to achieve the desired result. Rather, the steps depicted in the flowcharts may be performed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps. It should also be noted that the features and functions of two or more devices according to this disclosure may be embodied in one device. Conversely, the features and functions of one device described above may be further divided and embodied by multiple devices.

[0124] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0125] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A strategy evaluation and optimization method, characterized in that, The strategy evaluation and optimization methods include: Obtain the structured configuration parameters and corresponding business execution result data of the current policy scenario and bind the data to obtain the bound current policy scenario data; The data of the current policy scenario after binding is quantitatively evaluated and logically attributed using an evaluation and attribution mechanism based on multi-agent collaboration, so as to obtain the quantitative evaluation results and logical attribution results of the current policy scenario. A visual performance dashboard is generated and displayed based on the quantitative evaluation results and logical attribution results of the current strategy scenario. Based on the logical attribution results of the current strategy scenario, a pre-defined optimization agent outputs structured optimization suggestions, and based on the visualized results dashboard, the structured configuration parameters of the current strategy scenario are optimized and updated using the structured optimization suggestions.

2. The strategy evaluation and optimization method according to claim 1, characterized in that, The process of obtaining the structured configuration parameters and corresponding business execution result data of the current policy scenario and binding them to obtain the bound current policy scenario data includes: Real-time monitoring of the running status of policy scenarios in the SOP rule engine; If the running status of the policy scenario is detected as running or completing a cycle, the structured configuration parameters of the policy scenario are parsed. Retrieve the business execution result data corresponding to the strategy scenario from the pre-set data platform wide table; The structured configuration parameters of the strategy scenario are bound to the corresponding business execution result data to obtain the current strategy scenario data after binding.

3. The strategy evaluation and optimization method according to claim 1, characterized in that, The intelligent agent includes a quantitative analysis intelligent agent. The quantitative evaluation result includes the pure incremental value corresponding to the current policy scenario and the scores of each funnel stage. The quantitative evaluation and logical attribution of the bound current policy scenario data using a multi-agent collaborative evaluation and attribution mechanism to obtain the quantitative evaluation result and logical attribution result of the current policy scenario include: Using the aforementioned quantitative analysis agent, virtual control group data is extracted from the unreached silent customer group, and the bound current strategy scenario data is used as experimental group data. By comparing the experimental group data and the virtual control group data, the pure incremental value corresponding to the current strategy scenario is calculated; Each funnel stage in the current strategy scenario is scored to obtain a score for each funnel stage in the current strategy scenario.

4. The strategy evaluation and optimization method according to claim 3, characterized in that, The intelligent agent includes a logical attribution intelligent agent. The quantitative evaluation and logical attribution of the bound current policy scenario data using a multi-agent collaborative evaluation and attribution mechanism, to obtain the quantitative evaluation result and logical attribution result of the current policy scenario, includes: The overall score of the current strategy scenario is calculated based on the scores of each funnel stage under the current strategy scenario. If the overall score of the current strategy scenario is lower than a preset score threshold, the logical attribution agent is used to perform logical attribution on the bound current strategy scenario data to obtain the logical attribution result of the current strategy scenario.

5. The strategy evaluation and optimization method according to claim 4, characterized in that, The step of using the logical attribution agent to perform logical attribution on the bound current policy scenario data to obtain the logical attribution result of the current policy scenario includes: The low-scoring funnel stage is determined based on the scores of each funnel stage in the current strategy scenario. Using the logical attribution agent, cross-logical diagnosis is performed on the structured configuration parameters of the low-score funnel and the policy scenario to obtain the logical attribution result of the current policy scenario.

6. The strategy evaluation and optimization method according to claim 1, characterized in that, The visualization results dashboard includes at least one of a dynamic incremental funnel, an interpretable attribution module, and a strategy health radar chart; the dynamic incremental funnel is used to compare and display the experimental group funnel and the virtual control group funnel in a parallel manner; the interpretable attribution module is used to visualize the logical attribution results; and the strategy health radar chart is used to comprehensively display the health scores of multi-dimensional indicators.

7. The strategy evaluation and optimization method according to claim 1, characterized in that, The step of using a preset optimization agent to output structured optimization suggestions based on the logical attribution results of the current strategy scenario, and using the structured optimization suggestions to optimize and update the structured configuration parameters of the current strategy scenario based on the visualized performance dashboard, includes: Based on the logical attribution results of the current strategy scenario, the preset optimization agent is used to search the strategy knowledge base to obtain historically high-efficiency strategies. Output the structured optimization suggestions based on the historically high-efficiency strategies; Based on the structured configuration parameters of the current strategy scenario and the structured optimization suggestions, parameter tuning instruction cards are generated and displayed in the visualization results dashboard, so that users can optimize and update the structured configuration parameters of the current strategy scenario through the parameter tuning instruction cards in the visualization results dashboard.

8. A strategy evaluation and optimization device, characterized in that, The strategy evaluation and optimization device includes: The acquisition unit is used to acquire the structured configuration parameters and corresponding business execution result data of the current policy scenario and bind the data to obtain the bound current policy scenario data; The evaluation and attribution unit is used to perform quantitative evaluation and logical attribution on the bound current policy scenario data using a multi-agent collaborative evaluation and attribution mechanism, so as to obtain the quantitative evaluation result and logical attribution result of the current policy scenario. The visualization unit is used to generate and display a visualization performance dashboard based on the quantitative evaluation results and logical attribution results of the current strategy scenario. The optimization unit is used to output structured optimization suggestions using a preset optimization agent based on the logical attribution results of the current strategy scenario, and to optimize and update the structured configuration parameters of the current strategy scenario based on the structured optimization suggestions using the visualization results dashboard.

9. An apparatus comprising: processor; And a memory configured to store computer-executable instructions, which, when executed, cause the processor to perform the strategy evaluation and optimization method according to any one of claims 1 to 7.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the strategy evaluation and optimization method according to any one of claims 1 to 7.

Images