An in-app advertisement triggering dynamic adjustment method and system based on user churn risk score

Abstract

The application relates to the field of Internet application advertisement putting, in particular to an application-in-advertisement triggering dynamic adjustment method and system based on user loss risk score, which comprises the following steps: a user enters an application and generates an operation behavior, user behavior data, advertisement history data and application context data are collected; when the application is ready to trigger an advertisement, the current loss risk score of the user is calculated; according to the loss risk score, the user is divided into four risk levels; advertisement triggering strategies are executed for different risk levels; user behavior data after the advertisement triggering is collected, and it is judged whether the user continues to use the application or completes the current task; according to the result of whether the user continues to use or completes the task, the weight of the corresponding advertisement strategy is adjusted, and the risk score parameter and the advertisement triggering strategy are updated; before the next advertisement triggering, the above evaluation process is re-executed, and the application can guarantee user experience and optimize user retention rate.

Description

Technical Field

[0001] This invention relates to the field of internet application advertising, and in particular to a method and system for dynamically adjusting in-application ad triggers based on user churn risk scores. Background Technology

[0002] In-app advertising is currently the mainstream monetization method in mobile applications, mini-games, content communities, utility software, short video applications, live streaming applications, and e-commerce applications. Common ad formats include rewarded ads, interstitial ads, banner ads, content recommendation ads, reward unlock ads, and task reward ads. Existing ad delivery systems generally use fixed rules to trigger ads, that is, to execute ad display actions according to fixed pages, fixed time intervals, fixed number of operations, fixed level nodes, fixed task nodes, or fixed ad positions. For example, in mini-game scenarios, the system often displays rewarded ads at nodes such as when the user completes a level, fails the game, the character revives, or claims a reward; in news applications, interstitial ads usually pop up after the user reads a specified number of articles; in utility applications, reward unlock ads are often displayed before the user activates advanced features; and in short video and content applications, ad content is inserted into the content stream at fixed intervals. The above fixed ad triggering methods have simple implementation logic, but they have many shortcomings in actual use, as follows: Fixed ad trigger rules cannot identify the risk of real-time user churn. Users may be in the initial stage of use, on a critical task execution path, experiencing consecutive operation failures, staying on the page for a long time, or about to exit the application. Forcibly displaying ads in these situations can easily cause users to interrupt their current operation or even close the application directly.

[0003] A single, fixed advertising strategy cannot adapt to the diverse user states. Users with low churn risk can accept regular advertising content normally, users with medium churn risk are more suited to help ads related to the current task, and users with high churn risk need to have their ad display delayed or their ad push frequency reduced. Existing fixed strategies cannot provide differentiated control for different user states.

[0004] Current advertising strategies only focus on metrics such as ad impressions, click-through rates, and ad revenue, without establishing a closed-loop feedback mechanism for user behavior after ad display. Data on user behavior after ad display—such as whether they continue using the app, complete the current task, exit the app, or revisit the next day—is not included in subsequent updates and iterations of the ad triggering strategy.

[0005] New users and critical task paths are easily interrupted by ads. Showing ads too early to new users before they have received positive feedback will seriously affect the app's first-day retention rate. When users are in states such as before payment, before task submission, the first game, key points of content reading, or before tool operation is completed, ad interruption will directly reduce task completion rate.

[0006] Improper ad triggering timing or excessive frequency can lead to ineffective ad displays, decreased ad completion rates, and reduced overall user retention, severely impacting the app's user experience. Conversely, excessively low ad triggering frequency can weaken the app's monetization revenue. Current technology cannot achieve a dynamic balance between user retention and ad triggering.

[0007] Therefore, the industry needs a technical solution that can identify user churn risk before ad triggering and dynamically adjust ad triggering strategies based on user risk status. Summary of the Invention

[0008] To overcome the problems of existing technologies, such as inability to identify user churn risk, single strategy, lack of closed-loop feedback, easy interruption of user operations, and low user retention rate, the purpose of this invention is to provide a method and system for dynamically adjusting in-app ad triggering based on user churn risk scoring. By collecting multi-dimensional data to quantitatively calculate user churn risk and complete risk stratification, differentiated ad triggering strategies are configured for different risk levels. At the same time, the strategy is dynamically updated by combining user behavior data after ad triggering, ultimately ensuring user experience and optimizing user retention rate.

[0009] This invention is implemented using the following scheme: A method for dynamically adjusting in-app ad triggers based on user churn risk scoring, comprising the following steps: The method steps are as follows: Step 1: Users enter the application and perform actions; user behavior data, advertising history data, and application context data are collected. Step 2: When the application is ready to trigger an ad, calculate the user's current churn risk score based on the collected user behavior data, ad history data, and application context data; Step 3: Based on the churn risk score, users are divided into four risk levels: low risk, medium risk, high risk, and very high risk. Step 4: Implement ad triggering strategies for different risk levels. The ad triggering strategies are as follows: Display monetization ads to low-risk users normally and record ad trigger results; Show helpful or benefit-based ads related to the current task to users at medium risk, and record the ad trigger results; Delay ad display or reduce ad frequency for high-risk users, and record the reasons for delay or downgrade; For users with extremely high risk, suspend ad display and prioritize protecting their current operations, and record the reason for suspension and the protection strategy; Step 5: Collect user behavior data after the ad is triggered to determine whether the user continues to use the application or completes the current task; Step 6: Adjust the weight of the corresponding advertising strategy based on whether the user continues to use the service or completes the task, and update the risk scoring parameters and advertising triggering strategy; Step 7: Before the next ad trigger, repeat the evaluation process from Step 2 to Step 6.

[0010] A dynamic adjustment system for in-app advertising triggers based on user churn risk scoring, the system comprising: The user behavior collection module is used to collect user behavior data, advertising history data, and application context data when a user enters the application and performs an operation. The churn risk scoring module is used to calculate the user's current churn risk score based on collected user behavior data, advertising history data, and application context data when the application is about to trigger an ad. The user risk stratification module is used to classify users into four risk levels: low risk, medium risk, high risk, and extremely high risk, based on the churn risk score. The advertising strategy matching module is used to execute advertising triggering strategies for different risk levels. The advertising triggering strategies are as follows: Display monetization ads to low-risk users normally and record ad trigger results; Show helpful or benefit-based ads related to the current task to users at medium risk, and record the ad trigger results; Delay ad display or reduce ad frequency for high-risk users, and record the reasons for delay or downgrade; For users with extremely high risk, suspend ad display and prioritize protecting their current operations, and record the reason for suspension and the protection strategy; The post-advertisement behavior feedback module is used to collect user behavior data after an advertisement is triggered, and to determine whether the user continues to use the application or completes the current task. The strategy update module is used to adjust the weight of the corresponding advertising strategy and update the risk score parameters and advertising triggering strategy based on whether the user continues to use the service or the result of completing the task. The evaluation module is used to re-execute the evaluation process of the above modules before the next ad trigger.

[0011] Furthermore, the user churn risk score is obtained by weighting multiple sub-scores, each corresponding to different user behavior states. Specifically, the operation obstruction risk score characterizes the degree of continuous user failures, repeated clicks, or task progress obstruction; the operation fatigue risk score characterizes the degree of prolonged user dwell time, ineffective operation, or decreased operation pace; the ad aversion risk score characterizes the degree of users closing ads, skipping ads, exiting after an ad, or a decrease in the rate of complete ad viewing; the new user vulnerability risk score characterizes the degree to which new users have not yet formed a stable usage path or have not yet received their first positive feedback; the critical task interruption risk score characterizes whether users are in critical processes that should not be interrupted, such as before payment, before submission, the first game, or before completing a task; and the activity and stability deduction score characterizes the degree of recent user activity, stable operation, or active acceptance of ads. The system calculates the final risk score based on the above scores and selects a strategy of normal display, benefit-based display, delayed frequency reduction, or suspension of ads based on the risk score.

[0012] The beneficial effects of this invention are as follows: This invention provides a method and system for dynamically adjusting in-app ad triggering based on user churn risk scoring. It can achieve quantitative identification of user churn risk, obtain a standardized churn risk score through multi-dimensional data weighted calculation, accurately identify the real-time churn status of users, abandon the traditional blind fixed ad triggering mode, and reduce invalid ad push from the source.

[0013] This invention categorizes user groups according to risk levels and configures different strategies such as normal display, task-related ads, delayed and reduced frequency ads, and paused ads, taking into account the user experience of users in different states and significantly reducing the probability of user churn caused by ads.

[0014] This invention comprehensively collects user behavior data throughout the entire advertising lifecycle, forming a traceable record of advertising events. It dynamically updates risk parameters, risk thresholds, and advertising strategies based on feedback data, enabling continuous iteration and optimization of advertising control capabilities. It implements advertising protection strategies for users at key nodes such as new users, payments, and task submissions, effectively improving new user retention rates and user task completion rates, and preventing excessive advertising frequency and inappropriate timing from damaging the application experience. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a structural block diagram of the system of the present invention. Detailed Implementation

[0016] The invention will now be further described with reference to the accompanying drawings.

[0017] See Figure 1A method for dynamically adjusting in-app ad triggers based on user churn risk scoring, the method comprising the following steps: Step 1: Users enter the application and perform actions; user behavior data, advertising history data, and application context data are collected. Step 2: When the application is ready to trigger an ad, calculate the user's current churn risk score based on the collected user behavior data, ad history data, and application context data; Step 3: Based on the churn risk score, users are divided into four risk levels: low risk, medium risk, high risk, and very high risk. Step 4: Implement ad triggering strategies for different risk levels. The ad triggering strategies are as follows: Display monetization ads to low-risk users normally and record ad trigger results; Show helpful or benefit-based ads related to the current task to users at medium risk, and record the ad trigger results; Delay ad display or reduce ad frequency for high-risk users, and record the reasons for delay or downgrade; For users with extremely high risk, suspend ad display and prioritize protecting their current operations, and record the reason for suspension and the protection strategy; Step 5: Collect user behavior data after the ad is triggered to determine whether the user continues to use the application or completes the current task; Step 6: Adjust the weight of the corresponding advertising strategy based on whether the user continues to use the service or completes the task, and update the risk scoring parameters and advertising triggering strategy; Step 7: Before the next ad trigger, repeat the evaluation process from Step 2 to Step 6.

[0018] The present invention will be further described below with reference to a specific embodiment: Step 1: After a user enters the application, continuously record the user's behavioral data within the application. Behavioral data includes, but is not limited to: current page or functional module, current operation path, current page dwell time, current session total duration, number of consecutive operation failures, time without operation, number of clicks on the back or close button, page switching frequency, current task completion progress, whether the user is in the novice stage, and whether the user is on a critical conversion path or critical task path.

[0019] Record user behavior related to ads, including but not limited to: the number of times ads were displayed that day; the time of the most recent ad display; the type of the most recent ad; whether the user clicked on the ad entry; whether the user watched the ad in its entirety; whether the user actively closed the ad; whether the user claimed ad rewards; whether the user continued to use the app after the ad was triggered; whether the user exited the app after the ad was triggered; and the historical ad completion rate.

[0020] Record user source and current environment, such as natural access, search entry, sharing entry, short video entry, push entry, activity page entry, repeat visit entry, current network status, device performance status, page loading time, etc.

[0021] Step 2: When the application is ready to trigger an ad, the ad is not displayed immediately. Instead, the user's current churn risk score is calculated based on the data collected in Step 1.

[0022] A churn risk score can consist of multiple sub-scores, for example: 1. Operational obstruction risk score: Calculated based on data such as the number of consecutive operation failures, the duration of task incompleteness, and repeated clicks; 2. Operational fatigue risk score: calculated based on data such as dwell time, inactivity time, and increased click interval; 3. Ad Aversion Risk Score: Calculated based on data such as recent ad closing, ad exits, and a decrease in the rate of complete ad viewing; 4. New User Vulnerability Risk Score: Calculated based on whether the user is a new user and whether the initial positive feedback path has been completed; 5. Risk score for interruption of critical tasks: Calculated based on whether the user is in a state such as before payment, before submission, first game, incomplete tool operation, or at a critical point in content reading; 6. Activity and Stability Deduction Points: Calculated based on data such as users' continuous completion of tasks, active clicking on ad reward entry points, and stable operation.

[0023] The final risk score is calculated using rule-based weighting, threshold judgment, statistical models, or machine learning models. The final risk score can be represented as a numerical value from 0 to 100, or as a discrete level such as low, medium, or high.

[0024] This embodiment uses a rule-weighted approach to calculate a churn risk score ranging from 0 to 100. The calculation formula is as follows: Churn Risk Score = Operational Obstacle Risk Score × 25% + Operational Fatigue Risk Score × 15% + Ad Aversion Risk Score × 25% + New User Vulnerability Risk Score × 15% + Critical Task Interruption Risk Score × 20% - Active and Stable Deduction Score × 10%; Among the risk scores for operational obstruction, 0 to 2 consecutive failed operations score 20 points, 3 to 4 consecutive failed operations score 60 points, and 5 or more consecutive failed operations score 80 points. In the risk score for operational fatigue, 0 to 60 seconds of remaining on the same page without effective action scores 0 points, 60 to 120 seconds scores 60 points, and more than 120 seconds scores 80 points. In the risk score for ad aversion, 0 points are awarded for watching an ad completely, 40 points for closing or skipping an ad, and 40 points are awarded for repeatedly closing ads within a short period, exiting the application within 10 seconds of an ad being displayed, or within 30 minutes. Exiting after seeing more than two ads will result in 80 points; in the new user vulnerability risk score, new users or users who have not yet formed a stable usage path will receive 70 points, while old users with stable recent retention will receive 20 points; in the critical task interruption risk score, users will receive 70 points when they are at critical process nodes such as before payment, before submission, the first game, or when the task is not completed, while non-critical process nodes will receive 20 points; in the activity and stability deduction score, users will have 10 to 30 points deducted if they have active behaviors such as logging in, interacting, watching, or receiving rewards in the past 3 days, and no deduction will be made if there is no active behavior. According to the above rules, the final churn risk score R is represented as: R = w1F1 + w2F2 + w3F3 + w4F4 + w5F5 - w6A, where F1 is the risk score for operational obstruction, F2 is the risk score for operational fatigue, F3 is the risk score for ad aversion, F4 is the risk score for new user vulnerability, F5 is the risk score for critical task interruption, A is the active and stable deduction score, and w1 to w6 are the corresponding weights. Each weight can be preset by the backend or adjusted based on the application's historical ad completion rate, post-ad exit rate, task completion rate, and user retention data. In the example, R less than 40 is considered low risk, R greater than or equal to 40 and less than 70 is considered medium risk, and R greater than or equal to 70 is considered high risk; when R is greater than or equal to 85, or when the user is at a critical task node and simultaneously exhibits post-ad exit behavior, it can be considered extremely high risk. Each sub-score is normalized to a score range of 0 to 100, and the final score is limited to a score range of 0 to 100.

[0025] Step 3: Divide users into different statuses based on their risk scores. For example: 1. Low-churn-risk users: Users whose current behavior is stable, have a high acceptance of ads, and are unlikely to churn after being triggered by ads; 2. Users at risk of churn: Users may experience slight obstacles or hesitation, and need to be guided through advertising benefits related to the current task; 3. Users at high risk of churn: Users may be stuck, fatigued, have been continuously closing ads, or are preparing to exit. Forcibly interrupting the ad at this time may lead to churn. 4. Users at extremely high churn risk: Users have shown a clear tendency to quit. Ad triggering should be paused and priority should be given to ensuring that users complete their current operations.

[0026] Risk stratification thresholds can be configured in the backend or dynamically adjusted based on historical performance.

[0027] Step 4: Match ad triggering strategies based on user risk status Choose an ad triggering strategy based on the user's risk status.

[0028] For users with low churn risk, ads can be displayed normally, for example: 1. Display recommended advertisements; 2. Display ads that double rewards; 3. Display banner ads; 4. Display regular interstitial ads; 5. Display advertisements for claiming benefits.

[0029] For users at risk of churn, prioritize displaying help or benefit-related ads that are relevant to the current task, such as: 1. Display "Watch an ad to get a prompt" when the user's task is blocked; 2. Display "Watch ads to reduce waiting time" while the user is waiting; 3. Display "Watch ads to obtain temporary benefits" when user resources are insufficient; 4. Display "Watch an ad to get a chance to retry" after a user's operation fails; 5. Display "Watch an ad to receive extra rewards" when the user is close to completing the task.

[0030] For users at high risk of churn, the following strategies can be implemented: 1. Delayed ad display; 2. Reduce the frequency of ad displays; 3. Do not display interrupting interstitial ads; 4. Change forced ads to weak entry point ads; 5. Provide non-advertising assistance first, and then reassess whether to display ads after the user's condition is restored.

[0031] For users at extremely high risk of churn, ad display can be paused, and their current tasks can be prioritized, for example: 1. Pause this ad triggering; 2. Provide users with one free notification or free benefit; 3. Lower the threshold for the current task; 4. Guide the user to complete the current operation; 5. Wait until the user's condition stabilizes before re-entering the ad trigger judgment process.

[0032] Step 5: Implement ad triggering, delaying, downgrading, or pausing controls. Execute specific control actions based on the matching results from step 4.

[0033] When the strategy is display advertising, the advertising SDK or application advertising service is invoked to display the corresponding advertisement. Ad types can include rewarded ads, interstitial ads, banner ads, content recommendation ads, reward unlock ads, task reward ads, etc.

[0034] When the strategy is to delay the display, record the reason for the delay and the delay time, and re-evaluate it after the user completes the current key operation, the risk score decreases, or the new triggering conditions are met.

[0035] When the strategy is to downgrade the display, the strong interruption ad will be downgraded to a weak entry ad, a corner entry ad, a task entry ad, or an ad that is only displayed after the user actively clicks on it.

[0036] When the strategy is to pause the display, cancel the current ad trigger and record the reason for the pause. For users who pause ads multiple times consecutively, the priority of subsequent ad triggers can be reduced to avoid interference caused by frequent triggering of judgments.

[0037] Step 6: After the ad is triggered, record the ad results and subsequent user behavior, including but not limited to: whether the ad was successfully loaded; whether the user clicked on the ad entry; whether the user watched the ad in its entirety; whether the user actively closed the ad; whether the user claimed the ad reward; whether the user continued to use the application after the ad; whether the user completed the current task after the ad; whether the user exited the application after the ad; the duration of the user's stay after the ad; and whether the user returned within the preset time.

[0038] The above results are written into the ad trigger log and user status record. Each ad trigger can form a traceable ad event record, recording the risk score before the ad trigger, the trigger strategy, the ad type, the trigger result, and the post-ad behavior.

[0039] Step 7: Update the subsequent strategy based on the results of Step 6. If a user exits the application within 10 seconds of an ad display in 3 out of the last 5 ad displays, or exits after 2 ad displays within 30 minutes, the user is deemed to have frequent exit behavior, and their ad aversion risk score is increased to 80. Subsequently, the user's ad trigger frequency is reduced over the next 24 hours. For example, the interstitial ad trigger interval is adjusted from once every 10 minutes to once every 30 minutes, or the interrupting interstitial ads are changed to incentivized ads, weak entry ads, or delayed ads. If the user continues to use the application and completes the current task after 3 consecutive ad triggers, the ad aversion risk score is gradually reduced, and some ad triggering strategies are restored.

[0040] If users at a certain risk level continue to use and complete tasks after an ad is triggered, it indicates that the ad strategy for that level is relatively effective, and the trigger weight of that strategy can be maintained or appropriately increased.

[0041] If users at a certain risk level frequently exit after an ad is triggered, it indicates that the ad triggering is too aggressive. In such cases, the risk threshold for that level can be increased, the ad frequency reduced, the ad timing delayed, or the ad type adjusted.

[0042] If new users have a high exit rate after being shown ads before completing their first positive feedback, the new user ad protection period can be extended. A high exit rate means that the exit rate of new users after being shown ads before completing their first positive feedback reaches or exceeds 40%, or that a single new user exits the application within 10 seconds of being shown ads in 2 out of their last 3 ad triggers.

[0043] If a user actively clicks on a certain type of rewarded ad and watches it in its entirety, the risk weight of that type of ad disturbing the user can be reduced.

[0044] Through the aforementioned closed loop, the ad triggering strategy can be continuously adjusted for different user states.

[0045] See Figure 2 A dynamic adjustment system for in-app advertising triggers based on user churn risk scoring, the system comprising: The user behavior collection module is used to collect user behavior data, advertising history data, and application context data when a user enters the application and performs an operation. The churn risk scoring module is used to calculate the user's current churn risk score based on collected user behavior data, advertising history data, and application context data when the application is about to trigger an ad. The user risk stratification module is used to classify users into four risk levels: low risk, medium risk, high risk, and extremely high risk, based on the churn risk score. The advertising strategy matching module is used to execute advertising triggering strategies for different risk levels. The advertising triggering strategies are as follows: Display monetization ads to low-risk users normally and record ad trigger results; Show helpful or benefit-based ads related to the current task to users at medium risk, and record the ad trigger results; Delay ad display or reduce ad frequency for high-risk users, and record the reasons for delay or downgrade; For users with extremely high risk, suspend ad display and prioritize protecting their current operations, and record the reason for suspension and the protection strategy; The post-advertisement behavior feedback module is used to collect user behavior data after an advertisement is triggered, and to determine whether the user continues to use the application or completes the current task. The strategy update module is used to adjust the weight of the corresponding advertising strategy and update the risk score parameters and advertising triggering strategy based on whether the user continues to use the service or the result of completing the task. The evaluation module is used to re-execute the evaluation process of the above modules before the next ad trigger.

[0046] In one embodiment of the present invention, the user churn risk score is obtained by weighted summation of operation obstruction risk score, operation fatigue risk score, advertising aversion risk score, new user vulnerability risk score, critical task interruption risk score, and activity stability deduction score, and the calculation formula is as follows: Churn Risk Score = Operational Obstacle Risk Score × 25% + Operational Fatigue Risk Score × 15% + Ad Aversion Risk Score × 25% + New User Vulnerability Risk Score × 15% + Critical Task Interruption Risk Score × 20% - Active and Stable Deduction Score × 10%; Among the risk scores for operational obstruction, 0 to 2 consecutive failed operations score 20 points, 3 to 4 consecutive failed operations score 60 points, and 5 or more consecutive failed operations score 80 points. In the risk score for operational fatigue, 0 to 60 seconds of remaining on the same page without effective action scores 0 points, 60 to 120 seconds scores 60 points, and more than 120 seconds scores 80 points. In the risk score for ad aversion, 0 points are awarded for watching an ad completely, 40 points for closing or skipping an ad, and 40 points are awarded for repeatedly closing ads within a short period, exiting the application within 10 seconds of an ad being displayed, or within 30 minutes. Exiting after seeing more than two ads will result in 80 points; in the new user vulnerability risk score, new users or users who have not yet formed a stable usage path will receive 70 points, while old users with stable recent retention will receive 20 points; in the critical task interruption risk score, users will receive 70 points when they are at critical process nodes such as before payment, before submission, the first game, or when the task is not completed, while non-critical process nodes will receive 20 points; in the activity and stability deduction score, users will have 10 to 30 points deducted if they have active behaviors such as logging in, interacting, watching, or receiving rewards in the past 3 days, and no deduction will be made if there is no active behavior. According to the above rules, the final churn risk score R is represented as: R = w1F1 + w2F2 + w3F3 + w4F4 + w5F5 - w6A, where F1 is the risk score for operational obstruction, F2 is the risk score for operational fatigue, F3 is the risk score for ad aversion, F4 is the risk score for new user vulnerability, F5 is the risk score for critical task interruption, A is the active and stable deduction score, and w1 to w6 are the corresponding weights. Each weight can be preset by the backend or adjusted based on the application's historical ad completion rate, post-ad exit rate, task completion rate, and user retention data. In the example, R less than 40 is considered low risk, R greater than or equal to 40 and less than 70 is considered medium risk, and R greater than or equal to 70 is considered high risk; when R is greater than or equal to 85, or when the user is at a critical task node and simultaneously exhibits post-ad exit behavior, it can be considered extremely high risk. Each sub-score is normalized to a score range of 0 to 100, and the final score is limited to a score range of 0 to 100.

[0047] During system implementation, the churn risk scoring module also stores segmented threshold tables and weight configuration tables for each sub-score. The segmented threshold table includes at least the ranges for consecutive operation failures, page dwell time, ad closing or exit behavior, new user status, key process status, and active behavior deduction ranges. The weight configuration table includes at least the weights for operation obstruction risk scores, operation fatigue risk scores, ad aversion risk scores, new user vulnerability risk scores, key task interruption risk scores, and active stability deduction scores. The churn risk scoring module reads the above configurations before each ad trigger and combines them with the user's current session data to generate a churn risk score and risk level.

[0048] Each sub-score is normalized to a range of 0 to 100 points, with a maximum deduction of 10 points for active and stable performance, and the final score is limited to a range of 0 to 100 points.

[0049] Based on the previous embodiment, the threshold for classifying the risk levels is as follows: 0 to 30 points is low churn risk; 31 to 60 points is medium churn risk; 61 to 80 points is high churn risk; and 81 to 100 points is extremely high churn risk.

[0050] In one embodiment of the present invention, the collection of user behavior data after the advertisement is triggered includes: whether the advertisement is successfully loaded, whether the user clicks on the advertisement entry, whether the advertisement is watched in its entirety, whether the advertisement is actively closed, whether the advertisement reward is claimed, whether the user continues to use the application after the advertisement, whether the current task is completed, whether the application is exited, the duration of the user stays after the advertisement, and whether the user revisits within a preset time. The above results are written into the ad trigger log and user status record. Each ad trigger can form a traceable ad event record. The ad event record is used to record the risk score before ad triggering, triggering strategy, ad type, triggering result and post-ad behavior.

[0051] In one embodiment of the present invention, the updating of risk scoring parameters and advertising triggering strategies includes: If the user continues to use the application and completes the task after the ad is triggered, the trigger weight of the ad strategy for the corresponding risk level will be maintained or increased. If a user exits the application within 10 seconds of the ad being displayed 3 out of the last 5 ads, or exits the application after 2 ads appearing within 30 minutes, the risk threshold for the corresponding risk level will be increased, the ad frequency will be reduced, or the ad type will be adjusted. If the post-ad exit rate of new users before their first positive feedback reaches a preset threshold, the new user ad protection period will be extended.

[0052] The solution can be adapted to various in-app advertising scenarios: 1. In mini-games, user behavior data may include the number of times a level is failed, the level completion status, the need for revival, the need for hints, and the duration of time spent in a level; advertising strategies may include rewarded videos, revival ads, hint ads, and reward doubling ads.

[0053] 2. In short video applications, user behavior data may include video viewing time, swiping frequency, dwell time, and following or liking behavior; advertising strategies may include in-feed ads, interactive ads, and benefit ads.

[0054] 3. In tool applications, user behavior data may include feature usage progress, waiting time, number of exports, and number of failed retries; advertising strategies may include benefit unlocking ads, waiting time reduction ads, and export acceleration ads.

[0055] 4. In news or content applications, user behavior data may include the number of articles read, reading time, return behavior, and collection / comment behavior; advertising strategies may include in-read ads, recommended ads, and rewarded reading ads.

[0056] 5. In e-commerce or live streaming applications, user behavior data may include browsing path, dwell time, add-to-cart behavior, and exit tendency; advertising strategies may include coupon ads, event ads, and content recommendation ads.

[0057] The above description is only a preferred embodiment of the present invention. All equivalent changes and modifications made within the scope of the claims of the present invention should be included in the scope of the present invention.

Claims

1. A method for dynamically adjusting in-app ad triggering based on user churn risk scoring, characterized in that, The method steps are as follows: Step 1: Users enter the application and perform actions; user behavior data, advertising history data, and application context data are collected. Step 2: When the application is ready to trigger an ad, calculate the user's current churn risk score based on the collected user behavior data, ad history data, and application context data; Step 3: Based on the churn risk score, users are divided into four risk levels: low risk, medium risk, high risk, and very high risk. Step 4: Implement ad triggering strategies for different risk levels. The ad triggering strategies are as follows: Display monetization ads to low-risk users normally and record ad trigger results; Show helpful or benefit-based ads related to the current task to users at medium risk, and record the ad trigger results; Delay ad display or reduce ad frequency for high-risk users, and record the reasons for delay or downgrade; For users with extremely high risk, suspend ad display and prioritize protecting their current operations, and record the reason for suspension and the protection strategy; Step 5: Collect user behavior data after the ad is triggered to determine whether the user continues to use the application or completes the current task; Step 6: Adjust the weight of the corresponding advertising strategy based on whether the user continues to use the service or completes the task, and update the risk scoring parameters and advertising triggering strategy; Step 7: Before the next ad trigger, repeat the evaluation process from Step 2 to Step 6.

2. The method for dynamically adjusting in-app ad triggering based on user churn risk scoring according to claim 1, characterized in that, The user churn risk score is obtained by weighted summation of the following scores: operational obstruction risk score, operational fatigue risk score, advertising aversion risk score, new user vulnerability risk score, critical task interruption risk score, and activity stability deduction score. The calculation formula is as follows: Churn Risk Score = Operational Obstacle Risk Score × 25% + Operational Fatigue Risk Score × 15% + Ad Aversion Risk Score × 25% + New User Vulnerability Risk Score × 15% + Critical Task Interruption Risk Score × 20% - Active and Stable Deduction Score × 10%; Among the risk scores for operational obstruction, 0 to 2 consecutive failed operations score 20 points, 3 to 4 consecutive failed operations score 60 points, and 5 or more consecutive failed operations score 80 points. In the risk score for operational fatigue, 0 to 60 seconds of remaining on the same page without effective action scores 0 points, 60 to 120 seconds scores 60 points, and more than 120 seconds scores 80 points. In the risk score for ad aversion, 0 points are awarded for watching an ad completely, 40 points for closing or skipping an ad, and 40 points are awarded for repeatedly closing ads within a short period, exiting the application within 10 seconds of an ad being displayed, or within 30 minutes. Exiting after seeing more than two ads will result in 80 points; in the new user vulnerability risk score, new users or users who have not yet formed a stable usage path will receive 70 points, while old users with stable recent retention will receive 20 points; in the critical task interruption risk score, users will receive 70 points when they are at critical process nodes such as before payment, before submission, the first game, or when the task is not completed, while non-critical process nodes will receive 20 points; in the activity and stability deduction score, users will have 10 to 30 points deducted if they have active behaviors such as logging in, interacting, watching, or receiving rewards in the past 3 days, and no deduction will be made if there is no active behavior. According to the above rules, the final churn risk score R is represented as: R = w1F1 + w2F2 + w3F3 + w4F4 + w5F5 - w6A, where F1 is the risk score for operational obstruction, F2 is the risk score for operational fatigue, F3 is the risk score for ad aversion, F4 is the risk score for new user vulnerability, F5 is the risk score for critical task interruption, A is the active and stable deduction score, and w1 to w6 are the corresponding weights. Each weight can be preset by the backend or adjusted based on the application's historical ad completion rate, post-ad exit rate, task completion rate, and user retention data. In the example, R less than 40 is considered low risk, R greater than or equal to 40 and less than 70 is considered medium risk, and R greater than or equal to 70 is considered high risk; when R is greater than or equal to 85, or when the user is at a critical task node and simultaneously exhibits post-ad exit behavior, it can be considered extremely high risk. Each sub-score is normalized to a score range of 0 to 100, and the final score is limited to a score range of 0 to 100.

3. The method for dynamically adjusting in-app ad triggering based on user churn risk scoring according to claim 1, characterized in that, The risk level classification thresholds are as follows: 0 to 30 points indicate low churn risk; 31 to 60 points indicate medium churn risk; 61 to 80 points indicate high churn risk; and 81 to 100 points indicate extremely high churn risk.

4. The method for dynamically adjusting in-app ad triggering based on user churn risk scoring according to claim 1, characterized in that, The collected user behavior data after the advertisement is triggered includes: whether the advertisement is successfully loaded, whether the user clicks on the advertisement entry, whether the advertisement is watched in its entirety, whether the advertisement is actively closed, whether the advertisement reward is claimed, whether the user continues to use the application after the advertisement, whether the current task is completed, whether the application is exited, the duration of the user stays after the advertisement, and whether the user revisits within the preset time. The above results are written into the ad trigger log and user status record. Each ad trigger can form a traceable ad event record. The ad event record is used to record the risk score before ad triggering, triggering strategy, ad type, triggering result and post-ad behavior.

5. The method for dynamically adjusting in-app ad triggering based on user churn risk scoring according to claim 1, characterized in that, The updated risk scoring parameters and ad triggering strategies include: If the user continues to use the application and completes the task after the ad is triggered, the trigger weight of the ad strategy for the corresponding risk level will be maintained or increased. If a user exits the application within 10 seconds of the ad being displayed 3 out of the last 5 ads, or exits the application after 2 ads appearing within 30 minutes, the risk threshold for the corresponding risk level will be increased, the ad frequency will be reduced, or the ad type will be adjusted. If the post-ad exit rate of new users before their first positive feedback reaches a preset threshold, the new user ad protection period will be extended.

6. A dynamic adjustment system for in-app advertising triggers based on user churn risk scoring, characterized in that, The system includes: The user behavior collection module is used to collect user behavior data, advertising history data, and application context data when a user enters the application and performs an operation. The churn risk scoring module is used to calculate the user's current churn risk score based on collected user behavior data, advertising history data, and application context data when the application is about to trigger an ad. The user risk stratification module is used to classify users into four risk levels: low risk, medium risk, high risk, and extremely high risk, based on the churn risk score. The advertising strategy matching module is used to execute advertising triggering strategies for different risk levels. The advertising triggering strategies are as follows: Display monetization ads to low-risk users normally and record ad trigger results; Show helpful or benefit-based ads related to the current task to users at medium risk, and record the ad trigger results; Delay ad display or reduce ad frequency for high-risk users, and record the reasons for delay or downgrade; For users with extremely high risk, suspend ad display and prioritize protecting their current operations, and record the reason for suspension and the protection strategy; The post-advertisement behavior feedback module is used to collect user behavior data after an advertisement is triggered, and to determine whether the user continues to use the application or completes the current task. The strategy update module is used to adjust the weight of the corresponding advertising strategy and update the risk score parameters and advertising triggering strategy based on whether the user continues to use the service or the result of completing the task. The evaluation module is used to re-execute the evaluation process of the above modules before the next ad trigger.

7. The in-app advertising triggering dynamic adjustment system based on user churn risk scoring according to claim 6, characterized in that, The user churn risk score is obtained by weighted summation of the following scores: operational obstruction risk score, operational fatigue risk score, advertising aversion risk score, new user vulnerability risk score, critical task interruption risk score, and activity stability deduction score. The calculation formula is as follows: Churn Risk Score = Operational Obstacle Risk Score × 25% + Operational Fatigue Risk Score × 15% + Ad Aversion Risk Score × 25% + New User Vulnerability Risk Score × 15% + Critical Task Interruption Risk Score × 20% - Active and Stable Deduction Score × 10%; Among the risk scores for operational obstruction, 0 to 2 consecutive failed operations score 20 points, 3 to 4 consecutive failed operations score 60 points, and 5 or more consecutive failed operations score 80 points. In the risk score for operational fatigue, 0 to 60 seconds of remaining on the same page without effective action scores 0 points, 60 to 120 seconds scores 60 points, and more than 120 seconds scores 80 points. In the risk score for ad aversion, 0 points are awarded for watching an ad completely, 40 points for closing or skipping an ad, and 40 points are awarded for repeatedly closing ads within a short period, exiting the application within 10 seconds of an ad being displayed, or within 30 minutes. Exiting after seeing more than two ads will result in 80 points; in the new user vulnerability risk score, new users or users who have not yet formed a stable usage path will receive 70 points, while old users with stable recent retention will receive 20 points; in the critical task interruption risk score, users will receive 70 points when they are at critical process nodes such as before payment, before submission, the first game, or when the task is not completed, while non-critical process nodes will receive 20 points; in the activity and stability deduction score, users will have 10 to 30 points deducted if they have active behaviors such as logging in, interacting, watching, or receiving rewards in the past 3 days, and no deduction will be made if there is no active behavior. According to the above rules, the final churn risk score R is represented as: R = w1F1 + w2F2 + w3F3 + w4F4 + w5F5 - w6A, where F1 is the risk score for operational obstruction, F2 is the risk score for operational fatigue, F3 is the risk score for ad aversion, F4 is the risk score for new user vulnerability, F5 is the risk score for critical task interruption, A is the active and stable deduction score, and w1 to w6 are the corresponding weights. Each weight can be preset by the backend or adjusted based on the application's historical ad completion rate, post-ad exit rate, task completion rate, and user retention data. In the example, R less than 40 is considered low risk, R greater than or equal to 40 and less than 70 is considered medium risk, and R greater than or equal to 70 is considered high risk; when R is greater than or equal to 85, or when the user is at a critical task node and simultaneously exhibits post-ad exit behavior, it can be considered extremely high risk. Each sub-score is normalized to a score range of 0 to 100, and the final score is limited to a score range of 0 to 100.

8. The in-app advertising triggering dynamic adjustment system based on user churn risk scoring according to claim 6, characterized in that, The risk level classification thresholds are as follows: 0 to 30 points indicate low churn risk; 31 to 60 points indicate medium churn risk; 61 to 80 points indicate high churn risk; and 81 to 100 points indicate extremely high churn risk.

9. The in-app advertising triggering dynamic adjustment system based on user churn risk scoring according to claim 6, characterized in that, The collected user behavior data after the advertisement is triggered includes: whether the advertisement is successfully loaded, whether the user clicks on the advertisement entry, whether the advertisement is watched in its entirety, whether the advertisement is actively closed, whether the advertisement reward is claimed, whether the user continues to use the application after the advertisement, whether the current task is completed, whether the application is exited, the duration of the user stays after the advertisement, and whether the user revisits within the preset time. The above results are written into the ad trigger log and user status record. Each ad trigger can form a traceable ad event record. The ad event record is used to record the risk score before ad triggering, triggering strategy, ad type, triggering result and post-ad behavior.

10. The in-app advertising triggering dynamic adjustment system based on user churn risk scoring according to claim 6, characterized in that, The updated risk scoring parameters and ad triggering strategies include: If the user continues to use the application and completes the task after the ad is triggered, the trigger weight of the ad strategy for the corresponding risk level will be maintained or increased. If a user exits the application within 10 seconds of the ad being displayed 3 out of the last 5 ads, or exits the application after 2 ads appearing within 30 minutes, the risk threshold for the corresponding risk level will be increased, the ad frequency will be reduced, or the ad type will be adjusted. If the post-ad exit rate of new users before their first positive feedback reaches a preset threshold, the new user ad protection period will be extended.

Images