Marketing promotion method based on member churn prediction

Abstract

This invention relates to the field of digital marketing and discloses a marketing promotion method based on member churn prediction, comprising: Step 1, collecting member behavior data, transaction data, social interaction data, and advertising data; Step 2, preprocessing the data, including data cleaning, feature selection, and data standardization; Step 3, establishing a member churn prediction model to predict whether members will churn; Step 4, establishing a promotion channel effectiveness prediction model to evaluate the effectiveness of each promotion channel; Step 5, calculating the weight value of each promotion channel, the weight value reflecting the cost-effectiveness of the promotion channel; Step 6, ranking the promotion channels based on the weight values. Through accurate analysis based on member churn prediction, this invention can allocate corresponding advertising budgets to each promotion channel, ensuring that resources of different channels are rationally allocated, thereby enabling promotional activities to more accurately match the target audience and avoiding resource waste.

Description

Technical Field

[0001] This invention relates to the field of digital marketing technology, specifically a marketing promotion method based on member churn prediction. Background Technology

[0002] With the booming development of e-commerce and the internet industry, membership marketing has become an important means for businesses to improve customer loyalty, increase sales, and expand market share. To improve marketing effectiveness, businesses need to attract more potential customers and retain existing customers through precise marketing and promotion methods. However, effectively identifying potential churned members, optimizing promotion channels, and rationally allocating marketing resources remain major challenges in the current marketing field. Existing traditional marketing methods rely heavily on manual design and experience-based judgment, failing to fully utilize big data and artificial intelligence technologies, resulting in inefficiency and resource waste in many marketing activities.

[0003] Currently, there are several technical solutions on the market for churn prediction and marketing strategy optimization based on member behavior data. Traditional churn prediction methods mostly rely on historical data analysis, user tagging, and some simple statistical models, such as logistic regression and decision trees. These methods predict whether members are likely to churn by analyzing data such as purchase frequency, activity level, and browsing behavior, and adjust marketing strategies accordingly. Meanwhile, some technical solutions are also beginning to combine multi-source data (such as social media and user interactions) to optimize the selection of promotional channels and budget allocation. However, these existing technologies still have certain limitations. They cannot accurately predict the impact of different promotional channels on member churn, and they are difficult to automatically adjust strategies based on real-time data, failing to meet the rapidly changing market demands.

[0004] Existing technical solutions mainly rely on static data analysis and manual intervention to predict member churn and adjust promotional strategies, but these methods have the following problems:

[0005] Insufficient evaluation of promotion channel effectiveness: Most existing solutions only focus on predicting member churn and neglect the comprehensive evaluation of the effectiveness of different promotion channels, resulting in a lack of precision in the allocation of promotion resources and an inability to guarantee the optimal effect of marketing activities.

[0006] Lack of real-time dynamic adjustment mechanism: Existing methods usually make decisions based on historical data and fail to adjust promotion strategies in real time according to member behavior and market changes, resulting in marketing activities being slow to respond to the rapidly changing market environment. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a marketing promotion method based on member churn prediction. This method solves the problem that existing methods typically rely on historical data for decision-making and fail to adjust promotion strategies in real time based on member behavior and market changes, resulting in slow response of marketing activities in the face of a rapidly changing market environment.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a marketing promotion method based on member churn prediction, comprising:

[0009] Step 1: Collect member behavior data, transaction data, social interaction data, and advertising data;

[0010] Step 2: Preprocess the data, including data cleaning, feature selection, and data standardization;

[0011] Step 3: Establish a member churn prediction model to predict whether members will churn;

[0012] Step 4: Establish a promotion channel performance prediction model and evaluate the performance of each promotion channel;

[0013] Step 5: Calculate the weight value of each promotion channel, which reflects the cost-effectiveness of the promotion channel;

[0014] Step 6: Sort the promotion channels based on the weight values;

[0015] Step 7: Adjust the promotion strategy based on the ranking results, and optimize the promotion campaign by adjusting the budget for different channels;

[0016] Step 8: Monitor the implementation process of the promotion activities and adjust strategies based on actual feedback results.

[0017] Preferably, the member churn prediction model is established using a deep neural network or an XGBoost model, and the churn prediction formula is:

[0018]

[0019] Among them, P 流失 (i) represents the churn probability of the i-th member, σ is the Sigmoid function, and w j Let X be the weight of the j-th feature. ij Let be the value of the i-th member on the j-th feature, and b be the bias term.

[0020] Preferably, the promotion channel effectiveness prediction model is established using a regression model, and the regression formula is:

[0021] y = β0 + β1X1 + β2X2 + ... + β n X n

[0022] Where y is the performance indicator of the promotion channel, X1, X2, ..., X n For the characteristics of this promotion channel, β0, β1, ..., β n These are the coefficients of the regression model.

[0023] Preferably, the promotion budget allocation formula is:

[0024] Budget j =W j Total Budget

[0025] Among them, the budget j Let $\frac{j}{j}$ be the promotion budget for the $j$-th channel, and $\frac{j}{j}$ be the total budget for the entire marketing campaign.

[0026] Preferably, the optimized promotion strategy is based on A / B testing or a multi-armed slot machine algorithm. The A / B test selects the optimal strategy by comparing the effects of different promotion strategies.

[0027] Preferably, the multi-armed slot machine algorithm is used to dynamically adjust the distribution ratio of each promotion channel to balance exploration and utilization, and to ensure that the distribution strategy is adjusted based on historical data and real-time feedback.

[0028] Preferably, the monitoring steps include real-time monitoring of the promotion effects of each channel and adjusting the delivery strategy based on indicators such as churn rate and conversion rate.

[0029] Preferably, the weight value W of the promotion channel j The calculation formula is:

[0030]

[0031] Among them, W j Let P be the weight value of the j-th promotion channel. 流失 Let C be the predicted probability of member churn for the j-th channel. 转化率 For the conversion rate and cost of this channel j This refers to the promotion costs for this channel.

[0032] Preferably, the method further includes regularly updating the member churn prediction model and the promotion channel effectiveness prediction model to cope with changes in the market and user behavior.

[0033] Preferably, the method adjusts the promotion strategy based on user behavior data and market feedback.

[0034] This invention provides a marketing and promotion method based on member churn prediction. It has the following beneficial effects:

[0035] 1. Through precise analysis based on member churn prediction, this invention can allocate corresponding budgets to various promotional channels, ensuring that resources of different channels are rationally allocated, thereby enabling promotional activities to more accurately match the target audience and avoid resource waste.

[0036] 2. This invention automatically predicts member churn and promotional channel effectiveness by utilizing technologies such as deep learning and regression analysis, and adjusts promotional budgets and strategies based on the results, reducing manual intervention and improving the efficiency and response speed of promotional activities.

[0037] 3. By predicting and accurately matching the effects of various promotion channels, this invention can effectively avoid the waste of resources in inefficient channels, enabling marketing expenses to be used more accurately and effectively, thereby reducing marketing costs. The monitoring and feedback mechanism continuously adjusts and optimizes promotion strategies to cope with market changes and changes in user behavior, thus possessing strong adaptability and flexibility, and can be adjusted in a timely manner to meet different marketing needs. Attached Figure Description

[0038] Figure 1 Flow chart of the method of the present invention. Detailed Implementation

[0039] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] Example 1:

[0041] Please see the appendix Figure 1 This invention provides a marketing promotion method based on member churn prediction, including:

[0042] Step 1: Collect member behavior data, transaction data, social interaction data, and advertising data;

[0043] Step 2: Preprocess the data, including data cleaning, feature selection, and data standardization;

[0044] Step 3: Establish a member churn prediction model to predict whether members will churn;

[0045] Step 4: Establish a promotion channel performance prediction model and evaluate the performance of each promotion channel;

[0046] Step 5: Calculate the weight value of each promotion channel. The weight value reflects the cost-effectiveness of the promotion channel.

[0047] Step 6: Sort the promotion channels based on their weight values;

[0048] Step 7: Adjust the promotion strategy based on the ranking results, and optimize the promotion campaign by adjusting the budget for different channels;

[0049] Step 8: Monitor the implementation process of the promotion activities and adjust strategies based on actual feedback results.

[0050] Membership churn prediction models are built using deep neural networks or XGBoost models. The churn prediction formula is as follows:

[0051]

[0052] Among them, P 流失 (i) represents the churn probability of the i-th member, σ is the Sigmoid function, and w j Let X be the weight of the j-th feature. ij Let be the value of the i-th member on the j-th feature, and b be the bias term.

[0053] The promotion channel effectiveness prediction model is established using a regression model, and the regression formula is:

[0054] y = β0 + β1X1 + β2X2 + ... + β n X n

[0055] Where y is the performance indicator of the promotion channel, X1, X2, ..., X n For the characteristics of this promotion channel, β0, β1, ..., β n These are the coefficients of the regression model.

[0056] The formula for allocating the promotion budget is:

[0057] Budget j =W j Total Budget

[0058] Among them, the budget j Let $\frac{j}{j}$ be the promotion budget for the $j$-th channel, and $\frac{j}{j}$ be the total budget for the entire marketing campaign.

[0059] Optimize promotion strategies based on A / B testing or multi-armed slot machine algorithms. A / B testing compares the effects of different promotion strategies to select the optimal strategy.

[0060] The multi-armed slot machine algorithm is used to dynamically adjust the allocation ratio of various promotion channels to balance exploration and utilization, and ensure that the allocation strategy is adjusted based on historical data and real-time feedback.

[0061] The monitoring process includes real-time monitoring of the promotional effectiveness across various channels and adjusting the campaign strategy based on metrics such as churn rate and conversion rate.

[0062] The weight value W of the promotion channel j The calculation formula is:

[0063]

[0064] Among them, W j Let P be the weight value of the j-th promotion channel. 流失 Let C be the predicted probability of member churn for the j-th channel. 转化率 For the conversion rate and cost of this channel j This refers to the promotion costs for this channel.

[0065] The method also includes regularly updating the member churn prediction model and the promotion channel effectiveness prediction model to respond to changes in the market and user behavior.

[0066] The method adjusts promotion strategies based on user behavior data and market feedback.

[0067] In one embodiment, step 1: multi-source data acquisition

[0068] This method first acquires a large amount of multi-source data through a data acquisition module. Data sources include, but are not limited to:

[0069] Member basic information: member registration information, account level, historical purchase behavior, member points, etc.

[0070] Member behavior data: Members' browsing, clicking, and search history on the platform, product click-through rate, page dwell time, etc.

[0071] Transaction data: Member's historical purchase records, order frequency, spending amount, types of goods purchased, average transaction amount, etc.

[0072] Social interaction data: Members' social interaction behaviors, such as commenting, liking, sharing, joining groups, etc.

[0073] Advertising data: Advertising data for various promotional channels, including ad impressions, clicks, conversion rates, and costs.

[0074] User feedback data: User feedback data collected through surveys, customer service records, and other means helps to better understand members' needs and satisfaction.

[0075] All data should be collected while ensuring privacy and security, and updated in real time.

[0076] Step 2: Data Preprocessing

[0077] The collected raw data typically contains missing values, outliers, and duplicate data, requiring cleaning and preprocessing before subsequent use. Preprocessing includes the following steps:

[0078] Missing value handling: Missing data records are handled by interpolation (such as mean filling or median filling) or by deleting missing data records.

[0079] Outlier detection: Statistical methods (such as Z-Score, IQR) are used to detect and correct outlier data to ensure data quality.

[0080] Data standardization: Standardizing numerical data to give data of different units the same scale. A commonly used method is Z-score standardization.

[0081]

[0082] Where X represents the original data, μ represents the mean, and σ represents the standard deviation.

[0083] Feature engineering involves extracting meaningful features from raw data. For example, calculating purchase frequency, spending amount, and average purchase cycle based on members' transaction data. The L1 regularization (Lasso) algorithm can be used to select the most relevant features.

[0084] Step 3: Building a Member Churn Prediction Model

[0085] This invention employs machine learning techniques to construct a member churn prediction model. Based on the data preprocessing results in step 2, appropriate features are selected, and algorithms such as Deep Neural Networks (DNN) or Random Forest are used for churn prediction modeling. The construction steps of the churn prediction model are as follows:

[0086] Dataset partitioning: The dataset is divided into training and testing sets. Typically, 70% of the data is used for training and 30% for validation.

[0087] Model selection: You can choose machine learning algorithms such as deep neural networks, decision trees, and support vector machines (SVM). Deep neural networks typically include an input layer, multiple hidden layers, and an output layer.

[0088] Model training: Training is performed using the backpropagation algorithm. The loss function typically used is the cross-entropy loss function.

[0089]

[0090] Among them, y i p is the true label of the i-th sample. iThis represents the predicted churn probability for this sample.

[0091] Model evaluation: The predictive performance of the model is evaluated using metrics such as AUC, precision, and recall.

[0092] Step 4: Building a Prediction Model for Promotion Channel Effectiveness

[0093] The effectiveness of promotional channels needs to be estimated using predictive models. Specifically, this involves building a predictive model based on historical advertising data and the channel's fundamental characteristics (such as click-through rate, conversion rate, and cost of goods sold). We use a linear regression model, and the prediction formula is as follows:

[0094] y j =β0+β1x1+β2x2+…+β n x n

[0095] Among them, y j Let x1, x2, ..., xj be the performance score of the j-th promotion channel. n Let β0, β1, ..., β be the characteristics of this channel. n These are the regression coefficients. The least squares method is used to estimate the regression coefficients, thereby predicting the promotion effect.

[0096] Step 5: Calculate the weight of promotion channels

[0097] Based on the member churn prediction results and the predicted performance values ​​of each promotion channel, a weight value for each promotion channel is calculated. The weight calculation formula is as follows:

[0098]

[0099] Among them, W j Let P be the weight value of the j-th promotion channel. 流失 Let C be the predicted probability of member churn for the j-th channel. 转化率 For the conversion rate and cost of this channel j This refers to the promotion costs for this channel.

[0100] The weight value reflects the cost-effectiveness of each channel; a higher weight indicates that a channel is more likely to bring higher conversion rates or lower costs.

[0101] Step 6: Promotion Budget Allocation and Optimization

[0102] Based on the weight values ​​calculated in step 5, allocate corresponding budgets to each promotion channel. The specific budget allocation formula is: Budget j =W j Total Budget

[0103] Adjust the spending amount for each promotional channel based on the channel's budget and weight to ensure maximum promotional effectiveness.

[0104] Furthermore, A / B testing and a multi-armed slot machine algorithm were used to optimize the promotion strategy. In A / B testing, members were randomly grouped and experimented with different combinations of channels to evaluate the effectiveness of different strategies and select the best strategy. In the multi-armed slot machine algorithm, the budgets for each channel were dynamically adjusted to achieve optimal long-term returns.

[0105] Step 7: Real-time monitoring and feedback

[0106] During the marketing campaign, the effectiveness of each promotional channel is monitored in real time, and key metrics such as conversion rate, click-through rate, and ROI are collected promptly. Based on real-time data feedback, the budget and placement strategies for each channel are dynamically adjusted to ensure that the promotional campaign remains highly efficient.

[0107] Step 8: Model Update and Maintenance

[0108] To ensure the adaptability and long-term effectiveness of the models, the churn prediction model and the promotion channel effectiveness prediction model need to be updated regularly. The update process includes retraining with the latest data and adjusting the model hyperparameters.

[0109] Comparative Example 1:

[0110] Improvements and experimental verification of the "promotion channel weight calculation" step;

[0111] Original steps and instructions:

[0112] This invention uses the following formula for calculating the weight of promotion channels:

[0113]

[0114] Among them, W j Let P be the weight value of the j-th promotion channel. 流失 Let C be the predicted probability of member churn for the j-th channel. 转化率 For the conversion rate and cost of this channel j This refers to the promotion costs for this channel.

[0115] This formula comprehensively considers member churn risk, the effectiveness of promotional channels, and investment costs, providing a quantitative basis for budget allocation. The formula will be gradually improved, experimentally verified, and the overall optimality of the proposed solution in terms of performance and practical application will be demonstrated.

[0116] Improvement Solution 1: Introduce the Click-Through Rate (CTR) feature.

[0117] Improvements:

[0118] Try introducing click-through rate (CTR) to make the weight calculation formula more comprehensively reflect the channel's performance:

[0119]

[0120] Experimental steps:

[0121] Data preparation: Collect the following data for each promotion channel:

[0122] P 流失 Calculated by the churn prediction model.

[0123] C 转化率 Actual conversion rate of promotion channels.

[0124] CTR j Click-through rate of the advertisement.

[0125] cost j The cost of advertising through this channel.

[0126] Weight Calculation: The weight value for each channel is calculated using the improved formula.

[0127] Model evaluation: The performance of the improvement plan is verified by comprehensively evaluating the click-through rate, conversion rate and ROI (Return on Investment) of the promotional campaign.

[0128]

[0129] Table 1

[0130] Analysis and Conclusion:

[0131] Performance degradation: The F1-Score of the model decreased slightly after the introduction of CTR, indicating that CTR is highly correlated with conversion rate and the marginal contribution of the new feature to the model is low.

[0132] Increased resource consumption: CTR calculation increased resource consumption by 15%, and its effect on optimizing promotion strategies was limited.

[0133] Final conclusion: Improved solution 1 performs poorly in terms of resource consumption and performance improvement, and is inferior to the original solution of this invention.

[0134] Comparative Example 2:

[0135] Experimental steps:

[0136] Data preparation and preprocessing:

[0137] Collect at least six months of member behavior data from e-commerce platforms, social media, etc., including purchase records, browsing history, social interactions, etc.

[0138] Use data processing tools such as Python to preprocess the data, fill in missing values, remove duplicate data, and perform standardization.

[0139] Model training and evaluation:

[0140] Train multiple machine learning models (logistic regression, random forest, support vector machine, etc.) using historical data, and evaluate the accuracy of the models through cross-validation.

[0141] AUC (area under the curve) and F1-score were used as evaluation metrics to select the best model for churn prediction.

[0142] Weight calculation and optimization:

[0143] The trained churn prediction model is used to output the churn probability of each member across various promotional channels.

[0144] Calculate the weight value of each channel, sort the channels according to the weight, and allocate the budget accordingly.

[0145] Promotion strategy implementation and effect monitoring:

[0146] Implement optimized promotion strategies, regularly monitor conversion rates across various channels, and adjust budget allocation strategies accordingly.

[0147] By monitoring member behavior data in real time, we can analyze the promotion effect, adjust strategies in a timely manner, and ensure maximum return on investment.

[0148] Experimental results

[0149] Through comparative experiments, the promotion method of Example 1 was compared with the traditional manual design promotion strategy and other optimization schemes. The experimental results are shown in the table below:

[0150]

[0151] Table 2

[0152] Experimental Analysis and Conclusions:

[0153] Accuracy and effectiveness:

[0154] Example 1 performs comparably to other solutions in terms of accuracy, precision, and recall, with its F1-score being slightly higher than other improved solutions, indicating that it achieves a better balance between precision and recall.

[0155] In practical applications, improved accuracy and F1-score mean more precise member churn prediction, which can effectively optimize promotion strategies and avoid over-investment or waste of resources.

[0156] Computational complexity:

[0157] Compared with Improved Scheme 2 (multi-source data fusion) and Improved Scheme 3 (deep learning model), Implementation Example 1 has the lowest computational complexity and is suitable for real-time prediction of large-scale user data.

[0158] Example 1 can complete data processing and prediction in a short time, making it suitable for marketing campaigns that require rapid response.

[0159] Training time:

[0160] Example 1 has the shortest training time, requiring only 1 hour, indicating that the scheme has higher computational efficiency and can quickly adapt to changing market demands.

[0161] Budget utilization rate:

[0162] Because the promotion strategy was refined and optimized, the budget utilization rate of Example 1 was 98.5%, which is much higher than other options, indicating that the promotion funds were optimally allocated and the return on investment was improved.

[0163] Conclusion of the best implementation example

[0164] Taking all indicators into account, Example 1 demonstrates superior performance in terms of accuracy, computational efficiency, and budget utilization, proving itself to be the optimal example. Compared to other improved solutions, Example 1 not only achieves comparable predictive accuracy but also exhibits significant advantages in computational efficiency and resource utilization. Therefore, Example 1 is determined to be the optimal example of this invention and is suitable for marketing and promotional activities with high real-time requirements and large user groups.

[0165] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A marketing and promotion method based on member churn prediction, characterized in that, include: Step 1: Collect member behavior data, transaction data, social interaction data, and advertising data; Step 2: Preprocess the data, including data cleaning, feature selection, and data standardization; Step 3: Establish a member churn prediction model to predict whether members will churn; Step 4: Establish a promotion channel performance prediction model and evaluate the performance of each promotion channel; Step 5: Calculate the weight value of each promotion channel, which reflects the cost-effectiveness of the promotion channel; Step 6: Sort the promotion channels based on the weight values; Step 7: Adjust the promotion strategy based on the ranking results, and optimize the promotion campaign by adjusting the budget for different channels; Step 8: Monitor the implementation process of the promotion activities and adjust strategies based on actual feedback results; The weight value W of the promotion channel j The calculation formula is: Among them, W j Let P be the weight value of the j-th promotion channel. 流失 Let C be the predicted probability of member churn for the j-th channel. 转化率 For the conversion rate and cost of this channel j This refers to the promotion costs for this channel.

2. The marketing promotion method based on member churn prediction according to claim 1, characterized in that, The member churn prediction model is established using the XGBoost model, and the member churn prediction formula is as follows: Among them, P 流失 (i) represents the churn probability of the i-th member, σ is the Sigmoid function, and w j Let X be the weight of the j-th feature. ij Let be the value of the i-th member on the j-th feature, and b be the bias term.

3. The marketing promotion method based on member churn prediction according to claim 1, characterized in that, The promotion channel effectiveness prediction model is established using a regression model, and the formula for the regression model is: y=β0+β1X1+β2X2+…+β n X n Where y is the performance indicator of the promotion channel, X1, X2, ..., X n For the characteristics of this promotion channel, β0, β1, ..., β n These are the coefficients of the regression model.

4. The marketing promotion method based on member churn prediction according to claim 1, characterized in that, The formula for allocating the budget is as follows: Budget j =W j Total Budget Among them, the budget j Let $\frac{j}{j}$ be the promotion budget for the $j$-th channel, and $\frac{j}{j}$ be the total budget for the entire marketing campaign.

5. The marketing promotion method based on member churn prediction according to claim 1, characterized in that, The optimized promotion strategy is based on the multi-armed slot machine algorithm; The multi-armed slot machine algorithm is used to dynamically adjust the distribution ratio of each promotion channel to balance exploration and utilization, and ensure that the distribution strategy is adjusted based on historical data and real-time feedback.

6. The marketing promotion method based on member churn prediction according to claim 5, characterized in that, The monitoring steps include real-time monitoring of the promotion effects of various channels and adjusting the delivery strategy based on indicators such as churn rate and conversion rate.

7. The marketing promotion method based on member churn prediction according to claim 1, characterized in that, The method also includes regularly updating the member churn prediction model and the promotion channel effectiveness prediction model to cope with changes in the market and user behavior.

8. The marketing promotion method based on member churn prediction according to claim 1, characterized in that, The method adjusts promotion strategies based on user behavior data and market feedback.

Images