Customer relationship management optimization method based on artificial intelligence
Through the AI-driven customer relationship management method, the problems of manual dependence and strategy rigidity in traditional customer relationship management have been solved, accurate quantification of customer behavior and real-time strategy matching have been achieved, and the scientific nature of customer management and the adaptability of strategies have been improved.
Patent Information
- Application Number
- CN202510929616.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-07
- Publication Date
- 2025-10-10
AI Technical Summary
Traditional customer relationship management relies on manual experience, resulting in one-sided customer portraits, delayed strategy generation and a lack of dynamic tuning. It is unable to respond to real-time changes in customer status, and strategy thresholds are solidified for a long time, causing the strategy effectiveness to gradually become ineffective.
Adopting an AI-based customer relationship management method, we establish a multi-dimensional index evaluation system through time window division, data collection, data analysis and management optimization to achieve accurate quantification of customer behavior and real-time strategy matching.
It achieves objective stratification of customer value and deep feature mining, eliminating human subjectivity. The system completes behavioral analysis and strategy matching within seconds, ensuring close coordination between strategy and real-time behavior, and optimizes strategy accuracy and adaptability through a dynamic feedback loop.
Smart Images
Figure CN120765254A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and more specifically, to an artificial intelligence-based customer relationship management optimization method. Background Art
[0002] Customer relationship management has long relied on manual experience for customer value assessment and strategy formulation. Traditional solutions typically use a discrete operational process: first, operations personnel manually collect basic customer behavior data from multiple independent systems, then use basic data analysis tools to perform simple data statistics and report generation, and finally formulate standardized strategies based on general rules summarized from experience, and manually push preferential information or membership benefits.
[0003] This model has significant limitations: first, manual analysis cannot integrate multi-dimensional behavioral characteristics, resulting in a one-sided customer portrait; second, strategy generation relies on fixed rules and requires manual intervention, with a delay of several hours to several days from data collection to strategy execution, and is unable to respond to real-time changes in customer status; third, strategy thresholds are fixed for a long time, and there is a lack of a dynamic tuning mechanism based on effect feedback, which causes initially effective strategies to gradually become ineffective as the market changes. Summary of the Invention
[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a customer relationship management optimization method based on artificial intelligence, which solves the problems raised in the above-mentioned background technology through the following scheme.
[0005] To achieve the above objectives, the present invention provides the following technical solution: a customer relationship management optimization method based on artificial intelligence, comprising:
[0006] S1: Management time division: used to determine the relationship management time of target customers as the target time zone, and divide the target time zone into various time windows according to the time window division method;
[0007] S2: Data Collection: This is used to collect purchase behavior data, interaction behavior data, response behavior data, and loyalty behavior data for each time window, and clean the collected data.
[0008] S3: Data analysis: Analyze the data collected in S2 using a preset mathematical model, including the purchase behavior comprehensive index, interactive behavior quality index, response behavior effectiveness index, and loyalty value index;
[0009] S4: Management optimization: Management optimization is performed based on the data analysis results of S3.
[0010] Preferably, the purchase behavior data includes average order value, monthly purchase frequency, number of days since the last purchase, and total historical spending; the interactive behavior data includes weekly website visits, average session duration, page views per visit, and number of times items are added to the shopping cart but not purchased; the response behavior data includes email open rate, marketing campaign click-through rate, number of coupon uses, and number of advertising interactions; the loyalty behavior data includes customer relationship duration, repeat purchase ratio, number of new customers recommended, and number of times participation in loyalty activities.
[0011] Preferably, the purchasing behavior data directly collects the average order value through the enterprise order database or the transaction record table of the CRM system. The specific method is to calculate the average of all the customer's order amounts; collect the monthly purchase frequency by counting the number of orders placed by the customer each month according to the time range through the same system and calculating the average, usually with the past 12 months as a cycle; query the customer's most recent order date through the order database and subtract it from the current date to collect the number of days since the most recent purchase; directly collect the historical total consumption amount by summarizing the total amount field of all the customer's historical orders, and ensure data deduplication and unified currency units.
[0012] Preferably, the interactive behavior data is collected through the website or APP's tracking logs and analysis tools to collect the number of website visits per week, and customer visit events are aggregated by natural week; the average session duration is calculated through the session record table to collect the average value of the duration of each session, and robot traffic needs to be excluded; the number of page view events for each session is counted through the behavior log and the average is calculated to collect the page views for each visit; the number of times the shopping cart is added but not purchased is collected by filtering the records with the status of added but not settled through the e-commerce system shopping cart event log and counting the number of times on a monthly basis.
[0013] Preferably, the response behavior data directly collects the email open rate through the push report of the email marketing system, which is calculated as the number of independent users who opened the email divided by the total number of deliveries; collects the marketing activity click-through rate through the activity report of the advertising platform, which is calculated as the number of ad clicks divided by the number of impressions; collects the number of coupon uses on a monthly basis by customer ID through the redemption records of the enterprise coupon management system; and collects the number of advertising interactions through the social media API, including the monthly accumulation of likes, shares, and comments.
[0014] Preferably, the loyalty behavior data collects the customer relationship duration through the difference between the registration date field and the current date in the customer master data table of the CRM system, and the unit is converted into days; the repeat purchase ratio is collected by marking the first purchase identifier in the order database and counting the non-first order ratio; the number of recommended new customers is collected by counting the referrer ID in the successful recommendation relationship table recorded by the referral program tracking system; and the number of times the number of participation in loyalty activities is collected by filtering the participation event type by month through the activity participation log of the loyalty management system.
[0015] Preferably, the purchase behavior comprehensive index is specifically represented as: P1: average order value, P2: monthly purchase frequency, P3: number of days since the last purchase, P4: total historical consumption, c p : time smoothing constant, k p : dimension calibration constant.
[0016] Preferably, the interaction behavior quality index is specifically represented as: I1: number of website visits per week, I2: average session length, I3: page views per visit, I4: number of abandoned carts, c i : abandoned cart frequency smoothing constant, k i : dimension calibration constant.
[0017] Preferably, the response behavior effectiveness index is specifically represented as: R1: email open rate, R2: marketing click rate, R3: number of coupon uses, R4: number of ad interactions, c r : response smoothing constant, k r : dimension calibration constant.
[0018] Preferably, the loyalty value index is specifically represented as: L1: customer relationship length, L2: repeat purchase proportion, L3: number of new customers referred, L4: number of loyalty program participation, λ: referral and program weight constant, k l : dimension calibration constant.
[0019] Preferably, the customer purchase behavior index V p less than or equal to 30 triggers a 7x exclusive coupon and free shipping incentive to activate repeat purchases; V p greater than or equal to 70 pushes limited high-end goods and a buy-three-get-one-free bundled sale; the interaction behavior index V i less than or equal to 20, the customer starts a page loading acceleration engine and inserts an intelligent guidance pop-up window at the shopping cart stage; V i greater than or equal to 65, exclusive Beta function test permissions are opened and content creation incentives are provided; the response behavior index V r less than or equal to 25, the customer switches to a text-based simplified marketing material and uses a combination of SMS and push notifications for outreach; V r greater than or equal to 60, double-point time-limited privileges are opened and interactive lottery activities are designed; the loyalty index V l less than or equal to 35, the customer receives a tiered growth package including a $10 cashback on the first order and a gift on the second order; V lIf the number is greater than or equal to 75, the Black Card VIP qualification will be granted, which includes an exclusive customer service channel and a 5% commission reward based on the number of successful referrals. All strategy thresholds are dynamically calibrated every quarter based on the 25% and 75% percentiles.
[0020] Technical effects and advantages of the present invention:
[0021] 1. This invention uses an AI-driven, multi-dimensional indexation evaluation system to accurately quantify customer purchasing behavior, interaction quality, response effectiveness, and loyalty value. This completely eliminates the subjectivity and bias of manual experience, achieves objective stratification of customer value, and deeply explores its characteristics, providing a scientific basis for precise strategies.
[0022] 2. Relying on a fully automated real-time calculation process, the system completes customer behavior analysis, index diagnosis, and strategy matching within seconds, automatically triggering highly relevant intervention measures. This completely resolves the problem of lost strategy timeliness caused by manual delays in traditional solutions, ensuring close coordination between customer management strategies and real-time behavior status.
[0023] 3. This invention establishes a dynamic feedback loop between index thresholds and strategy effects, automatically calibrates the judgment intervals and matching strategies of the four major indexes based on historical execution data every quarter, continuously optimizes strategy accuracy and adaptability, and fundamentally breaks through the long-term performance degradation bottleneck caused by rigid rules in traditional solutions, thus realizing the self-evolution of the customer management system. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] Figure 1 It is a schematic diagram of the overall structure of the present invention. DETAILED DESCRIPTION
[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.
[0026] refer to Figure 1 The AI-based customer relationship management optimization method shown includes:
[0027] S1: Management time division: used to determine the relationship management time of target customers as the target time zone, and divide the target time zone into various time windows according to the time window division method.
[0028] S2: Data collection: used to collect purchase behavior data, interaction behavior data, response behavior data and loyalty behavior data in each time window, and perform data cleaning on the collected data.
[0029] The purchasing behavior data includes the average order value, monthly purchase frequency, the number of days since the last purchase, and the total historical consumption amount; the interactive behavior data includes the number of website visits per week, the average session duration, the number of page views per visit, and the number of times the shopping cart is added but not purchased; the response behavior data includes the email open rate, the marketing campaign click-through rate, the number of coupon uses, and the number of advertising interactions; the loyalty behavior data includes the length of customer relationships, the repeat purchase rate, the number of new customers recommended, and the number of times participation in loyalty activities.
[0030] The purchasing behavior data is collected directly from the enterprise order database or the transaction record table of the CRM system, with the specific method being to calculate the average amount of all customer orders; to collect the monthly purchase frequency by counting the number of orders placed by the customer each month according to the time range through the same system and calculating the average, usually with the past 12 months as a cycle; to collect the number of times since the last purchase by querying the customer's most recent order date through the order database and subtracting it from the current date; and to directly collect the historical total consumption amount by summarizing the total amount field of all historical orders of the customer, ensuring data deduplication and unified currency units.
[0031] The interactive behavior data is collected through the website or APP's tracking logs and analysis tools to collect the number of website visits per week, and customer visit events are aggregated by natural week; the average session duration is calculated through the session record table to collect the average session duration, and robot traffic needs to be excluded; the number of page view events for each session is counted through the behavior log and the average is calculated to collect the page views for each visit; the e-commerce system shopping cart event log is filtered out for records with a status of added but not settled, and the number of times added to the shopping cart but not purchased is collected by monthly statistics.
[0032] The response behavior data is collected directly through the push report of the email marketing system to obtain the email open rate, which is calculated as the number of independent users who opened the email divided by the total number of deliveries; the marketing activity click-through rate is collected through the activity report of the advertising platform, which is calculated as the number of ad clicks divided by the number of impressions; the number of coupon usages is collected on a monthly basis by customer ID through the redemption records of the enterprise coupon management system; the number of advertising interactions is collected through the social media API, including the monthly accumulation of likes, sharing and comments.
[0033] The loyalty behavior data is collected by collecting the customer relationship duration through the difference between the registration date field and the current date in the customer master data table of the CRM system, and the unit is converted into days; the repeat purchase ratio is collected by counting the non-first order ratio through marking the first purchase identifier in the order database; the number of recommended new customers is collected by counting the referrer ID through the successful recommendation relationship table recorded by the referral program tracking system; the number of times the number of participation in loyalty activities is collected by filtering the participation event type by month through the activity participation log of the loyalty management system.
[0034] S3: Data analysis: Analyze the data collected in S2 through a preset mathematical model, including the comprehensive purchase behavior index, interactive behavior quality index, response behavior effectiveness index, and loyalty value index.
[0035] The purchasing behavior comprehensive index is specifically expressed as: P1: average order value, P2: monthly purchase frequency, P3: number of days since the last purchase, P4: total historical spending, c p : time smoothing constant, k p : Dimensional calibration constant.
[0036] The purchasing behavior comprehensive index measures customer consumption value by integrating four key indicators: average order value reflects the quality of a single transaction, monthly purchase frequency reflects consumption activity, and the square root of the product of the two compresses the impact of extreme values to form a basic consumption capacity value; the logarithm of historical total consumption amount (adding 1 to avoid invalid calculation when new customers have not consumed) represents long-term contribution, and its logarithmic form weakens the absolute dominance of high-value consumption; the number of days of recent purchase is placed in the denominator as a negative indicator, and its square root structure (adding 1 to smooth the risk of division by zero caused by purchases on the same day) realizes the reverse adjustment of recent activity - the shorter the number of days, the smaller the denominator and the higher the index. In the constant calibration link, the time smoothing constant is fixed at 1 day (the minimum recording unit), and the dimensional constant is reversed based on the median of historical customer consumption values to ensure that the output value is within the business interpretable range (such as 0-100 points).
[0037] The interactive behavior quality index is specifically expressed as: I1: Number of website visits per week, I2: Average session duration, I3: Page views per visit, I4: Number of shopping cart abandonments, c i : Abandon the number of smoothing constants, k i : Dimensional calibration constant.
[0038] The Engagement Quality Index integrates three positive indicators and one negative indicator: the product of weekly visits, average session duration, and page views per visit represents engagement intensity, and is raised to the 2 / 3 power to achieve sublinear scaling (to prevent any one indicator from being overly dominant). Shopping cart abandonment is placed in the denominator as a negative indicator, directly penalizing abandonment (by adding 1 to avoid the denominator being invalid when there are zero abandonments). Its linear structure ensures that each additional abandonment weakens the index value. The derivation balances the contributions of the three positive indicators through the 2 / 3 power, creating a nonlinear synergistic effect between visit frequency, duration, and depth. The denominator for abandonment uses a linear penalty rather than a square root to emphasize its direct negative impact on conversion rate. The dimensional constant is calibrated based on the 90th percentile of historical engagement data, so that a high score represents high-quality engagement.
[0039] The response behavior efficacy index is specifically expressed as: R1: email opening rate, R2: marketing click-through rate, R3: number of coupon usage, R4: number of advertising interactions, c r : response smoothing constant, k r : Dimensional calibration constant.
[0040] The response behavior effectiveness index hierarchically processes two types of indicators: the product of email open rate and marketing click-through rate reflects the efficiency of the response link, and its logarithmic transformation (adding 1 to ensure that the zero value is valid) compresses the percentage magnitude to a reasonable range; the number of coupon uses and the number of advertising interactions are directly summed to quantify the total amount of response actions, and the numerator multiplies the logarithmic result of the response efficiency by the sum of the response actions to form a "high efficiency + high frequency" combination effect; the denominator takes the square root of the sum of the response actions (adding 1 to prevent zero value), and introduces a saturation adjustment mechanism - when the response action is extremely high, the gain decreases to avoid excessive rewarding low-frequency and high-response rate behaviors. In the derivation, the nested structure of multiplication and square root is used to naturally construct a dynamic balance between response quality and action quantity. The smoothing constant is set to 1, and the dimensional constant needs to be calibrated to match the peak data of the marketing activity.
[0041] The loyalty value index is specifically expressed as: L1: Customer relationship duration, L2: Repeat purchase ratio, L3: Number of new customers recommended, L4: Number of participation in loyalty activities, λ: Recommendation and activity weight constant, k l : Dimensional calibration constant.
[0042] The loyalty value index is decomposed into a time-based and behavioral value-added module: the logarithm of the customer relationship duration (plus 1 to handle new customers) reflects the marginal diminishing effect, and multiplied by the repeat purchase ratio (converted to a decimal) strengthens the long-term repurchase value; the geometric mean (square root) of the number of new customers recommended and the number of loyalty activities quantifies the synergistic effect of word-of-mouth and investment, and a weight constant is added in front to balance the contribution of the two types of behaviors. In the derivation, the absolute duration advantage of long-life cycle customers is compressed by logarithm, and the geometric mean weakens the order of magnitude difference between the number of recommendations and the number of activities; the weight constant needs to be optimized through regression analysis (such as maximizing the correlation with customer retention rate), and the dimensional constant scales the output range based on the top loyal customer data.
[0043] S4: Management optimization: Management optimization is performed based on the data analysis results of S3.
[0044] The customer purchasing behavior index V p When the price is less than or equal to 30, a 30% discount coupon and free shipping incentives will be immediately triggered to activate repeat purchases; V p If the number is greater than or equal to 70, limited edition high-end products and buy three get one free bundle sales will be pushed; the interactive behavior index V i For customers with less than or equal to 20, the page loading acceleration engine is activated and a smart guide pop-up window is inserted in the shopping cart section; iIf the response index is greater than or equal to 65, exclusive Beta function testing rights will be opened and content creation incentives will be provided; the response behavior index V r For customers less than or equal to 25, the company switches to simplified text and picture marketing materials and adopts SMS and push dual-channel reach; V r If the loyalty index is greater than or equal to 60, a limited-time privilege of double points will be opened and an interactive lottery event will be designed; the loyalty index V l For customers with less than or equal to 35, a step-by-step growth gift package will be implemented, including a $10 cashback on the first order and a gift for the second order; l If the number is greater than or equal to 75, the Black Card VIP qualification will be granted, which includes an exclusive customer service channel and a 5% commission reward based on the number of successful referrals. All strategy thresholds are dynamically calibrated every quarter based on the 25% and 75% percentiles.
[0045] The present invention first divides management time, sets the relationship management cycle of target customers as the target time zone, and subdivides it into multiple continuous time windows by using the time window division method; then performs data acquisition, and the system automatically collects the customer's purchase behavior data, interactive behavior data, response behavior data and loyalty behavior data in each time window, and cleans the original data to ensure quality; then enters the data analysis stage, processes the cleaned data through a preset mathematical model, and calculates the purchase behavior comprehensive index that represents the customer's consumption value, the interactive behavior quality index that reflects the degree of participation, and the response behavior index that quantifies the effectiveness of marketing feedback. It is an effectiveness index and a loyalty value index for evaluating long-term relationships; finally, management optimization is carried out according to the index analysis results: when the purchase behavior index is low, exclusive coupons and free shipping are triggered to activate repeat purchases; when it is high, high-end product bundle sales are pushed; for customers with a low interaction index, page acceleration and shopping cart guidance pop-ups are enabled; for customers with a high interaction index, testing permissions are opened and creative incentives are provided; for customers with a low response index, simplified materials are switched and dual-channel contact is enabled; for customers with a high response index, double points and interactive lotteries are opened; for customers with a low loyalty index, a step-by-step growth gift package is implemented; for customers with a high loyalty index, black card VIP qualifications are granted and referral commissions are issued. All strategy thresholds are dynamically calibrated every quarter based on historical data.
[0046] Secondly: The drawings of the embodiments disclosed in the present invention only involve structures related to the embodiments disclosed in the present invention. Other structures may refer to conventional designs. The same embodiment and different embodiments of the present invention may be combined with each other without conflict.
[0047] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. The customer relationship management optimization method based on artificial intelligence is characterized by: include: S1: Management time division: used to determine the relationship management time of target customers as the target time zone, and divide the target time zone into various time windows according to the time window division method; S2: Data Collection: This is used to collect purchase behavior data, interaction behavior data, response behavior data, and loyalty behavior data for each time window, and clean the collected data. S3: Data analysis: Analyze the data collected in S2 using a preset mathematical model, including the purchase behavior comprehensive index, interactive behavior quality index, response behavior effectiveness index, and loyalty value index; S4: Management optimization: Management optimization is performed based on the data analysis results of S3.
2. The artificial intelligence-based customer relationship management optimization method according to claim 1, characterized in that: The purchase behavior data includes average order value, monthly purchase frequency, number of days since the last purchase, and historical total spending; the interactive behavior data includes weekly website visits, average session duration, page views per visit, and number of times items are added to the shopping cart but not purchased; The response behavior data includes email open rate, marketing campaign click-through rate, number of coupon uses and number of advertising interactions; the loyalty behavior data includes customer relationship duration, repeat purchase rate, number of new customers recommended and number of participation in loyalty activities.
3. The artificial intelligence-based customer relationship management optimization method according to claim 2, characterized in that: The purchasing behavior data is collected directly from the enterprise order database or the transaction record table of the CRM system, and the specific method is to calculate the average order value of all customer orders. The monthly purchase frequency is collected by counting the number of orders placed by customers per month by time range and calculating the average through the same system, usually taking the past 12 months as a period; Query the customer's most recent order date in the order database and subtract it from the current date to collect the number of days since the most recent purchase; directly collect the historical total consumption amount by summarizing the total amount field of all the customer's historical orders. Ensure data deduplication and consistency of currency units.
4. The artificial intelligence-based customer relationship management optimization method according to claim 2, characterized in that: The interactive behavior data is collected through the website or app's tracking logs and analysis tools to collect the number of website visits per week, aggregating customer visit events by natural week; the average duration of each session is calculated through the session record table to collect the average session duration, excluding robot traffic; The number of page views in each session is counted through behavioral logs and the average is calculated to collect the page views for each visit; the number of times the shopping cart is added but not purchased is collected by filtering the records with the status of added but not settled through the e-commerce system shopping cart event logs and counting the number of times on a monthly basis.
5. The artificial intelligence-based customer relationship management optimization method according to claim 2, characterized in that: The response behavior data is collected directly through the push report of the email marketing system, and the email open rate is calculated as the number of unique users who opened the email divided by the total number of emails delivered; the marketing campaign click-through rate is collected through the activity report of the advertising platform, and is calculated as the number of ad clicks divided by the number of impressions; the number of coupon usages is collected by customer ID and aggregated monthly through the redemption records of the enterprise coupon management system; The number of advertising interactions is collected through social media APIs, including likes, shares, and comments, which are accumulated monthly.
6. The artificial intelligence-based customer relationship management optimization method according to claim 2, characterized in that: The loyalty behavior data is collected by the difference between the registration date field and the current date in the customer master data table of the CRM system, and the unit is converted into days; By marking the first purchase identification in the order database, the proportion of non-first orders is counted and the proportion of repeat purchases is collected; the number of new customers recommended is collected by counting the referrer ID through the successful recommendation relationship table recorded by the referral program tracking system; the number of times participation in loyalty activities is collected by filtering the participation event type by month through the activity participation log of the loyalty management system.
7. The artificial intelligence-based customer relationship management optimization method according to claim 1, characterized in that: The purchasing behavior comprehensive index is specifically expressed as: P1: average order value, P2: monthly purchase frequency, P3: number of days since the last purchase, P4: total historical spending, c p : time smoothing constant, k p : Dimensional calibration constant.
8. The artificial intelligence-based customer relationship management optimization method according to claim 1, characterized in that: The interactive behavior quality index is specifically expressed as: I1: Number of website visits per week, I2: Average session duration, I3: Page views per visit, I4: Number of shopping cart abandonments, c i : Abandon the number of smoothing constants, k i : Dimensional calibration constant.
9. The artificial intelligence-based customer relationship management optimization method according to claim 1, characterized in that: The response behavior efficacy index is specifically expressed as: R1: email opening rate, R2: marketing click-through rate, R3: number of coupon usage, R4: number of advertising interactions, c r : response smoothing constant, k r : Dimensional calibration constant.
10. The artificial intelligence-based customer relationship management optimization method according to claim 1, characterized in that: The loyalty value index is specifically expressed as: L1: Customer relationship duration, L2: Repeat purchase ratio, L3: Number of new customers recommended, L4: Number of participation in loyalty activities, λ: Recommendation and activity weight constant, k l : Dimensional calibration constant.