Member data management system

Abstract

The invention provides a member data management system, which belongs to the technical field of data management, captures member behavior changes in real time through a time sequence behavior analysis engine, solves the problem of cross-system data islands by utilizing a data consanguinity tracking module, and realizes second-level updating of member portraits by adopting a real-time portrait updating component. The system further comprises a dynamic right and interest matching device and an intelligent content generation unit, personalized right and interest recommendation can be generated according to real-time portraits and external event data, and compliant marketing content can be automatically produced. According to the invention, the technical problems of static label management, data integration difficulty, portrait updating lagging and the like of the existing member management system are solved, and accurate dynamic management of the whole life cycle of members is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of membership data management, and in particular to a membership data management system. Background Art

[0002] Existing membership management systems have problems such as static label management, data silo effect, and lagging portrait updates. Traditional systems rely on batch data processing, cannot capture changes in membership behavior in real time, and have difficulty integrating data across systems, resulting in low marketing accuracy and poor member experience. Therefore, a membership data management system is proposed to solve the above problems. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to provide a membership data management system to solve at least the above problems.

[0004] The technical solution adopted by the present invention is as follows:

[0005] A membership data management system, comprising:

[0006] A time-series behavior analysis engine, a data lineage tracking module, a real-time portrait update component, a dynamic rights and interests matching device, and an intelligent content generation unit;

[0007] The time-series behavior analysis engine is used to capture changes in membership behavior patterns in real time through multi-dimensional behavior sequence analysis and generate a dynamic label set that evolves over time;

[0008] The data lineage tracking module constructs a cross-system data entity relationship network based on knowledge graph technology to achieve metadata-level association of CRM, ERP, and POS systems;

[0009] The real-time portrait update component is data-connected to the time-series behavior analysis engine and uses a stream-batch integrated processing architecture to achieve second-level updates of the membership portrait;

[0010] The dynamic rights and interests matching device is respectively connected to the real-time portrait update component and the data lineage tracking module, and is used to generate a scenario-based rights and interests recommendation plan;

[0011] The intelligent content generation unit is communicatively connected to the dynamic rights and interests matching device and is used to achieve automated production of marketing content.

[0012] Further, the time-series behavior analysis engine includes:

[0013] A behavior mutation detection sub-module, which is used to compare the deviation degree of the current behavior sequence from the historical benchmark through a sliding window algorithm;

[0014] A life cycle stage determination sub-module, which is used to automatically divide the membership life cycle stage through RFM index clustering analysis;

[0015] Cold start guidance sub-module, used to build a personalized guidance path based on the behavior patterns of similar member groups;

[0016] Among them, the behavior mutation detection sub-module is configured to trigger a warning signal when it detects that the consumption frequency has decreased by more than 40% within 3 consecutive cycles.

[0017] Furthermore, the five-stage model established by the life cycle stage determination sub-module includes: [[ID=u8]]

[0018] Introduction stage, characterized by the behavior pattern within 30 days after the first registration;

[0019] Growth stage, characterized by a monthly growth rate of consumption frequency exceeding 15%;

[0020] Maturity stage, characterized by the consumption behavior being stable within a preset threshold range;

[0021] Decline stage, characterized by the key indicators decreasing by more than 20% in 2 consecutive cycles;

[0022] Churn stage, characterized by reaching the preset threshold of inactive days.

[0023] Furthermore, the data lineage tracking module includes:

[0024] Metadata collection component, used to connect to each business system through the adapter pattern and extract metadata;

[0025] Graph construction component, used to store data entity nodes and relationship edges using a graph database;

[0026] Conflict detection component, used to implement a three-layer verification mechanism;

[0027] The three-layer verification mechanism includes:

[0028] Uniqueness verification rule, used to ensure the unique correspondence of member identifiers across systems;

[0029] Business logic verification rule, used to verify the business logic consistency between data;

[0030] Timeliness verification rule, used to check the timeliness of data updates.

[0031] Furthermore, the conflict detection component further includes:

[0032] Confidence score unit, used to score and classify the detected data conflicts;

[0033] Data repair workflow engine, used to execute different processing strategies according to the confidence score;

[0034] Among them, for data conflicts with confidence scores higher than the first threshold, automatic repair is performed, and for those with confidence scores lower than the second threshold, a manual intervention task order is generated.

[0035] Further, the real-time portrait update component includes:

[0036] A streaming data processing interface for accessing data streams of multiple message protocols;

[0037] An interest migration analysis unit for calculating an interest migration index through a category association graph;

[0038] A life cycle node association unit for establishing a dynamic time node network;

[0039] The interest migration analysis unit is configured to trigger portrait update when the migration index of the comparison between the recent 7-day consumption record and the historical preference exceeds a preset threshold.

[0040] Further, it also includes:

[0041] A hybrid query optimizer for implementing a joint query of a graph database and a relational database;

[0042] The hybrid query optimizer includes:

[0043] A query rewriting unit for converting a graph query statement into a relational query statement;

[0044] A response time control unit for ensuring that the cross-database query response time does not exceed 200 ms;

[0045] A cache management unit for caching and optimizing high-frequency query results.

[0046] Compared with the prior art, the beneficial effects of the present invention are:

[0047] The present invention solves the technical problems existing in the existing membership management system, such as static label management, difficult data integration, and lagging portrait update, and realizes the accurate dynamic management of the entire membership life cycle. Description of the Drawings

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only the preferred embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0049] Figure 1 It is a schematic diagram of the overall structure of a membership data management system proposed in an embodiment of the present invention. Detailed Embodiments

[0050] The principles and features of the present invention will be described below in conjunction with the accompanying drawings. The listed embodiments are only used to explain the present invention and are not intended to limit the scope of the present invention.

[0051] Referring to Figure 1 , the present invention provides a membership data management system, which is characterized by including:

[0052] A time-series behavior analysis engine, a data lineage tracking module, a real-time portrait update component, a dynamic rights and interests matching device, and an intelligent content generation unit;

[0053] The time-series behavior analysis engine is used to capture the changes in the membership behavior patterns in real time through multi-dimensional behavior sequence analysis and generate a dynamic tag set that evolves over time;

[0054] The data lineage tracking module constructs a cross-system data entity relationship network based on knowledge graph technology to achieve metadata-level association of CRM, ERP, and POS systems;

[0055] The real-time portrait update component is data-connected to the time-series behavior analysis engine and uses a stream-batch integrated processing architecture to achieve second-level updates of the membership portrait;

[0056] The dynamic rights and interests matching device is respectively connected to the real-time portrait update component and the data lineage tracking module and is used to generate a scenario-based rights and interests recommendation plan;

[0057] The intelligent content generation unit is communicatively connected to the dynamic rights and interests matching device and is used to achieve automated production of marketing content.

[0058] Exemplarily, the time-series behavior analysis engine generates the label "Healthy Lifestyle Transformer" by analyzing the consumption sequence of member "Ms. Zhang" in the past 3 months (buying imported fruits every Tuesday evening → switching to domestic fruits + fitness equipment in the past 2 weeks) and combining it with the category association graph. If it detects that member "Mr. Li" has not visited the store for 3 consecutive weeks (with an average monthly consumption of 4 times in the past), it triggers the "Potential Churn Warning" label and pushes it to the dynamic rights and interests matching device. For the cross-system association of the data lineage tracking module, when the POS system records that "Mr. Wang" buys baby milk powder, the module automatically associates the "New Father" label in the CRM and the milk powder inventory data in the ERP to construct a knowledge graph node of "Parenting Needs". For conflict detection, when it is found that the birthday information of member "Ms. Zhao" in the CRM and ERP systems differs by 3 days, it activates the aging verification rule and automatically repairs it using the birthday data in the latest POS consumption record. For the real-time portrait update component's stream-batch processing, by capturing the behavior of member "Ms. Chen" browsing air fryers on the APP in real time (stream processing) and combining it with the batch data of historical purchases of small household appliances (batch processing), it updates the portrait dimension of "Kitchen Upgrade Needs" within 87 ms, thus achieving a second-level update of the member portrait. For interest migration, when it detects that member "Mr. Zhou" has switched from consuming digital products to outdoor equipment and calculates that the migration index reaches 82% (threshold 75%), it triggers an adjustment of the portrait dimension weights. For the scenario-based recommendation of the dynamic rights and interests matching device, by targeting the healthy transformation label of "Ms. Zhang", it matches a 50% discount coupon for fitness courses (validated by ERP inventory data) + a fruit full reduction coupon (linked with the POS system) to form a "Healthy Living Package". For lifecycle adaptation, by identifying that "Mr. Li" has entered the churn period, it automatically upgrades the rights and interests to "Exclusive Customer Service + Double Points" and verifies the feasibility of the points redemption rule through the data lineage module. For the personalized production of the intelligent content generation unit, based on the rights and interests matching result, it calls the "Health Care" template and inserts the yoga mat model that "Ms. Zhang" recently searched for to generate the copy: "The yoga mat you are concerned about has arrived! Save 200 yuan with the fitness course coupon >>". For channel adaptation, according to the APP usage habits of member "Mr. Li", it generates a push notification instead of a text message and inserts the navigation link to his historical high-frequency consumption store. In this system, when member "Mr. Li" receives the churn warning rights and interests package, the system traces through the data lineage module that he redeemed double points to buy a newly launched coffee machine, and the real-time portrait update component immediately marks the "Coffee Lover" label, triggering subsequent recommendations for coffee bean peripherals rights and interests, forming a closed-loop operation.

[0059] The time-series behavior analysis engine includes:

[0060] The behavior mutation detection sub-module is used to compare the deviation degree of the current behavior sequence from the historical benchmark through the sliding window algorithm;

[0061] The lifecycle stage determination sub-module is used to automatically divide the membership lifecycle stage through RFM index clustering analysis;

[0062] The cold start guidance sub-module is used to construct a personalized guidance path based on the behavior patterns of similar membership groups;

[0063] Among them, the behavior mutation detection sub-module is configured to trigger a warning signal when it detects that the consumption frequency drops by more than 40% within 3 consecutive cycles.

[0064] The five-stage model established by the lifecycle stage determination sub-module includes:

[0065] The introduction stage, which is characterized by the behavior pattern within 30 days after the first registration;

[0066] The growth stage, which is characterized by a monthly growth rate of consumption frequency exceeding 15%;

[0067] The maturity stage, which is characterized by the consumption behavior being stable within a preset threshold range;

[0068] The decline stage, which is characterized by the key indicators dropping by more than 20% in 2 consecutive cycles;

[0069] The churn stage, which is characterized by reaching a preset threshold of inactive days.

[0070] Exemplarily, for the behavior mutation detection sub-module, regarding the behavior of member A of an e-commerce platform who used to consume 4 times per month on average in the past 6 months and only consumed 1 time in the last 3 weeks (0.3 times per week on average), the behavior mutation detection sub-module can adopt a sliding window algorithm to calculate the consumption frequency with a 3-week window (current window value: 0.3 times / week vs historical benchmark: 1 time / week). When it detects a 70% decrease (exceeding the preset threshold of 40%) and it lasts for 3 cycles, the system generates a warning signal, pushes it to the operation staff and initiates the process of retaining lost customers; for the life cycle stage determination sub-module, different life cycles are determined by the following settings. Introduction stage: 28 days after user B registers and only completes 1 first order purchase, the new user exclusive gift package activation process is pushed. Growth stage: The monthly consumption frequency growth rate of user C in the last 3 months reaches 22% (from 2 times → 3 times → 4 times), and the membership level rights and interests are upgraded. Maturity stage: User D's monthly average consumption in the last 6 consecutive months is stably in the range of 3 - 5 times, and exclusive activities for loyal users are recommended. Decline stage: The consumption amount of user E has decreased by 25% in the last 2 months (from ￥800 → ￥600), and the consumption interval has extended, and the awakening mechanism is started to send exclusive coupons. Churn stage: User F has no login record for 90 consecutive days (exceeding the preset threshold of 60 days) and is transferred to the special operation pool for silent users. Then, based on RFM (Recency / Frequency / Monetary), through clustering analysis, a certain beauty brand realizes the improvement of the membership hierarchical operation efficiency through this model; for the cold start guidance sub-module, when new user G has no consumption behavior after registration, it can identify the group with a similar profile to G (such as: 25 - 30 years old / metropolitan area / caring about skin care products). By pushing a "￥5 no-threshold coupon for new users", if it is not used, a "list of popular skin care products" is pushed 3 days later. It can also dynamically adjust the recommended content in combination with the real-time profile update. In this system, when it detects the failure of cold start guidance, it automatically triggers the data lineage tracking module to check the cross-system behavior of users (such as whether they consume in offline stores), so as to achieve the operation coordination of the whole channel.

[0071] The data lineage tracking module includes:

[0072] The metadata collection component is used to connect to each business system through the adapter mode and extract metadata;

[0073] The graph construction component is used to store data entity nodes and relationship edges using a graph database;

[0074] The conflict detection component is used to implement a three-layer verification mechanism;

[0075] The three-layer verification mechanism includes:

[0076] The uniqueness verification rule is used to ensure the unique correspondence of the member identifier across systems;

[0077] The business logic verification rule is used to verify the business logic consistency between data;

[0078] The timeliness verification rule is used to check the timeliness of data updates.

[0079] Exemplarily, for the metadata collection component, assuming that there is fragmented data of the same member in the CRM system (member ID: A123), ERP system (customer number: B456), and POS system (payment account: C789) of a retail enterprise, the metadata collection component extracts the metadata fields of each system in real time through the adapter pattern (such as the JDBC / API interface), for example:

[0080] CRM: Member registration time, contact information, level label

[0081] ERP: Historical order records, return records, customer complaint records

[0082] POS: Real-time consumption amount, payment channel, shopping basket items

[0083] For the graph construction component, the collected metadata is converted into nodes and relationship edges in a graph database (such as Neo4j):

[0084] Nodes: Member node (A123), order node (B456), product node (such as "coffee machine")

[0085] Relationship edges: A123 - "consume" -> order B456 (amount $150), order B456 - "contain" -> coffee machine (quantity 1)

[0086] Through this network, the behavior path of member A123 purchasing a coffee machine can be traced, and the data lineage of the member in different systems can be associated.

[0087] For the uniqueness verification rule in the three-layer verification mechanism of the conflict detection component, when it is detected that the member ID (A123) in the CRM system and the customer number (B456) in the ERP system match the same entity through the mobile phone number, but there is an unassociated payment account (C789) in the POS system, the system triggers a uniqueness conflict. At this time, the system confirms that C789 belongs to A123 through secondary verification of the mobile phone number and automatically completes the association relationship; for the business logic verification rule, when an order is recorded as "completed" in the ERP, but the payment status of the order shows "uncleared" in the POS system, the business logic verification rule identifies the conflict through a predefined "order status - payment status" mapping table (such as "completed" must correspond to "cleared") and generates an exception log for verification; for the timeliness verification rule, in the POS system where transaction data is synchronized every 15 minutes, if a certain synchronization is delayed by more than 30 minutes (the timeliness threshold), the system marks this batch of data as "lagged", gives priority to processing other real-time stream data, and simultaneously triggers an alarm to notify the operation and maintenance team.

[0088] The conflict detection component further includes:

[0089] A confidence scoring unit for scoring and classifying the detected data conflicts;

[0090] A data repair workflow engine for executing a differential processing strategy according to the confidence score;

[0091] Among them, for data conflicts with a confidence score higher than the first threshold, automatic repair is performed, and for those with a confidence score lower than the second threshold, a manual intervention task sheet is generated.

[0092] Exemplarily, in a certain retail enterprise, the member data is scattered in the CRM system (recording member basic information), the POS system (recording consumption flow), and the ERP system (recording inventory and order data). Due to different data update frequencies and rules in each system, the "level label" of the same member conflicts in the three systems:

[0093] CRM system: Marks member A as a "gold member" (rule: historical consumption reaches 5000 yuan)

[0094] POS system: Marks member A as an "ordinary member" (because the recent consumption has not been synchronized to the CRM)

[0095] ERP system: Marks member A as a "diamond member" (due to misassociating the order data of other members)

[0096] By using the adapter pattern to extract the metadata of member A from the three systems, it is found that the "level label" is inconsistent; after performing uniqueness verification and confirming that the data of the three systems belongs to the same entity through the member ID (such as mobile phone number + ID card number), business logic verification is carried out: checking the member level calculation rules of each system, it is found that: the CRM rule is reasonable (historical consumption meets the standard)

[0097] The POS system has not synchronized the latest consumption record (the timeliness verification fails)

[0098] The ERP system misassociates the order (the business logic verification fails)

[0099] Then the following operations of the confidence scoring unit are carried out:

[0100] Conflict Dimension Scoring Criteria Current Score Data Source Priority CRM (Master Data System) > POS > ERP 80 Time Freshness The CRM data update lags behind the POS system by 3 days -20 Rule Rigor ERP error associated with the order (human operation error) -30 Impact Scope Affects core operations such as rights and interests issuance and inventory reservation +40 Total Score 70

[0101] For the decision-making of the data repair workflow engine, set the following thresholds: First threshold = 85 (automatic repair), Second threshold = 50 (manual intervention); the current conflict score is 70: between the first threshold and the second threshold, trigger semi-automated processing: Automatically repair the POS system: Synchronize the member level of the POS system to "Gold Member" (because the CRM is the main data source and the business rules are correct), and generate a manual intervention task order: Task type: Data error correction

[0102] Priority: High (because it affects the rights and interests distribution)

[0103] Suggested operation: Verify the order association logic of the ERP system and manually correct the incorrect data

[0104] Associated evidence chain: Provide screenshots of CRM consumption records and a list of ERP error order IDs

[0105] The real-time portrait update component includes:

[0106] A streaming data processing interface for accessing data streams of multiple message protocols;

[0107] An interest migration analysis unit for calculating the interest migration index through a category association graph;

[0108] A lifecycle node association unit for establishing a dynamic time node network;

[0109] The interest migration analysis unit is configured to trigger portrait update when the migration index of the comparison between the consumption records in the last 7 days and the historical preferences exceeds a preset threshold.

[0110] Exemplarily, as a data access layer, the streaming data processing interface can support multi-protocol adaptation such as Kafka, MQTT, HTTP / 2, etc., so as to capture user behavior events (such as clicks, add-to-carts, payments) and business data (such as orders, returns) in real time, and then convert the original events into structured feature vectors (such as user ID, behavior type, timestamp, product category), and distribute them to the downstream analysis units.

[0111] For the interest migration analysis unit, a category association graph can be constructed: Based on the commodity co-occurrence matrix and deep learning embedding, construct a transfer probability matrix between categories, and calculate the migration index. When the migration index within a continuous 7-day sliding window exceeds the dynamic threshold, mark the portrait update requirement.

[0112] For the lifecycle node association unit, a dynamic network can be constructed, including key nodes such as registration, first order, repeat purchase, silence (no activity for 30 days), churn (no activity for 90 days), etc., and then calculate the transfer probability between nodes through survival analysis, and combine with RNN to predict the occurrence time of the next node.

[0113] For portrait enhancement: use the life cycle stage as a portrait dimension (such as "potential churn users") and integrate it with the interest migration results to generate a multi-dimensional portrait that includes short-term interest shifts and long-term value predictions.

[0114] This embodiment also includes:

[0115] Hybrid query optimizer, used to implement joint queries of graph databases and relational databases;

[0116] The hybrid query optimizer includes:

[0117] A query rewriting unit, used to convert graph query statements into relational query statements;

[0118] Response time control unit, used to ensure that cross-database query response time does not exceed 200ms;

[0119] The cache management unit is used to optimize the cache of high-frequency query results.

[0120] For example, a relational query involves querying member transaction records (such as order amounts and product categories) from a MySQL database, and a graph query involves querying member social relationship chains (such as friend networks and community influence) from a Neo4j graph database. The query rewrite unit converts graph query statements into relational query statements. The response time control unit can split graph queries and relational queries into two subtasks, which are executed in parallel through a thread pool. The result set is compressed, i.e., columnar storage is used for intermediate results (such as the Apache Arrow format), thereby reducing data transmission volume. When monitoring finds that "Diamond Member" queries account for 45% of daily requests, the cache management unit can perform cache tiering:

[0121] L1 cache: Redis stores the query results of the last hour (TTL = 3600s)

[0122] L2 cache: Disk cache stores historical high-frequency queries (such as the monthly Top 100 members)

[0123] When a membership level change or social relationship update is detected, the relevant cache can be automatically cleared.

[0124] 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. A membership data management system, characterized in that, Including: A time-series behavior analysis engine, a data lineage tracking module, a real-time portrait update component, a dynamic rights and interests matching device, and an intelligent content generation unit; The time-series behavior analysis engine is used to capture changes in member behavior patterns in real time through multi-dimensional behavior sequence analysis and generate a dynamic tag set that evolves over time; The data lineage tracking module constructs a cross-system data entity relationship network based on knowledge graph technology to achieve metadata-level association of CRM, ERP, and POS systems; The real-time portrait update component is data-connected to the time-series behavior analysis engine and uses a stream-batch integrated processing architecture to achieve second-level updates of member portraits; The dynamic rights and interests matching device is respectively connected to the real-time portrait update component and the data lineage tracking module and is used to generate a scenario-based rights and interests recommendation plan; The intelligent content generation unit is communicatively connected to the dynamic rights and interests matching device and is used to realize automated production of marketing content.

2. The membership data management system according to claim 1, characterized in that, The time-series behavior analysis engine includes: A behavior mutation detection sub-module, which is used to compare the deviation degree between the current behavior sequence and the historical benchmark through a sliding window algorithm; A life cycle stage determination sub-module, which is used to automatically divide the member life cycle stage through RFM index clustering analysis; A cold start guidance sub-module, which is used to construct a personalized guidance path based on the behavior patterns of similar member groups; Among them, the behavior mutation detection sub-module is configured to trigger an early warning signal when it detects that the consumption frequency drops by more than 40% within 3 consecutive cycles.

3. The member data management system according to claim 2, characterized in that: The five-stage model established by the life cycle stage determination sub-module includes: The introduction stage, which is characterized by the behavior pattern within 30 days after the first registration; The growth stage, which is characterized by a monthly growth rate of consumption frequency exceeding 15%; The maturity stage, which is characterized by stable consumption behavior within a preset threshold range; The decline stage, which is characterized by a decrease in key indicators by more than 20% in 2 consecutive cycles; The churn stage, which is characterized by reaching a preset threshold of inactive days.

4. The membership data management system according to claim 1, wherein The data lineage tracking module includes: A metadata collection component, which is used to connect to each business system through the adapter mode and extract metadata; A graph construction component, which is used to store data entity nodes and relationship edges using a graph database; A conflict detection component, which is used to implement a three-layer verification mechanism; The three-layer verification mechanism includes: A uniqueness verification rule, which is used to ensure the unique correspondence of member identifiers across systems; A business logic verification rule, which is used to verify the business logic consistency between data; A timeliness verification rule, which is used to check the timeliness of data updates.

5. The member data management system according to claim 4, characterized in that: The conflict detection component further includes: A confidence score unit, which is used to score and classify detected data conflicts; A data repair workflow engine, which is used to execute a differential processing strategy based on the confidence score; Among them, automatic repair is performed on data conflicts with a confidence score higher than the first threshold, and a manual intervention task order is generated for those with a confidence score lower than the second threshold.

6. The membership data management system according to claim 1, wherein The real-time portrait update component includes: A stream data processing interface, which is used to access data streams of multiple message protocols; An interest migration analysis unit, which is used to calculate the interest migration index through a category association graph; A life cycle node association unit, which is used to establish a dynamic time node network; The interest migration analysis unit is configured to trigger a portrait update when it detects that the migration index of the consumption records in the last 7 days compared with the historical preferences exceeds a preset threshold.

7. The member data management system according to any one of claims 1 to 6, characterized in that: Also includes: Hybrid query optimizer, used to implement joint queries of graph databases and relational databases; The hybrid query optimizer includes: A query rewriting unit, used to convert graph query statements into relational query statements; Response time control unit, used to ensure that cross-database query response time does not exceed 200ms; The cache management unit is used to optimize the cache of high-frequency query results.

Images