Multi-source data-based SCRM customer intelligent portraying system and method

Abstract

The invention discloses an SCRM customer intelligent portraying system and method based on multi-source data, and the system comprises a multi-source data collection and integration module which is used for collecting and cleaning the internal and external multi-source data of an enterprise; the client dynamic label engine is used for generating labels based on business rules and calculating dynamic weights based on data sources and recency; the client intelligent portrait generation and updating module generates a unified portrait through a multi-dimensional portrait model and a portrait fusion device and tracks version evolution; the interactive portrait application and feedback module provides visual display, intelligent early warning and strategy reaching functions and optimizes a label rule through a feedback closed loop; according to the method, internal and external multi-source data of an enterprise are integrated and uniformly coded, so that a data island phenomenon is broken, a multi-dimensional portrait model covering basic identities, transaction values, social behaviors and dynamic intentions is constructed, and a uniform customer panoramic view is formed.

Description

Technical Field

[0001] This invention relates to the fields of big data and artificial intelligence technology, and in particular to an SCRM customer intelligent profiling system and method based on multi-source data. Background Technology

[0002] In existing customer relationship management, enterprises typically use customer profiling methods based on internal business data for management. Some systems attempt to integrate social media information or other external data sources. These methods often rely on single or limited data channels, such as customer transaction records or basic interaction data from social platforms, and generate static tags through predefined rules to support marketing decisions.

[0003] The aforementioned methods suffer from a lack of effective integration between internal enterprise systems, social media platforms, and third-party data sources, resulting in significant data silos and fragmented customer profiles that fail to form a unified view. Furthermore, the generation and updating of customer profiles are often batch-processed or periodic operations, failing to capture real-time changes in customer behavior, leading to outdated profile information and impacting the timeliness of marketing efforts. Tag generation is typically based on fixed rules, neglecting factors such as the reliability of data sources and the recentity of behavior, resulting in unreasonable tag weights and insufficient profile accuracy.

[0004] Therefore, in response to the problems mentioned above, this invention proposes an SCRM customer intelligent profiling system and method based on multi-source data. Summary of the Invention

[0005] To overcome the problems of insufficient data integration, static profiles, and unreasonable tag weights in existing technologies, this invention proposes an SCRM customer intelligent profiling system and method based on multi-source data. By integrating internal and external data, dynamically calculating tag weights, updating profiles in real time, and introducing a feedback loop, the system achieves comprehensive and accurate customer profiles.

[0006] The technical solution of this invention is: an SCRM customer intelligent profiling system based on multi-source data, comprising: The multi-source data acquisition and integration module is used to collect and clean customer data from internal enterprise systems, social platforms and third-party data sources. The internal enterprise systems include CRM and ERP systems, and the social platform data includes customer posts, interactive behaviors and community relationships. The customer dynamic tagging engine connects to the multi-source data acquisition and integration module to perform real-time or near real-time calculations on incoming customer data based on predefined business rules, generate customer tags, and calculate a dynamic weight for each tag. The dynamic weight is determined based on the reliability of the data source and the recentity of the behavior. The customer intelligent profile generation and update module is connected to the customer dynamic tag engine and includes a multi-dimensional profile model, a profile fusion unit, and a profile version manager. The multi-dimensional profile model defines at least four dimensions: basic identity, transaction value, social behavior, and dynamic intent. The profile fusion unit is used to merge weighted tags from the customer dynamic tag engine into the multi-dimensional profile model to generate a unified customer intelligent profile. The profile version manager is used to record the historical versions of the customer profile and track its evolution trend. The interactive profiling application and feedback module connects to the customer intelligent profiling generation and update module, including a visual interface for displaying customer intelligent profiling to users; an insight and early warning center for automatically generating insight or early warning information based on profiling changes; a strategy outreach interface for selecting customers based on profiling and connecting to external marketing tools to execute outreach actions; and a feedback closed-loop channel for receiving user feedback on the accuracy of the profiling and transmitting it to the customer dynamic tagging engine for optimizing tag generation rules. It is worth noting that the multi-source data acquisition and integration module includes: An internal data interface is used to connect to the enterprise's internal systems in batches at predetermined intervals; Social data collectors use authorized API interfaces to collect data from social platforms and can parse text and related information in dynamic content. A third-party data access gateway is used to securely access legitimate third-party data sources, including enterprise registration information and industry opinion reports. The data cleaning and standardization unit is used to deduplicatize, handle outliers, convert formats, and unify customer identification codes for the collected heterogeneous data. The unified customer identification code is preferably achieved by matching mobile phone numbers, email addresses, or open IDs. The customer dynamic tagging engine includes: A tag rule library is used to store predefined tag generation rules based on explicit business logic. These rules allow for logical AND and OR combinations of transaction data and social behavior data. The calculation module is used to perform event-driven scanning and calculation of incremental data streams based on the tag rule base, and dynamically add, modify or remove customer tags; The tag weight calculation unit is used to calculate the dynamic weight of each tag based on a preset weight strategy. The weight strategy stipulates that the base weight of tags derived from transaction data is greater than the base weight of tags derived from social data, and for tags of the same behavior type, their weight values ​​decrease over time according to a preset decay function.

[0007] Preferably, the profile version manager identifies the amount of change in key dimensions by comparing the current version of the customer profile with at least one historical version. When the amount of change exceeds a preset threshold, a visual prompt is displayed in the visualization interface in the form of highlighting, flashing, or a trend chart.

[0008] Preferably, the strategy outreach interface provides a graphical drag-and-drop tool that allows users to define target customer groups based on a combination of one or more profile dimensions and tags, and push the customer group list and predefined marketing content to at least one external marketing tool among WeChat, email, or SMS platforms with one click.

[0009] This invention provides a method for intelligent profiling of SCRM customers based on multi-source data, comprising the following steps: S1 collects customer data from internal enterprise systems, social platforms, and third-party data sources, and performs cleaning and standardization processing. S2, based on predefined business rules, performs real-time or near-real-time calculations on the processed data to generate customer tags, and calculates a dynamic weight for each tag based on the data source and the probabilities of the behavior. The predefined business rules include composite rules that logically combine customer transaction data with social interaction data. For example, the rule for identifying "high-value potential customers" is defined as: customers show strong interest in specific products on social platforms and their affiliated companies match the target customer profile. S3 integrates weighted tags into a predefined multidimensional profile model to generate a smart customer profile and saves the profile version for tracking its evolution. The multidimensional profile model includes at least four dimensions: basic identity, transaction value, social behavior, and dynamic intent. The dynamic intent dimension infers the customer's recent focus and potential needs by analyzing the customer's recent posts and comments on social platforms and semantically matching them with a predefined keyword library. S4 visualizes the generated intelligent customer profile, generates insights or alerts based on profile changes, supports customer outreach actions based on profiles, and collects user feedback to optimize tag generation rules. The collected user feedback includes "confirmation," "correction," or "ignore" operations on system-generated tags. This feedback data is recorded and forms training samples. System administrators or algorithms can manually adjust or semi-automatically optimize the rules in the tag rule base based on these samples, such as modifying rule thresholds or confidence levels, thereby forming a closed-loop learning mechanism to continuously improve the accuracy of profile generation.

[0010] The beneficial effects of this invention are: 1. This invention constructs a multi-source data collection and integration module that integrates internal enterprise systems, social platforms, and third-party data. By using data cleaning and standardized units for unified coding, it achieves comprehensive integration of customer information. The resulting multi-dimensional profile model can present a unified panoramic view of customers from multiple dimensions such as basic identity, transaction value, social behavior, and dynamic intent. This solves the problem of fragmented customer profiles and one-sided perspectives caused by the single and scattered data sources in existing technologies.

[0011] 2. This invention introduces a dynamic customer tagging engine, which uses its calculation module to scan and calculate data. Combined with a weight decay mechanism based on data recency, it achieves real-time updates and dynamic weight adjustments for customer tags. This enables intelligent customer profiles to respond sensitively to the latest changes in customer behavior and intent, ensuring the timeliness and accuracy of the profiles. This overcomes the problems of information lag and inability to capture dynamic customer journeys caused by batch processing and periodic updates in existing technologies.

[0012] 3. This invention distinguishes between tag generation and weight calculation. The tag weight calculation unit assigns dynamic weights to tags based on the reliability of the data source and the recentity of the behavior, which quantifies and distinguishes the importance of different information in the profile, thereby greatly improving the reliability and usability of the profile. At the same time, the system constructs a feedback loop through interactive profile application and feedback modules, which can continuously optimize the tag rule base by using user feedback on tag corrections, forming a system with self-learning and continuous improvement capabilities. This solves the problems of unreasonable tag weights and the difficulty of static rule systems to adapt to business changes in the prior art. Attached Figure Description

[0013] Figure 1 The diagram shown is a schematic representation of the system framework of this invention. Figure 2 The diagram shown is a schematic representation of the method flow of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but 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.

[0015] Please see Figure 1 and Figure 2 This invention provides an embodiment of an SCRM customer intelligent profiling system based on multi-source data: In this embodiment, the multi-source data acquisition and integration module is described in detail: For enterprise CRM and ERP systems, develop dedicated data connectors. These connectors extract key data such as customer tables, order tables, and customer service work order tables through the system's API or by directly reading database logs (authorization required) at a daily incremental synchronization frequency. For example, a synchronization task is started at 2:00 AM every night to synchronize customer data that has been added or changed in the past 24 hours to the system's data center.

[0016] For WeChat Work, you need to apply for WeChat Work API permissions to use its interface list to obtain the list of customers added by enterprise members, and then use the interface to retrieve the Moments posts, comments and likes posted by customers in near real-time (polling every 5 minutes). For public social media, you can collect public data according to your permissions through its official API (developer qualifications are required).

[0017] For third-party data, we cooperate with legitimate and compliant commercial data service providers (such as Tianyancha and Data Treasure) and use their encrypted API interfaces to input the client's company name or unified social credit code to query and supplement the client's business registration information, operating status, and equity structure in batches.

[0018] In this embodiment, the collected data is cleaned and standardized. For cases where the same customer's data is collected multiple times from different sources, priorities are set (e.g., CRM data takes precedence over ERP data). The highest priority complete record is retained, and discrepancies in other records are merged. Outlier handling is then performed, with business rules automatically identifying anomalies. For example, if a customer's "annual transaction amount" field value is identified as negative or significantly exceeding the industry average, the record is marked as "suspicious" and temporarily withheld from the database, awaiting manual review. Finally, all dates and times in the data are unified, and all currency amounts are converted to RMB.

[0019] A customer master data model is established, using mobile phone numbers as the primary matching key due to their high prevalence and stability in business activities. Corporate email addresses are used as a secondary matching key. For customers connected through open platforms (such as WeChat), their UnionID is used as a unique identifier. The system automatically performs a fuzzy matching algorithm; for example, numbers "138-0013-8000" and "13800138000" are identified as the same customer after standardization. Finally, a globally unique customer ID (GUID) is generated for each unique customer entity, and all subsequent data is associated with this GUID.

[0020] This embodiment describes the customer dynamic tagging engine in detail: (1) First, a tag rule base is built. The rule base is predefined using the "condition-action" logic. All rules are configured and maintained by business experts (such as sales directors and marketing managers) through the management backend interface. Some examples of the rules configured during implementation are as follows: Rule 1 (High-Value Customers): If the cumulative transaction amount in the past 12 months is >= 100,000 yuan AND there has been any interaction in the past 30 days, then the customer is tagged as a "High-Value Customer".

[0021] Rule 2 (Product A Interesters): If (mention of the keyword "Product A" in WeChat Moments or likes / comments containing "Product A" in the past 7 days) >= 2 times, then label them as "Product A Interesters".

[0022] Rule 3 (Serving High-Risk Customers): If the number of customer service tickets submitted in the past 15 days with "Complaint" as the "Ticket Type" is >= 1, then label the customer as "Serving High-Risk Customer".

[0023] (2) The calculation module starts working. This module continuously listens to the message queue from the data acquisition module. Once a new data event (such as "add a transaction record" or "customer posted a new Moments post") arrives, the engine immediately loads all rules related to the customer's GUID, checks whether the new data triggers any rule conditions, and if so, immediately executes the "tag" or "detag" action and writes the result to the database. For example, when a customer completes a large payment, the condition of rule 1 is immediately met, and the customer will be dynamically tagged as a "high-value customer" within minutes.

[0024] (3) The tag weight calculation unit designs a dynamic weight value W for each tag, and its calculation formula is as follows: ; Wherein, S (source credibility coefficient) is a preset static coefficient, set according to the objective credibility of the data source on which the tag is generated. In this invention, it is defined as: S=1.0 (transaction data), S=0.7 (customer service data), S=0.5 (social interaction data), S=0.8 (third-party authoritative enterprise data).

[0025] F(t) (the recency decay function) is a function that decays over time to reflect the timeliness of information. This invention uses an exponential decay model: ; Where t is the time elapsed since the tag was generated or last confirmed (in days). It is the decay rate constant, which can be adjusted according to business needs, such as for the fast-moving consumer goods industry. A value of 0.1 is acceptable for the heavy equipment industry. 0.01 is acceptable.

[0026] For example, a customer is labeled a "high-value customer" based on their transaction activity on a given day, and their initial weight... If no new related behaviors reinforce this label after 30 days, its weight will decrease. At this point, the importance of this tag in the profile decreases significantly; conversely, if another transaction occurs on the 29th day, the tag's t will be reset, and the weight will return to around 1.0.

[0027] In this embodiment, the customer intelligent profile generation and update module is described in detail: (1) The profile fusion device reads all currently valid tags and their weights of a customer's GUID periodically or triggered by a tag update event, and classifies these tags into four dimensions: basic identity, transaction value, social behavior and dynamic intent according to the mapping relationship table.

[0028] For the dynamic intent dimension, this invention implements an inference method based on a keyword library and simple statistics. The system maintains a business keyword library, including product names, competitor names, and industry terms (such as expansion, budget, bidding, comparison of XX products, etc.). When a customer's social media content is collected, it is segmented and matched with the keyword library. Successfully matched keywords are stored as intent signals in this dimension. Their initial weight is determined by the frequency of the keyword in recent content and the S-coefficient. For example, if a customer mentions competitor B twice in three WeChat Moments posts within a week, a signal of paying attention to competitor B will be generated in the dynamic intent dimension, and it will have a high weight.

[0029] (2) The profile version manager automatically creates a version for each customer profile. Whenever a profile undergoes a significant change (such as a change in the total weight of any dimension exceeding 10%), the system saves a snapshot of the current profile and adds a timestamp. In the visualization interface, business personnel can select a customer and view its profile's "historical evolution view". This view clearly displays the weight change curves of each core tag in a timeline format. For example, one can intuitively see the complete evolution path of a customer from a potential customer to a high-value customer, and then to a recently serviced high-risk customer.

[0030] In this embodiment, the interactive portrait application and feedback module will be described in detail: The visual interface displays a visual profile card for each customer. The top of the card shows the customer's basic information, and the bottom shows a radar chart or bar chart for four dimensions, intuitively displaying the strengths and weaknesses of the customer in each aspect. High-weighted tags and tags that have recently changed are highlighted with special colors (such as red).

[0031] The system's built-in early warning rules continuously scan for changes in all customer profiles. For example, a rule could be defined as: "IF 'Service Risk' category tag weights for 'High-Value Customers' increase by more than 0.5 THEN Generate a Level 1 Early Warning." Early warning information will be instantly pushed to the relevant customer management personnel via system messages, SMS, or WeChat notifications.

[0032] In the dashboard, sales staff can quickly identify a target customer group by selecting dimensions and tags (such as "High-Value Customers," "Product A Interested Individuals," and "Dynamic Intent" containing the keyword "Budget"). They can then write a targeted product description and click the "Push to WeChat Work" button. The system will then use the WeChat Work API to distribute the customer list and sales script template to the relevant sales staff's WeChat Work sidebar with a single click, guiding them to make precise outreach.

[0033] The feedback loop has three buttons next to the label of each customer profile: Confirm, Correct, and Ignore. When a salesperson finds that the system's label of "Product A Interested Person" for a customer is inaccurate (for example, after communication, it is found that the customer is actually interested in Product B), they can click the "Correct" button and select the correct label "Product B Interested Person" from the drop-down menu. This "correction" action will be recorded by the system, forming a feedback record: (Customer GUID, Original Label, Corrected Label, Operator, Timestamp).

[0034] System administrators can review these feedback records weekly and manually optimize the rule base accordingly. For example, if multiple people are found to have changed "Product A enthusiast" to "Product B enthusiast," and these customers have all mentioned a common keyword "mobile office," then the administrator can add or modify a rule: IF if the content contains "mobile office" THEN tag "Product B enthusiast," while lowering the confidence level of the original rule.

[0035] This invention provides a comparative example: This comparative study used the sales team of a medium-sized enterprise for a three-month comparative test. The comparative team used a traditional, static rule-based CRM system, while the experimental team used the present invention.

[0036]

[0037] As shown in the table above, the delays of more than 24 hours in the comparative model are due to reliance on manual data entry or daily batch data processing. For example, after a salesperson visits a customer, the information may not be updated in the CRM until the next day. The system's batch calculations are usually performed at night. In this invention, social interaction data is collected every 5 minutes, and transaction data is transmitted in real time through the system interface. Once the data enters the system, the customer dynamic tagging engine processes it immediately without waiting for the batch job window.

[0038] As shown in the table above, relying solely on transaction history to identify clues may lead to missing out on a large number of highly potential customers who have not yet made a purchase. This invention identifies "silent" customers (such as "product A enthusiasts" identified by Rule 2) who actively express interest and display buying signals through social behavior and dynamic intent dimensions. By updating information in a timely manner, sales personnel can reach out to customers when their interest is at its peak, thereby significantly improving the success rate of communication.

[0039] As shown in the table above, customer churn is a gradual process, not a sudden event. A warning signal of less than one week often indicates that the customer has already made up their mind to leave, and it's too late to try and win them back. A lead time of 3-4 weeks, however, provides ample time.

[0040] As shown in the table above, traditional customer screening relies on sales staff manually filtering through lists of thousands of customers, which is time-consuming and labor-intensive. The strategic outreach interface of this invention allows sales staff to quickly identify target groups (such as "all 'high-value customers' who are interested in 'Product A' and whose 'industry' is finance") through a graphical interface using multi-dimensional tags, reducing screening time from hours to minutes.

[0041] As can be seen from the table above, the dynamic weighting mechanism of this invention automatically reduces the interference of outdated and unreliable information. At the same time, the fusion of multi-source data plays a role in cross-validation (for example, a customer may be an ordinary customer in the CRM, but an industry opinion leader on social media, and the system can more comprehensively portray its value).

[0042] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An SCRM customer intelligent profiling system based on multi-source data, characterized in that, Including: The multi-source data acquisition and integration module is used to collect and clean customer data from internal enterprise systems, social platforms and third-party data sources. Internal enterprise systems include CRM and ERP systems, and social platform data includes customer posts, interactive behaviors and community relationships. The customer dynamic tagging engine connects to the multi-source data acquisition and integration module to perform real-time or near real-time calculations on incoming customer data based on predefined business rules, generate customer tags, and calculate a dynamic weight for each tag. The dynamic weight is determined based on the reliability of the data source and the recentity of the behavior. The customer intelligent profile generation and update module is connected to the customer dynamic tag engine and includes a multi-dimensional profile model, a profile fusion unit, and a profile version manager. The multi-dimensional profile model includes at least four dimensions: basic identity, transaction value, social behavior, and dynamic intent. The profile fusion unit is used to merge weighted tags from the customer dynamic tag engine into the multi-dimensional profile model to generate a unified customer intelligent profile. The profile version manager is used to record the historical versions of the customer profile and track its evolution trend. The interactive profiling application and feedback module connects with the customer intelligent profiling generation and update module. It is used to display customer intelligent profiling to users, automatically generate insights or early warning information based on profiling changes, and select customers based on the profiling and connect to external marketing tools to execute outreach actions.

2. The SCRM customer intelligent profiling system based on multi-source data according to claim 1, characterized in that, The customer dynamic tagging engine further includes: A tag rule library is used to store tag generation rules predefined based on explicit business logic; The calculation module is used to scan and calculate the incremental data stream based on the tag rule base and dynamically update customer tags; The tag weight calculation unit is used to calculate the dynamic weight of each tag based on a preset weight strategy.

3. The SCRM customer intelligent profiling system based on multi-source data according to claim 1, characterized in that: The interactive profiling application and feedback module also includes a feedback closed-loop channel, which is used to receive user feedback on the accuracy of the profile and transmit it to the customer dynamic tag engine for optimizing tag generation rules.

4. The SCRM customer intelligent profiling system based on multi-source data according to claim 1, characterized in that: The customer intent in the dynamic intent dimension is a potential demand signal inferred by analyzing the customer's recent behavior on social platforms and matching it with a predefined keyword library.

5. The SCRM customer intelligent profiling system based on multi-source data according to claim 1, characterized in that: The profile version manager identifies and highlights dimensions that have changed significantly in the interactive profile application and feedback module by comparing the current version of the customer profile with at least one historical version.

6. The SCRM customer intelligent profiling system based on multi-source data according to claim 1, characterized in that: The interactive profiling application and feedback module allows users to define target customer groups based on one or more profiling dimensions and tag combinations, and push the customer group list and predefined content to at least one of the following platforms: WeChat, email, or SMS.

7. A method for intelligent customer profiling in SCRM based on multi-source data, employing the intelligent customer profiling system for SCRM based on multi-source data as described in any one of claims 1-6, characterized in that, It includes the following steps: S1 collects customer data from internal enterprise systems, social platforms, and third-party data sources, and performs cleaning and standardization processing. S2, based on predefined business rules, performs real-time or near-real-time calculations on the processed data, generates customer tags, and calculates a dynamic weight for each tag based on data source and behavioral relevance; S3 integrates weighted labels into a predefined multidimensional profile model to generate a smart customer profile and saves profile versions for tracking their evolution. S4 visualizes the generated intelligent customer profiles, generates insights or alerts based on profile changes, supports customer outreach actions based on profiles, and collects user feedback to optimize tag generation rules.

8. The SCRM customer intelligent profiling method based on multi-source data according to claim 7, characterized in that: In step S2, the predefined business rules include rules that logically combine transaction data with social interaction data to identify customer groups with specific potential.

9. The SCRM customer intelligent profiling method based on multi-source data according to claim 7, characterized in that: In step S3, the multi-dimensional profile model includes at least four dimensions: basic identity, transaction value, social behavior, and dynamic intent. The dynamic intent dimension is filled by analyzing the matching degree between the customer's recent content on the social platform and predefined keywords.

10. The SCRM customer intelligent profiling method based on multi-source data according to claim 7, characterized in that: In step S4, the collected user feedback includes confirmation, correction, or ignoring of system-generated tags. This feedback is used to manually adjust or semi-automatically optimize predefined business rules.

Images