Customer revisit priority analysis method, device, equipment and medium

Abstract

The application relates to the technical field of data analysis, and discloses a customer follow-up priority analysis method, device, equipment and medium, which comprises the following steps: collecting online consultation information, offline interaction information and online behavior data of a target user, constructing behavior intention features of the user based on the online behavior data, analyzing quantified features of the user in multiple dimensions such as transaction urgency, asset scale and decision maturity from online consultation texts, and obtaining transaction intention feature values of the user by weighted fusion, generating transaction demand labels of the user according to the online consultation information, and generating customer group type labels according to the offline interaction information. The transaction demand labels and the transaction intention feature values are combined to calculate follow-up service priority scores of each user, and the customer group type labels are used together to generate a follow-up task queue and push the follow-up task queue to a client. The application can be applied to a financial technology business system platform. The application improves the efficiency and accuracy of customer follow-up priority analysis.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, and in particular to a method, apparatus, equipment and medium for customer follow-up priority analysis. Background Technology

[0002] In recent years, wealth management needs have become more diversified and personalized. The traditional operating model that relies on financial advisors to answer inquiries, judge intentions based on experience, and manually enter data into the CRM is no longer suitable for business development.

[0003] In the fintech field, most common lead management systems are general-purpose CRMs or simple automatic allocation rules, lacking specialized solutions tailored to the characteristics of financial wealth management, such as "long decision-making cycles, high capital thresholds, and highly segmented needs." Specifically, existing technologies have the following shortcomings: First, there is a lack of quantitative standards for judging intent. Advisors' assessments of clients' financial strength, urgency of needs, and decision maturity are often based on subjective experience, leading to the misjudgment or omission of high-value leads. Second, the timing of lead follow-up is difficult to capture automatically. Strong intent signals generated by clients during "window periods" such as fund maturity and year-end dividends cannot be identified in real time, causing delays in follow-up. Third, there is a disconnect between front-end marketing touchpoints (official website, app, offline events) and back-end CRM systems, requiring manual sorting and entry of leads, which is inefficient and prone to errors. Fourth, the high segmentation of wealth management needs, such as family trusts, overseas allocation, and retirement planning, makes it impossible to accurately label customer groups, resulting in a lack of targeted initial outreach. The development of fintech offers a technological solution to the aforementioned problems: Natural Language Processing (NLP) technology automatically extracts keywords related to assets, needs, and decision-making from consultation texts; user behavior analysis constructs online behavioral intention characteristics; multi-dimensional features are combined and weighted to generate a quantitative intention score; APIs enable real-time synchronization with CRM; and sales queues are automatically adjusted and follow-up tasks are generated based on business rules. Therefore, there is an urgent need for an intelligent, high-intent lead dynamic tagging and CRM automation integration system and method for financial wealth management to improve lead conversion efficiency and the productivity of financial advisors. Summary of the Invention

[0004] This invention provides a customer follow-up priority analysis method, apparatus, computer equipment, and medium to solve the problems of low efficiency and poor accuracy of existing customer follow-up priority analysis methods on the market.

[0005] Firstly, a method for prioritizing customer follow-up visits is provided, including: Acquire online consultation information, offline activity interaction information, and online behavior information of each target user; Based on the online behavior information, construct the behavioral intention characteristics of each target user; Identify the preset key fields in the online consultation information to obtain the key field set for each target user, wherein the preset key fields include asset key fields, demand key fields, and decision key fields; Based on the set of key fields, the data characteristics of each target user in multiple preset dimensions are analyzed. The data characteristics include the data characteristics of the target user in the dimensions of transaction demand, asset distribution and decision-making ability. The data characteristics of each target user in different data dimensions are weighted and summed to obtain the transaction intention feature value of each target user. Based on the online consultation information, transaction demand tags are constructed for each target user; based on the offline activity interaction information, customer group type tags are constructed for each target user. Analyze the priority score of the return visit service for each target user based on the transaction demand tags and transaction intention feature values ​​of each target user; A return visit task queue is generated based on the return visit service priority score and the customer group type label, and the return visit task queue is pushed to a preset client.

[0006] Secondly, a customer follow-up priority analysis device is provided, including: The data acquisition module is used to acquire online consultation information, offline activity interaction information, and online behavior information of each target user; The feature construction module is used to construct behavioral intention features of each target user based on the online behavior information, identify preset key fields in the online consultation information, and obtain a set of key fields for each target user. The preset key fields include asset key fields, demand key fields, and decision key fields. The intention analysis module is used to analyze the data characteristics of each target user in multiple preset dimensions based on the set of key fields. The data characteristics include the data characteristics of the target user in the dimensions of transaction demand, asset distribution and decision-making ability. The data characteristics of each target user in different data dimensions are weighted and summed to obtain the transaction intention feature value of each target user. The tag building module is used to build transaction demand tags for each target user based on the online consultation information, and to build customer group type tags for each target user based on the offline activity interaction information. The task construction module is used to analyze the return visit service priority score of each target user based on the transaction demand tags and transaction intention feature values ​​of each target user, generate a return visit task queue based on the return visit service priority score and the customer group type tag, and push the return visit task queue to a preset client.

[0007] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described customer return visit priority analysis method.

[0008] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the aforementioned customer return visit priority analysis method.

[0009] The aforementioned solution, implemented using customer follow-up priority analysis methods, devices, computer equipment, and storage media, comprehensively covers customer interaction data by collecting online consultation content, offline activity participation records, and online behavioral trajectories from multiple channels. This avoids profile bias caused by a single data source, laying the foundation for subsequent accurate assessment. Based on this online behavioral data, behavioral intention features reflecting the user's level of proactive attention are constructed, thereby transforming user behavior into objective values. This overcomes the ambiguity of traditional reliance on consultants' subjective observation, achieving automatic capture and quantitative expression of behavioral signals. Simultaneously, three key fields—assets, needs, and decisions—are identified from online consultation texts, and quantitative characteristics of users across multiple dimensions, such as transaction urgency, asset size, and decision maturity, are analyzed. This process transforms unstructured text into structured information, significantly reducing manual reading and annotation workload, ensuring comprehensiveness and objectivity in the assessment dimensions. Weighted fusion yields the user's transaction intention feature value, making the judgment of customer transaction intentions more scientific and consistent, addressing the problem of inconsistent experience-based judgment standards. Furthermore, transaction demand tags are generated based on online consultation information, and customer group type tags are generated based on offline activity interaction information. This forms a dual online and offline tagging system, which can precisely characterize customer needs and customer group categories, facilitating the adoption of differentiated service strategies and significantly improving marketing accuracy. Subsequently, by combining transaction demand tags and transaction intention feature values, a follow-up service priority score is calculated for each user. This ensures that high-intent, high-urgency, and high-value customers are followed up immediately, preventing the loss of valuable leads due to delays, while also improving the time and resource utilization efficiency of financial advisors. Finally, follow-up task queues are generated by prioritizing customers and matching personalized task prompts based on customer group type tags, and pushed to the financial advisor's client, improving the efficiency and accuracy of customer follow-up priority analysis. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of an application environment for a customer follow-up priority analysis method according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating a customer follow-up priority analysis method according to an embodiment of the present invention; Figure 3 This is a schematic diagram of a customer follow-up priority analysis device in one embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention; Figure 5 This is another structural schematic diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0012] 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 only some, not all, of the embodiments of the present invention. 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.

[0013] The customer follow-up priority analysis method provided in this invention can be applied to, for example... Figure 1In this application environment, the client communicates with the server via a network. The server can collect online consultation content, offline activity participation records, and online behavioral patterns of target users from multiple channels through the client. Based on this online behavioral data, behavioral intention features reflecting the user's level of proactive attention are constructed. Simultaneously, three key fields—assets, needs, and decisions—are identified from the online consultation text, and quantitative characteristics of users across multiple dimensions, such as transaction urgency, asset size, and decision maturity, are analyzed. These are then weighted and fused to obtain the user's transaction intention feature value. Furthermore, transaction demand tags are generated based on online consultation information, and customer group type tags are generated based on offline activity interaction information. Subsequently, combining the transaction demand tags and transaction intention feature values, a follow-up service priority score for each user is calculated. This score, along with the customer group type tag, is used to generate a follow-up task queue—sorted by priority and accompanied by personalized task prompts for different customer groups. Finally, the task queue is pushed to the financial advisor's client, improving the efficiency and accuracy of customer follow-up priority analysis. The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster consisting of multiple servers. The invention will be described in detail below through specific embodiments.

[0014] Please see Figure 2 As shown, Figure 2 A flowchart illustrating the customer follow-up priority analysis method provided in this embodiment of the invention includes the following steps: S1. Obtain online consultation information, offline activity interaction information, and online behavior information of each target user.

[0015] In this embodiment of the invention, the acquisition of online consultation information, offline activity interaction information and online behavior information of each target user involves comprehensively collecting the original interaction records of customers from multiple preset business channels, providing basic data for subsequent intention assessment, feature extraction and tag construction.

[0016] In the fintech field, online consultation information refers to text-based inquiries submitted by customers through channels such as official website messages, app online chat, and WeChat official account messages; offline activity interaction information refers to registration forms, contact information, or on-site communication records filled out by customers at offline lectures, exhibitions, and other scenarios; and online behavioral information refers to traceable behaviors generated by customers on digital channels such as official websites and apps, such as visiting product detail pages, downloading product manuals, using yield calculators, and scheduling face-to-face meetings. By aggregating these three types of information, a relatively complete picture of the interaction between customers and wealth management institutions can be drawn, laying a data foundation for subsequent multi-dimensional assessments.

[0017] In this embodiment of the invention, by simultaneously collecting online consultation texts, offline activity data, and online behavior logs, it is possible to comprehensively cover customer interaction traces across multiple channels, avoiding customer profiling bias caused by a single data source, and providing a rich, multi-dimensional, and reliable data foundation for subsequent evaluation.

[0018] S2. Construct behavioral intention characteristics for each target user based on the online behavior information.

[0019] In this embodiment of the invention, constructing behavioral intention features for each target user based on the online behavioral information includes: Obtain the user behavior events contained in the online behavior information to form a behavior event sequence; Analyze the event type of each behavioral event in the behavioral event sequence; Count the number of event sequences corresponding to different event types in the behavioral event sequence; The behavior score for each event type is obtained by multiplying the number of event sequences corresponding to each event type by the preset base weight for each event type. The behavioral scores for each event type are normalized to obtain a normalized behavioral feature value table. Generate behavioral intention features for each target user based on the normalized behavioral feature value table.

[0020] In detail, the process of acquiring user behavior events contained in the online behavior information and forming a behavior event sequence is achieved by embedding front-end tracking code or SDKs at front-end touchpoints such as official websites, apps, and WeChat official accounts. This automatically captures every interactive operation generated by each user while logged in, including page visits, button clicks, file downloads, form submissions, etc. The system groups the captured raw logs by user ID and arranges them in ascending order according to the event occurrence time, forming a strict time-series behavior event sequence based on users. Each event record includes fields such as event type, occurrence timestamp, target object identifier, and session ID.

[0021] In detail, the analysis of the event type of each behavioral event in the behavioral event sequence involves automatically classifying each event record in the behavioral event sequence according to a preset event type classification rule base. This rule base uses event name or URL path pattern as the matching basis to divide user behavior into several financial wealth management related types, such as "accessing product details page", "downloading product manual", "using the yield calculator", "application for appointment", "repeatedly viewing the same product", etc.

[0022] In detail, the step of counting the number of event sequences corresponding to different event types in the behavioral event sequence is to group and count the behavioral event sequences of each user according to the event type after the event type labeling is completed.

[0023] In detail, the step of multiplying the number of event sequences corresponding to each event type by the preset basic weight of each event type to obtain the behavior score for each event type is pre-assigned by business experts based on the contribution of each behavior to the intensity of intention (for example, "schedule a meeting" has a weight of 50 points, "download the instruction manual" has a weight of 20 points, and "visit the product page" has a weight of 5 points). The frequency of each event type for each user is obtained by multiplying it by the corresponding basic weight to obtain the user's original score for that event type.

[0024] In detail, the normalization of behavioral scores for each event type to obtain a normalized behavioral feature value table is to eliminate the impact of differences in the total amount of behavior among different users and inconsistencies in the weighting of different event types. The minimum-maximum normalization method is adopted: for each event type, the maximum and minimum values ​​of the original scores of all users in that type are first calculated, and then each user's original score is mapped to the interval of 0 to 100 according to the formula (original score - minimum value) / (maximum value - minimum value) × 100; if all users' scores in that type are 0, the normalized score is uniformly set to 0.

[0025] In detail, the step of generating behavioral intention features for each target user based on the normalized behavioral feature value table involves performing a secondary weighted summation of the normalized scores for each user across multiple event types. The secondary weights are also set by business experts based on the predictive power of each behavioral type for the final intention (e.g., "Schedule an appointment" weight 40%, "Download instruction manual" weight 30%, "Use calculator" weight 15%, "Visit late at night" weight 10%, "Visit product page" weight 5%). The system multiplies each user's normalized score by its corresponding weight and sums them up to obtain a single value between 0 and 100 as the user's "behavioral intention feature".

[0026] In this embodiment of the invention, by converting users' online behavior into quantifiable feature values, the degree of customers' active attention and interest in financial products is objectively reflected, overcoming the ambiguity of traditional reliance on advisors' subjective observation, and realizing the automatic capture and numerical expression of behavioral signals.

[0027] S3. Identify the preset key fields in the online consultation information to obtain the key field set for each target user, wherein the preset key fields include asset key fields, demand key fields, and decision key fields.

[0028] In this embodiment of the invention, the step of identifying preset key fields in the online consultation information to obtain a set of key fields for each target user includes: The online consultation texts are classified according to different preset channel types to obtain the original consultation text set. The original consultation text set is segmented to obtain a word sequence set; Each word in the word sequence set is tagged with its part of speech to generate a word sequence list after word segmentation; The segmented word sequence list is matched with a preset asset key field lexicon to obtain an asset key field set; The word segmentation sequence list is matched with the preset key field lexicon to obtain the key field set; The word segmentation sequence list is matched with a preset decision key field lexicon to obtain a decision key field set; The key field set is obtained by summarizing the set of key asset fields, the set of key demand fields, and the set of key decision fields.

[0029] In detail, the process of classifying the online consultation texts based on different preset channel types to obtain the original consultation text set involves reading all original online consultation records from the multi-channel access module. Each record is accompanied by a source channel identifier (such as official website messages, APP online consultations, WeChat official account messages, online customer service conversations, etc.). The system groups the records according to the preset channel type list, classifying consultation texts from the same channel into one category.

[0030] In detail, the step of segmenting the original consultation text set to obtain a word sequence set involves calling a Chinese word segmentation tool (such as a dictionary-based maximum matching algorithm or a statistical hidden Markov model word segmenter) to divide the continuous Chinese character sequence into independent word units for each text content in the original consultation text set.

[0031] In detail, the step of performing part-of-speech tagging on each word in the word sequence set to generate a word sequence list after word segmentation is to tag each word with its grammatical part of speech (such as noun, verb, numeral, adjective, etc.) based on the word segmentation results. This tagging can be performed using a part-of-speech tagging model based on conditional random fields or bidirectional LSTM.

[0032] In detail, the step of matching the segmented word sequence list with a pre-defined asset key field lexicon to obtain an asset key field set involves pre-constructing a lexicon covering asset-related keywords in the financial wealth management field. This lexicon includes explicit monetary expressions (such as "ten thousand", "hundred million", "yuan", "deposit", "real estate", "corporate dividends", "trust maturity", etc.) as well as asset size implications (such as "high net worth", "tens of millions", etc.). The system iterates through each user's word sequence, performing a complete string match or a synonym expansion match between each word and the asset key field lexicon. If a match is successful, the word and its contextual numerical information are extracted together. For example, if "8 million" or "tens of millions" is matched, it is added to the asset key field set, and the position and frequency of the word in the original text are recorded.

[0033] In detail, the step of matching the segmented word sequence list with a pre-built demand key field thesaurus to obtain a demand key field set uses a pre-constructed demand key field thesaurus. This thesaurus focuses on keywords that reflect the urgency and specific theme of customer needs, including time-sensitive words (such as "next week," "immediately," "due," "urgently needed," "window period"), financial goal words (such as "children's education," "retirement," "study abroad," "trust alternative"), and behavioral need words (such as "recommended product," "need a solution," "schedule a meeting"). The system matches the word sequence of each user, and the successfully matched words are included in the demand key field set, while recording the urgency level or demand theme category, providing raw materials for subsequent transaction demand dimension analysis.

[0034] In detail, the step of matching the word sequence list after word segmentation with a preset decision key field thesaurus to obtain a set of decision key fields is to use a pre-built decision key field thesaurus. This thesaurus contains professional terms that reflect customers’ in-depth comparison and rational judgment of financial products, such as “terms”, “fees”, “historical performance”, “risk control measures”, “contract sample”, “minimum investment threshold”, “exit mechanism”, “subscription fee”, “management fee”, “rating”, and “underlying assets”.

[0035] In detail, the process of summarizing the set of key fields for assets, the set of key fields for requirements, and the set of key fields for decisions to obtain the set of key fields involves merging and deduplicating the three sets obtained by the same user in the above three matching steps, and organizing them in a structured manner according to field categories (assets / requirements / decisions); during the merging process, the original occurrence frequency and position information of each field are retained for subsequent weighted or priority sorting.

[0036] In this embodiment of the invention, by extracting key fields such as assets, needs, and decisions from unstructured consultation texts, the original natural language is transformed into structured information, which greatly reduces the workload of manual reading and annotation, and provides accurate and standardized input for subsequent dimensional analysis.

[0037] S4. Analyze the data characteristics of each target user in multiple preset dimensions based on the set of key fields. The data characteristics include the data characteristics of the target user in the dimensions of transaction demand, asset distribution and decision-making ability. The data characteristics of each target user in different data dimensions are weighted and summed to obtain the transaction intention feature value of each target user.

[0038] In this embodiment of the invention, the step of analyzing the data characteristics of each target user across multiple preset dimensions based on the set of key fields includes: Extract the key fields related to the urgency of the requirement from the set of key requirement fields contained in the set of key requirement fields to obtain the urgency of the requirement field; Based on preset mapping rules, the urgency field of the demand is mapped to the characteristic value of the transaction demand; Extract the key fields related to asset size from the asset key field set contained in the key field set to obtain the asset size field; Extract the numerical features of the asset size field to obtain the asset distribution feature value; The percentage of preset professional fields appearing in the decision-making key field set included in the key field set is statistically analyzed to generate decision-making capability feature values; By summarizing the transaction demand characteristic values, the asset distribution characteristic values, and the decision-making ability characteristic values, data characteristics of each target user in multiple preset dimensions are obtained.

[0039] In detail, the step of extracting key fields related to the urgency of the need from the set of key fields containing the key field set to obtain the urgency field is to take a subset of "key fields of need" from the key field set generated in the previous step, perform semantic classification on each field in the subset, and identify keywords that clearly express the urgency of the time or the degree of urgency of the need, such as "urgently needed", "immediately", "right now", "due next week", "funds will arrive soon", "window period", "child will go abroad soon", etc.; pattern matching is performed through a predefined list of urgency keywords, or specific time expressions can be extracted by combining regular expressions (such as "within 3 days" or "next week").

[0040] In detail, the process of mapping the urgency field of demand to transaction demand feature values ​​based on preset mapping rules involves converting the urgency field obtained in the previous step into a value between 0 and 100 according to preset grading rules. For example, the mapping rules assign a base score of 60 points directly to words with strong urgency such as "urgently needed / immediately / right away". When a specific time limit is present, the score is based on the number of days remaining (40 points for 7 days, 20 points for 30 days, and 5 points for more than 30 days). When a window signal such as "funds due / deposited" is present, 30 points are added. The above scores are accumulated for each user, but the upper limit is no more than 100 points.

[0041] In detail, the step of extracting key fields related to asset size from the asset key field set included in the key field set to obtain the asset size field involves taking a subset of "asset key fields" from the key field set and filtering out keywords and numerical expressions that can directly or indirectly reflect the scale of a client's investable assets. These include explicit monetary figures (such as "8 million" or "120 million"), monetary unit words ("ten thousand", "hundred million", "yuan"), and qualitative descriptive words ("high net worth", "tens of millions", "business owner", "multiple properties", etc.). The combination of numbers and units is extracted using regular expressions, and qualitative keywords are recorded to form the asset size field set, providing a basis for subsequent numerical extraction.

[0042] In detail, the extraction of numerical features from the asset size field to obtain asset distribution feature values ​​involves standardizing the numerical information in the asset size field: First, different units (ten thousand, one hundred million) are uniformly converted to "ten thousand yuan", for example, "8 million" is converted to 800, and "120 million" is converted to 12,000; if the user does not have a specific number but has a qualitative description (such as "high net worth"), a reference value is assigned according to a preset mapping table (extremely high → 100 million yuan, high → 30 million yuan, medium → 5 million yuan, low → 500,000 yuan); then, the amount value is mapped to a feature value of 0 to 100 through a piecewise function, for example, feature value = min(100, amount / 100), where the upper limit of the amount is capped at 100 million yuan; for users without any asset signals, the feature value is set to 0; finally, the asset distribution feature value representing the potential asset size level of the customer is output.

[0043] In detail, the step of statistically analyzing the percentage of preset professional fields appearing in the decision-making key field set included in the key field set and generating decision-making capability feature values ​​involves extracting a subset of "decision-making key fields" from the key field set, counting the number of fields belonging to the preset "professional decision-making terminology library," and simultaneously recording the total length (or total number of effective words) of the user's consultation text. The preset professional decision-making terminology library includes terms such as "terms," ​​"fees," "historical performance," "risk control measures," "contract samples," "investment threshold," "exit mechanism," "subscription fee," "management fee," "rating," "underlying assets," "leverage ratio," and "liquidity."

[0044] In detail, the process of summarizing the transaction demand feature value, the asset distribution feature value, and the decision-making ability feature value to obtain the data features of each target user across multiple preset dimensions involves associating and integrating the three feature values ​​(transaction demand feature value, asset distribution feature value, and decision-making ability feature value) calculated above by user ID to form a data record containing values ​​across the three dimensions. This record can be stored in JSON or database row format, for example, {User ID: U1001, Transaction demand feature value: 85, Asset distribution feature value: 72, Decision-making ability feature value: 63}. This summary result will serve as the direct input for the subsequent "weighted summation of the data features of each target user across different data dimensions" to ultimately calculate the transaction intention feature value.

[0045] In this embodiment of the invention, by comprehensively evaluating users’ performance in multiple dimensions such as asset strength, urgency of needs, and decision-making ability, and weighting and integrating them to form a unified intention score, the judgment of customers’ transaction intentions becomes more objective and quantitative, completely changing the problem of strong subjectivity and inconsistent standards in traditional judgments based on experience.

[0046] S5. Construct transaction demand tags for each target user based on the online consultation information, and construct customer group type tags for each target user based on the offline activity interaction information.

[0047] In this embodiment of the invention, constructing transaction demand tags for each target user based on the online consultation information includes: The online consultation information is segmented into words to obtain a sequence of consultation terms; The sequence of consultation terms is matched with a pre-set keyword library for different demand types to obtain demand type matching results; Based on the matching results of the aforementioned demand types, the number of times each target user's consultation word sequence was matched in the keyword database for different demand types was calculated. The demand type corresponding to the keyword library with the most hits is identified as the main demand type of each target user. The keyword library is mapped to transaction demand tags based on preset tag mapping rules.

[0048] In detail, the step of performing word segmentation on the consultation text in the online consultation information to obtain a consultation word sequence is specifically as follows: after merging the original customer online consultation texts collected from multiple channels (including official website messages, APP online conversations, WeChat official account messages, etc.) by user ID, a Chinese word segmentation tool (such as a bidirectional LSTM-based neural network word segmenter or jieba word segmentation loaded with a custom dictionary for the financial wealth management field) is uniformly called for segmentation; during the word segmentation process, punctuation marks, meaningless stop words (such as "de", "le", "a", etc.), numbers, English words and other irrelevant characters are removed, continuous sentences are disassembled into independent word lists, and each user corresponds to one word sequence.

[0049] In detail, the step of matching the consultation word sequence with preset keyword libraries of different demand types to obtain demand type matching results is specifically as follows: a plurality of keyword libraries corresponding to financial wealth management demand types are constructed in advance, each keyword library corresponds to one demand type, for example, the "conservative wealth management" keyword library includes "conservative", "capital preservation", "fixed income", "trust", "fixed income-like" and other words, the "children's education fund planning" keyword library includes "children's education", "overseas study", "children's tuition", "education fund" and other words, and the "family trust demand" keyword library includes "family trust", "inheritance", "estate" and other words; the system performs string matching (supporting synonym expansion and fuzzy matching) between the consultation word sequence of each user and the keyword library of each demand type respectively, and records the specific keywords matched and the matching positions in the keyword library of each demand type for each user.

[0050] In detail, the step of counting the hit times of the consultation word sequence of each target user in keyword libraries of different demand types according to the demand type matching results is specifically as follows: the matching results of each user are aggregated and counted, the list of matched keywords of the user in the keyword library of each demand type is traversed, and the total number of hits for each keyword library is calculated (if the same keyword appears multiple times, the times are accumulated); for the matched keywords, different weights are assigned according to their importance (for example, the weight of "family trust" is higher than that of "conservative"), but the basic statistics are mainly based on the number of hits; finally, a statistical vector with the demand type as the dimension and the number of hits as the value is output, for example, a user gets 3 hits in the "conservative wealth management" keyword library, 5 hits in the "children's education fund planning" keyword library, and 0 hits in the "family trust demand" keyword library.

[0051] In detail, the demand type corresponding to the keyword library with the most hits is the main demand type of each target user. It is determined by comparing the hit counts of each demand type obtained in the previous step and selecting the demand type with the highest hit count as the user's main demand type. If two or more demand types have the same hit count, a decision is made according to a preset priority rule (e.g., family trust demand type > overseas allocation demand type > high-yield aggressive type > stable financial management type > children's education fund planning type > retirement planning type > insurance planning type), and the type with higher priority is selected. If the hit count of all demand types is 0, the user's main demand type is marked as "demand unclear" and subsequently included in the general cultivation pool.

[0052] In detail, the process of mapping the keyword library to transaction demand tags based on preset tag mapping rules involves, after confirming the keyword library corresponding to the main demand types, converting the name of the keyword library into a standard transaction demand tag string according to the preset tag mapping table. For example, the mapping rules are: "Stable financial management" keyword library → tag "Stable financial management", "Children's education fund planning" keyword library → tag "Children's education fund planning", "Family trust demand" keyword library → tag "Family trust demand", etc. Ultimately, a concise and unified transaction demand tag is generated for each user, facilitating subsequent display and filtering in the CRM system, task queue, and financial advisor workbench, thereby achieving refined service matching based on customer needs.

[0053] In this embodiment of the invention, the step of constructing customer group type tags for each target user based on the offline activity interaction information refers to using fields such as registration information, topics of interest, age, and occupation left by customers in offline activities (such as financial salons, lectures, exhibitions, etc.), and automatically identifying the customer group category (such as conservative financial management type, children's education fund planning type, family trust need type, etc.) through keyword matching and rule judgment, thereby generating a tag for each customer that reflects their core wealth management needs. The key to this step is that offline activities often directly reflect customers' true interests and intentions (such as checking "retirement planning" or filling in "children's overseas study funding arrangements" on the registration form), which is more certain than purely online behavior; extracting this information and matching it with a preset keyword library, combined with auxiliary information such as the customer's age and occupation, comprehensively judges the main customer group type tag that best matches the customer. The finally generated tag will be used for subsequent follow-up task customization—for example, automatically generating a task to send education fund plans to "children's education fund planning type" customers to achieve precise marketing.

[0054] In this embodiment of the invention, by generating transaction demand tags and customer group type tags, a dual-tag system is formed, which can accurately depict the customer's demand type and customer group attributes, making it easier to adopt differentiated service strategies and significantly improving marketing accuracy.

[0055] S6. Analyze the return visit service priority score of each target user based on the transaction demand tags and transaction intention feature values ​​of each target user.

[0056] In this embodiment of the invention, the step of analyzing the return visit service priority score of each target user based on the transaction demand tags and transaction intention feature values ​​of each target user includes: Based on a preset weight mapping rule, the transaction demand labels are mapped to weight scores to form transaction demand weight scores; The basic priority score of each target user is obtained by weighted summing of the transaction demand weight score and the transaction intention feature value. Based on the online consultation information, the transaction urgency level of each target user was analyzed; Extract the preset high-priority fields and preset low-priority fields from the key field set of each target user; The transaction urgency level, the high priority field, and the low priority field are respectively mapped to the transaction urgency score, high priority field score, and low priority field score of each target user; The weighted sum of the transaction urgency score, the high priority field score, and the low priority field score is used to obtain the corrected weight for each target user; The corrected weight of each target user is multiplied by the basic priority score to obtain the return visit service priority score for each target user.

[0057] In detail, the transaction demand labels are mapped to weighted scores based on preset weighted mapping rules, forming transaction demand weighted scores. These scores are assigned in advance by business experts based on the contribution of different demand types to the institution's commercial value and the urgency of conversion. Each transaction demand label is assigned a weighted score of 0 to 100. For example, "family trust demand type" is assigned 95 points due to its high average transaction value and short window period, "high-yield aggressive type" is assigned 85 points, "stable financial management type" is assigned 80 points, "children's education fund planning type" is assigned 75 points, "retirement planning type" is assigned 70 points, and "unclear demand type" is assigned 30 points.

[0058] In detail, the weighted summation of the transaction demand weight score and the transaction intention feature value to obtain the basic priority score for each target user is achieved by weighting and fusing the transaction demand weight score obtained in the previous step with the previously calculated transaction intention feature value (0-100 points) according to a preset ratio, for example, the intention feature value accounts for 70% of the weight and the demand weight score accounts for 30% of the weight. The system calculates for each user: Basic priority score = Transaction intention feature value × 0.7 + Transaction demand weight score × 0.3, and the result is rounded to the nearest integer and limited to the range of 0-100.

[0059] In detail, the analysis of the transaction urgency level of each target user based on the online consultation information involves classifying and evaluating the content reflecting time pressure in the user's online consultation text, extracting "urgency signals" (such as "urgently needed," "due next week," "funds will arrive soon," etc.) from the previously generated key demand fields, and combining them with specific time information (such as "within 3 days," "next Monday"). The system classifies the urgency level into three levels: high, medium, and low according to preset rules: high urgency is indicated by words such as "immediately / right now / today" or ≤3 days remaining; medium urgency is indicated by words such as "next week / nearly / within a month" or ≤30 days remaining; and low urgency is indicated by only "in the future / later" or no time signals.

[0060] In detail, the extraction of preset high-priority fields and preset low-priority fields from the key field set of each target user involves filtering specific words or phrases that are predefined by the system as "high-priority signals" and "low-priority signals" from the previously generated key field set. High-priority fields include strong intention words such as "assets in the tens of millions," "family business owner," and "schedule a meeting," while low-priority fields include weak intention words or negative signals such as "ask casually," "learn about it," and "ask for a friend." The system maintains two independent word lists to match the key fields of each user and records the list of matched high-priority fields and low-priority fields, providing raw materials for subsequent mapping.

[0061] In detail, mapping the transaction urgency level, the high-priority field, and the low-priority field to the transaction urgency score, high-priority field score, and low-priority field score of each target user respectively involves converting the non-numerical signal obtained in the previous step into a calculable numerical value: the transaction urgency level is mapped to a score, for example, high urgency → 30 points, medium urgency → 15 points, and low urgency → 0 points; the high-priority field score is accumulated based on the number of hits (each preset high-priority field +10 points, with a maximum of 50 points), and differential scores can be added based on different field weights; the low-priority field score is deducted based on the number of hits (each -5 points, with a minimum of -30 points).

[0062] In detail, the weighted summation of the transaction urgency score, the high-priority field score, and the low-priority field score to obtain the corrected weight for each target user is achieved by linearly combining the three score items according to a preset coefficient. For example, the corrected weight = 1 + (transaction urgency score + high-priority field score + low-priority field score) / 100, where the low-priority field score is usually negative. Thus, the baseline of the corrected weight is set to 1.0, which can be higher than 1.0 (e.g., a maximum of 1.8) or lower than 1.0 (e.g., a minimum of 0.7).

[0063] In detail, the step of multiplying the adjusted weight of each target user by the basic priority score to obtain the follow-up service priority score for each target user involves performing a multiplication operation on each user: Follow-up service priority score = Basic priority score × Adjusted weight, and rounding the result to the nearest integer within the range of 0 to 100; for example, if a user's basic score is 75 points and the adjusted weight is 1.3, then the final priority score is 98 points; this score integrates the customer's basic intention strength, demand commercial value, urgency, and strong / weak signal adjustments, and can more accurately reflect the order in which financial advisors should prioritize follow-up, with a higher score indicating a more urgent follow-up and a greater possibility of conversion.

[0064] In this embodiment of the invention, by combining the customer's intention-quantified score with the specific need type, the priority of follow-up visits is scientifically calculated to ensure that customers with high intention, high urgency, and high value can be followed up as soon as possible.

[0065] S7. Generate a return visit task queue based on the return visit service priority score and the customer group type label, and push the return visit task queue to a preset client.

[0066] In this embodiment of the invention, generating a return visit task queue based on the return visit service priority score of each target user and the customer group type label includes: Based on the online consultation information, obtain the names and contact information of each target user to obtain a basic user list; The basic user list is sorted according to the return visit service priority score to generate a sorted user list; Matching the customer group type tags with the preset task template library to obtain the follow-up task prompt text for each target user; The callback task prompt text is added to the sorted user list to obtain the callback task queue.

[0067] In detail, the step of obtaining the name and contact information of each target user based on the online consultation information to obtain a basic user list involves automatically extracting key contact information such as customer name (or nickname), mobile phone number, WeChat ID, and email address corresponding to each lead from the previously collected original online consultation records using regular expressions or preset field extraction rules. For heterogeneous data from different channels (official website messages, APP online consultation, WeChat official account), the data is converted according to a unified field mapping table, for example, mapping "tel", "phone", and "contact number" to the standard mobile phone number field.

[0068] In detail, the step of sorting the basic user list according to the follow-up service priority score to generate a sorted user list involves associating the basic user list obtained in the previous step with each user's follow-up service priority score (0-100) and arranging them in descending order of score. If multiple users have the same priority score, they are further sorted according to secondary rules such as transaction intention feature value, last consultation time, or urgency of need to ensure the refinement of the queue order. After sorting, a sorted user list with ranking number, user ID, name, contact information, and priority score is output. This list determines the order in which leads are displayed in the financial advisor's workbench.

[0069] In detail, the step of matching the customer group type tags with a preset task template library to obtain the follow-up task prompt text for each target user involves pre-building a task template library. This template library uses "customer group type tags" as the primary key and stores corresponding standard task description texts. For example, "Children's Education Fund Planning" corresponds to "Please call the customer, send the children's education fund planning proposal and schedule a meeting," "Family Trust Needs" corresponds to "Call immediately, send the family trust white paper and transfer to a dedicated advisor," and "Conservative Investment" corresponds to "Call today and recommend a list of recent conservative trust products." The system reads the customer group type tags for each user and performs precise matching in the template library. If a match is successful, the corresponding prompt text is returned directly. If the tag is "General" or "Needs Unclear," the default prompt text "Understand customer needs and provide a preliminary product introduction" is returned. Finally, a personalized follow-up task prompt text is generated for each user.

[0070] In detail, adding the task prompt text to the sorted user list to obtain the follow-up task queue involves appending the follow-up task prompt text generated for each user in the previous step as a new column to the corresponding record in the sorted user list, forming a complete task queue record. Each record includes user ID, name, contact information, priority score, customer group type tag, and specific follow-up task prompt text. A suggested completion time limit is automatically generated based on the priority score (e.g., P0 level requires contact within 1 hour, P1 level requires contact within today). After grouping the entire queue by consultant ID, it is pushed in real-time to a preset client (such as a financial advisor's CRM workbench, WeChat Work, or an app) via API or push notification. Consultants can open the client to see the priority-sorted list of customers to be followed up with and the standardized task guidance for each customer, thereby achieving automated generation and accurate assignment of follow-up tasks.

[0071] In this embodiment of the invention, differentiated follow-up tasks that match customer needs (such as sending specific product information, scheduling meetings, etc.) are automatically generated and pushed to the consultant's workbench in real time after being sorted by priority. This achieves full automation of the process from customer assessment to task assignment, reduces manual scheduling costs, and improves team efficiency and lead conversion rate.

[0072] In the fintech field, this solution integrates natural language processing, user behavior modeling, and automated workflow technologies to build an end-to-end intelligent marketing platform for banks, securities firms, trust companies, and third-party wealth management institutions. It automatically collects multi-terminal customer interaction data, quantitatively assesses asset strength, urgency of needs, and decision-making ability, and generates dual tags of "need stage + customer type." Ultimately, high-intent leads are synchronized to the CRM in real time and dynamically sorted, automatically assigning differentiated follow-up tasks. This application effectively solves problems such as subjective intent judgment, delayed follow-up timing, and system disconnect in traditional wealth management marketing, achieving a transformation from "experience-driven" to "data + algorithm-driven," and is a typical application scenario of fintech in intelligent marketing and customer relationship management.

[0073] As can be seen, the above solution proposes an intelligent high-intent lead dynamic tagging and CRM automation integration method for financial wealth management, aiming to solve the pain points of subjective intent judgment, delayed follow-up, and system disconnect in the traditional model. First, the system uniformly collects online consultation texts, offline interaction information, and online behavior logs from channels such as official website consultations, APP messages, and offline activities, forming a multi-source data foundation. Second, based on the frequency of online behavior information (such as visiting product pages, downloading manuals, and scheduling meetings), it calculates and weights the data to construct behavioral intent features; simultaneously, it extracts key fields such as assets, needs, and decision-making from consultation texts, quantifies and assesses the user's asset size, urgency of needs, and decision-making ability, and generates transaction intent feature values ​​through weighted fusion. Then, it generates transaction need tags based on a demand type thesaurus (such as conservative wealth management, family trust) matched with consultation content, and generates customer group type tags based on offline activity information. On this basis, it combines the transaction intent feature values ​​and demand tag weights to calculate a basic priority, and then introduces urgency level and high / low priority fields for correction, obtaining the final follow-up service priority score. Finally, users are sorted in descending order of priority, and preset task templates (such as "sending children's education fund plans") are automatically matched based on customer group type tags to generate a follow-up task queue with task prompts, which is pushed to the financial advisor's client in real time. The entire method realizes a closed loop of the entire process from lead acquisition, multi-dimensional quantitative assessment, dual tag generation to dynamic priority queue sorting and automatic task assignment, which significantly improves the conversion efficiency of high-net-worth client leads and team productivity.

[0074] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0075] In one embodiment, a customer follow-up priority analysis device is provided, which corresponds one-to-one with the customer follow-up priority analysis method in the above embodiments. For example... Figure 3 As shown, the customer follow-up priority analysis device includes a data acquisition module 101, a feature construction module 102, an intent analysis module 103, a tag construction module 104, and a task construction module 105. Detailed descriptions of each functional module are as follows: Data acquisition module 101 is used to acquire online consultation information, offline activity interaction information and online behavior information of each target user; The feature construction module 102 is used to construct behavioral intention features of each target user based on the online behavior information, identify preset key fields in the online consultation information, and obtain a set of key fields for each target user, wherein the preset key fields include asset key fields, demand key fields, and decision key fields. The intention analysis module 103 is used to analyze the data characteristics of each target user in multiple preset dimensions based on the set of key fields. The data characteristics include the data characteristics of the target user in the dimensions of transaction demand, asset distribution and decision-making ability. The data characteristics of each target user in different data dimensions are weighted and summed to obtain the transaction intention feature value of each target user. Tag building module 104 is used to build transaction demand tags for each target user based on the online consultation information, and to build customer group type tags for each target user based on the offline activity interaction information; The task construction module 105 is used to analyze the return visit service priority score of each target user based on the transaction demand tag and transaction intention feature value of each target user, generate a return visit task queue based on the return visit service priority score and the customer group type tag, and push the return visit task queue to a preset client.

[0076] In one embodiment, the feature construction module 102, when performing the step of constructing behavioral intention features for each target user based on the online behavior information, is specifically used for: Obtain the user behavior events contained in the online behavior information to form a behavior event sequence; Analyze the event type of each behavioral event in the behavioral event sequence; Count the number of event sequences corresponding to different event types in the behavioral event sequence; The behavior score for each event type is obtained by multiplying the number of event sequences corresponding to each event type by the preset base weight for each event type. The behavioral scores for each event type are normalized to obtain a normalized behavioral feature value table. Generate behavioral intention features for each target user based on the normalized behavioral feature value table.

[0077] In one embodiment, the feature construction module 102, when performing the process of identifying preset key fields in the online consultation information to obtain a set of key fields for each target user, is specifically used for: The online consultation texts are classified according to different preset channel types to obtain the original consultation text set. The original consultation text set is segmented to obtain a word sequence set; Each word in the word sequence set is tagged with its part of speech to generate a word sequence list after word segmentation; The segmented word sequence list is matched with a preset asset key field lexicon to obtain an asset key field set; The word segmentation sequence list is matched with the preset key field lexicon to obtain the key field set; The word segmentation sequence list is matched with a preset decision key field lexicon to obtain a decision key field set; The key field set is obtained by summarizing the set of key asset fields, the set of key demand fields, and the set of key decision fields.

[0078] In one embodiment, the intent analysis module 103, when performing the analysis of data characteristics of each target user across multiple preset dimensions based on the set of key fields, is specifically used for: Extract the key fields related to the urgency of the requirement from the set of key requirement fields contained in the set of key requirement fields to obtain the urgency of the requirement field; Based on preset mapping rules, the urgency field of the demand is mapped to the characteristic value of the transaction demand; Extract the key fields related to asset size from the asset key field set contained in the key field set to obtain the asset size field; Extract the numerical features of the asset size field to obtain the asset distribution feature value; The percentage of preset professional fields appearing in the decision-making key field set included in the key field set is statistically analyzed to generate decision-making capability feature values; By summarizing the transaction demand characteristic values, the asset distribution characteristic values, and the decision-making ability characteristic values, data characteristics of each target user in multiple preset dimensions are obtained.

[0079] In one embodiment, the tag building module 104, when performing the step of building transaction demand tags for each target user based on the online consultation information, is specifically used for: The online consultation information is segmented into words to obtain a sequence of consultation terms; The sequence of consultation terms is matched with a pre-set keyword library for different demand types to obtain demand type matching results; Based on the matching results of the aforementioned demand types, the number of times each target user's consultation word sequence was matched in the keyword database for different demand types was calculated. The demand type corresponding to the keyword library with the most hits is identified as the main demand type of each target user. The keyword library is mapped to transaction demand tags based on preset tag mapping rules.

[0080] In one embodiment, the task construction module 105, when performing the analysis of the return visit service priority score of each target user based on the transaction demand tags and transaction intention feature values ​​of each target user, is specifically used for: Based on a preset weight mapping rule, the transaction demand labels are mapped to weight scores to form transaction demand weight scores; The basic priority score of each target user is obtained by weighted summing of the transaction demand weight score and the transaction intention feature value. Based on the online consultation information, the transaction urgency level of each target user was analyzed; Extract the preset high-priority fields and preset low-priority fields from the key field set of each target user; The transaction urgency level, the high priority field, and the low priority field are respectively mapped to the transaction urgency score, high priority field score, and low priority field score of each target user; The weighted sum of the transaction urgency score, the high priority field score, and the low priority field score is used to obtain the corrected weight for each target user; The corrected weight of each target user is multiplied by the basic priority score to obtain the return visit service priority score for each target user.

[0081] In one embodiment, the task construction module 105, when executing the generation of the return visit task queue based on the return visit service priority score of each target user and the customer group type label, is specifically used for: Based on the online consultation information, obtain the names and contact information of each target user to obtain a basic user list; The basic user list is sorted according to the return visit service priority score to generate a sorted user list; Matching the customer group type tags with the preset task template library to obtain the follow-up task prompt text for each target user; The callback task prompt text is added to the sorted user list to obtain the callback task queue.

[0082] This invention provides a customer follow-up priority analysis device, aiming to address pain points in traditional models such as subjective intent judgment, delayed follow-up, and system disconnect. First, the system collects online consultation texts, offline interaction information, and online behavior logs from channels such as official website inquiries, APP messages, and offline activities, forming a multi-source data foundation. Second, based on the frequency of online behavior information (such as visiting product pages, downloading manuals, and scheduling meetings), the system calculates and weights the data to construct behavioral intent features. Simultaneously, it extracts key fields such as assets, needs, and decision-making from the consultation texts, quantifying and assessing the user's asset size, urgency of needs, and decision-making ability, and generating transaction intent feature values ​​through weighted fusion. Then, it generates transaction need tags by matching the consultation content with a demand type thesaurus (such as conservative financial management and family trust), and generates customer group type tags based on offline activity information. On this basis, it combines the transaction intent feature values ​​with the demand tag weights to calculate a basic priority, and then introduces urgency level and high / low priority fields for correction, obtaining the final follow-up service priority score. Finally, users are sorted in descending order of priority, and preset task templates (such as "sending children's education fund plan") are automatically matched based on customer group type tags to generate a follow-up task queue with task prompts, which is pushed to the financial advisor's client in real time. The entire method realizes a closed loop of the entire process from lead acquisition, multi-dimensional quantitative assessment, dual tag generation to dynamic sorting of priority queues and automatic task assignment, significantly improving the conversion efficiency of high-net-worth customer leads and team productivity. For specific limitations of the customer follow-up priority analysis device, please refer to the limitations of the customer follow-up priority analysis method above, which will not be repeated here. Each module in the above customer follow-up priority analysis device can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in the processor of the computer device in hardware form or independent of the processor, or it can be stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0083] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements the server-side functions or steps of a customer return visit priority analysis method.

[0084] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When executed by the processor, the computer program implements client-side functions or steps of a customer return visit priority analysis method.

[0085] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Acquire online consultation information, offline activity interaction information, and online behavior information of each target user; Based on the online behavior information, construct the behavioral intention characteristics of each target user; Identify the preset key fields in the online consultation information to obtain the key field set for each target user, wherein the preset key fields include asset key fields, demand key fields, and decision key fields; Based on the set of key fields, the data characteristics of each target user in multiple preset dimensions are analyzed. The data characteristics include the data characteristics of the target user in the dimensions of transaction demand, asset distribution and decision-making ability. The data characteristics of each target user in different data dimensions are weighted and summed to obtain the transaction intention feature value of each target user. Based on the online consultation information, transaction demand tags are constructed for each target user; based on the offline activity interaction information, customer group type tags are constructed for each target user. Analyze the priority score of the return visit service for each target user based on the transaction demand tags and transaction intention feature values ​​of each target user; A return visit task queue is generated based on the return visit service priority score and the customer group type label, and the return visit task queue is pushed to a preset client.

[0086] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: Acquire online consultation information, offline activity interaction information, and online behavior information of each target user; Based on the online behavior information, construct the behavioral intention characteristics of each target user; Identify the preset key fields in the online consultation information to obtain the key field set for each target user, wherein the preset key fields include asset key fields, demand key fields, and decision key fields; Based on the set of key fields, the data characteristics of each target user in multiple preset dimensions are analyzed. The data characteristics include the data characteristics of the target user in the dimensions of transaction demand, asset distribution and decision-making ability. The data characteristics of each target user in different data dimensions are weighted and summed to obtain the transaction intention feature value of each target user. Based on the online consultation information, transaction demand tags are constructed for each target user; based on the offline activity interaction information, customer group type tags are constructed for each target user. Analyze the priority score of the return visit service for each target user based on the transaction demand tags and transaction intention feature values ​​of each target user; A return visit task queue is generated based on the return visit service priority score and the customer group type label, and the return visit task queue is pushed to a preset client.

[0087] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0088] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0089] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0090] It should be noted that in the technical solutions disclosed in this invention, the acquisition of user information (personal image data (e.g., facial videos or pictures, facial feature videos or pictures, etc.) and personal privacy information (e.g., name, ID number, occupation, address, etc.)) is all completed with the user's knowledge and consent, and the acquisition of the relevant user information is legal and compliant.

[0091] Finally, it should be noted that any AI models, software tools, or components not belonging to this company appearing in the embodiments of this application are merely illustrative examples and do not represent actual use. All user personal information involved in the embodiments of this application is authorized (with knowledge and consent) by the relevant parties or fully authorized by all parties, and the executing entity can obtain it through various legal and compliant means. The collection, storage, use, processing, transmission, provision, and disclosure of the information, data, and signals involved all comply with relevant laws and regulations and do not violate public order and good morals. The above-described embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for prioritizing customer follow-up visits, characterized in that, include: Acquire online consultation information, offline activity interaction information, and online behavior information of each target user; Based on the online behavior information, construct the behavioral intention characteristics of each target user; Identify the preset key fields in the online consultation information to obtain the key field set for each target user, wherein the preset key fields include asset key fields, demand key fields, and decision key fields; Based on the set of key fields, the data characteristics of each target user in multiple preset dimensions are analyzed. The data characteristics include the data characteristics of the target user in the dimensions of transaction demand, asset distribution and decision-making ability. The data characteristics of each target user in different data dimensions are weighted and summed to obtain the transaction intention feature value of each target user. Based on the online consultation information, transaction demand tags are constructed for each target user; based on the offline activity interaction information, customer group type tags are constructed for each target user. Analyze the priority score of the return visit service for each target user based on the transaction demand tags and transaction intention feature values ​​of each target user; A return visit task queue is generated based on the return visit service priority score and the customer group type label, and the return visit task queue is pushed to a preset client.

2. The customer follow-up priority analysis method as described in claim 1, characterized in that, The step of constructing behavioral intention characteristics for each target user based on the online behavior information includes: Obtain the user behavior events contained in the online behavior information to form a behavior event sequence; Analyze the event type of each behavioral event in the behavioral event sequence; Count the number of event sequences corresponding to different event types in the behavioral event sequence; The behavior score for each event type is obtained by multiplying the number of event sequences corresponding to each event type by the preset base weight for each event type. The behavioral scores for each event type are normalized to obtain a normalized behavioral feature value table. Generate behavioral intention features for each target user based on the normalized behavioral feature value table.

3. The customer follow-up priority analysis method as described in claim 1, characterized in that, The process involves identifying preset key fields in the online consultation information to obtain a set of key fields for each target user, including: The online consultation texts are classified according to different preset channel types to obtain the original consultation text set. The original consultation text set is segmented to obtain a word sequence set; Each word in the word sequence set is tagged with its part of speech to generate a word sequence list after word segmentation; The segmented word sequence list is matched with a preset asset key field lexicon to obtain an asset key field set; The word segmentation sequence list is matched with the preset key field lexicon to obtain the key field set; The word segmentation sequence list is matched with a preset decision key field lexicon to obtain a decision key field set; The key field set is obtained by summarizing the set of key asset fields, the set of key demand fields, and the set of key decision fields.

4. The customer follow-up priority analysis method as described in claim 1, characterized in that, The step of analyzing the data characteristics of each target user across multiple preset dimensions based on the set of key fields includes: Extract the key fields related to the urgency of the requirement from the set of key requirement fields contained in the set of key requirement fields to obtain the urgency of the requirement field; Based on preset mapping rules, the urgency field of the demand is mapped to the characteristic value of the transaction demand; Extract the key fields related to asset size from the asset key field set contained in the key field set to obtain the asset size field; Extract the numerical features of the asset size field to obtain the asset distribution feature value; The percentage of preset professional fields appearing in the decision-making key field set included in the key field set is statistically analyzed to generate decision-making capability feature values; By summarizing the transaction demand characteristic values, the asset distribution characteristic values, and the decision-making ability characteristic values, data characteristics of each target user in multiple preset dimensions are obtained.

5. The customer follow-up priority analysis method as described in claim 1, characterized in that, The step of constructing transaction demand tags for each target user based on the online consultation information includes: The online consultation information is segmented into words to obtain a sequence of consultation terms; The sequence of consultation terms is matched with a pre-set keyword library for different demand types to obtain demand type matching results; Based on the matching results of the aforementioned demand types, the number of times each target user's consultation word sequence was matched in the keyword database for different demand types was calculated. The demand type corresponding to the keyword library with the most hits is identified as the main demand type of each target user. The keyword library is mapped to transaction demand tags based on preset tag mapping rules.

6. The customer follow-up priority analysis method as described in claim 1, characterized in that, The analysis of the return visit service priority score for each target user based on their transaction demand tags and transaction intention feature values ​​includes: Based on a preset weight mapping rule, the transaction demand labels are mapped to weight scores to form transaction demand weight scores; The basic priority score of each target user is obtained by weighted summing of the transaction demand weight score and the transaction intention feature value. Based on the online consultation information, the transaction urgency level of each target user was analyzed; Extract the preset high-priority fields and preset low-priority fields from the key field set of each target user; The transaction urgency level, the high priority field, and the low priority field are respectively mapped to the transaction urgency score, high priority field score, and low priority field score of each target user; The weighted sum of the transaction urgency score, the high priority field score, and the low priority field score is used to obtain the corrected weight for each target user; The corrected weight of each target user is multiplied by the basic priority score to obtain the return visit service priority score for each target user.

7. The customer follow-up priority analysis method as described in claim 1, characterized in that, The generation of the return visit task queue based on the return visit service priority score of each target user and the customer group type label includes: Based on the online consultation information, obtain the names and contact information of each target user to obtain a basic user list; The basic user list is sorted according to the return visit service priority score to generate a sorted user list; Matching the customer group type tags with the preset task template library to obtain the follow-up task prompt text for each target user; The callback task prompt text is added to the sorted user list to obtain the callback task queue.

8. A customer follow-up priority analysis device, characterized in that, include: The data acquisition module is used to acquire online consultation information, offline activity interaction information, and online behavior information of each target user; The feature construction module is used to construct behavioral intention features of each target user based on the online behavior information, identify preset key fields in the online consultation information, and obtain a set of key fields for each target user. The preset key fields include asset key fields, demand key fields, and decision key fields. The intention analysis module is used to analyze the data characteristics of each target user in multiple preset dimensions based on the set of key fields. The data characteristics include the data characteristics of the target user in the dimensions of transaction demand, asset distribution and decision-making ability. The data characteristics of each target user in different data dimensions are weighted and summed to obtain the transaction intention feature value of each target user. The tag building module is used to build transaction demand tags for each target user based on the online consultation information, and to build customer group type tags for each target user based on the offline activity interaction information. The task construction module is used to analyze the return visit service priority score of each target user based on the transaction demand tags and transaction intention feature values ​​of each target user, generate a return visit task queue based on the return visit service priority score and the customer group type tag, and push the return visit task queue to a preset client.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the customer return visit priority analysis method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the customer return visit priority analysis method as described in any one of claims 1 to 7.

Images