Insurance online channel flow attribution and budget allocation method based on big data analysis

Abstract

This invention provides a method for attributing traffic and allocating budgets for online insurance channels based on big data analysis, relating to the field of marketing data technology. Specific steps include: collecting fragmented online insurance marketing data from multiple sources and generating a comprehensive marketing dataset; classifying online marketing channels and matching them to corresponding conversion stages based on the user's insurance application process; adaptively configuring the backtracking period based on the decision-making time of different insurance types, correcting touchpoint weights using touchpoint screening rules and attenuation coefficients, and calculating the direct and indirect auxiliary contribution values ​​of channels using a time-series attribution algorithm; obtaining a comprehensive evaluation coefficient by combining the channel contribution value, customer acquisition cost per customer, and policy surrender risk value; and solving for the optimal budget allocation using an iterative algorithm, with the total marketing budget and single-channel allocation as constraints. This invention fully leverages the differentiated value of channel allocation, improves the quality of touchpoint data collection and attribution accuracy from the source, and optimizes the overall allocation structure of online insurance channels.

Description

Technical Field

[0001] This invention relates to the field of marketing data technology, specifically to a method for attributing traffic and allocating budgets for online insurance channels based on big data analysis. Background Technology

[0002] As the internet insurance industry continues to upgrade, insurance marketing models are gradually shifting from traditional offline direct sales and agency models to diversified online channels such as short video platforms, search engines, social media, third-party investment advisory platforms, and private domain traffic. Currently, insurance companies' online marketing budgets are increasing year by year, covering various types of insurance including property insurance, life insurance, health insurance, and accident insurance. However, online traffic is characterized by dispersed sources, long conversion paths, frequent user jumps, and overlapping channel traffic. Most insurance companies generally adopt a crude, average allocation or historical incremental allocation model for their marketing budgets, failing to accurately identify the true value of traffic from different channels. This easily leads to problems such as insufficient investment in high-quality channels, waste of inefficient channel resources, and duplicate billing for overlapping traffic, directly restricting the ROI of online customer acquisition. Therefore, leveraging big data technology to achieve refined traffic attribution and intelligent budget allocation is a core necessity for reducing costs and increasing efficiency in online insurance.

[0003] In the prior art, the omnichannel marketing funnel intelligent attribution model and automated budget allocation system disclosed in CN120807019A includes the following steps: The model includes a data acquisition module, used to acquire multi-channel marketing touchpoint interaction data, user behavior data, and channel type information through touchpoint recording devices, user identification devices, user-side behavior tracking devices, and channel classification devices deployed on marketing terminals; the path parsing module realizes data path parsing through time alignment unit time sequence synchronization, path association unit feature matching, and computing power scheduling unit dynamic adjustment; the attribution calculation module generates conversion attribution weights for each channel based on the matching data; and the budget mapping module establishes and updates the dynamic mapping relationship between weights and budgets. This system outputs attribution results through the model application module, and the budget execution module allocates the budget according to the dynamic mapping, which can accurately assess channel contribution, achieve scientific allocation of marketing budgets, and improve marketing efficiency and return on investment.

[0004] The existing technologies mentioned above have the following shortcomings: the existing technology has a general channel classification method, which does not refine the channels in combination with the insurance business scenario and complete the matching of each stage in combination with the insurance process. In addition, it uses a fixed backtracking period to collect touch point data, and lacks touch point screening and weight correction mechanisms. At the same time, the attribution model is simplistic and cannot break down the direct transaction contribution and indirect incubation contribution of the channel in a layered manner, resulting in low attribution accuracy and unbalanced budget allocation.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a method for attributing traffic and allocating budgets for online insurance channels based on big data analysis, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A method for attributing traffic and allocating budgets for online insurance channels based on big data analytics, comprising the following steps: S1. Collect fragmented multi-source marketing data of online insurance, and complete the association matching of heterogeneous data based on the user's unique identifier, insurance type, and ad placement timestamp as the three-level association primary key. After standardization processing, a full-domain marketing dataset is obtained. S2. Based on the full-domain marketing dataset, classify online marketing channels and, in conjunction with the user insurance purchase evolution process, match each online marketing channel to the corresponding conversion stage; S3. Based on the analysis of historical insurance samples of various insurance types using the full-domain marketing dataset, the exclusive retrospective period is adaptively configured according to the differentiated consumption decision-making time. The original interaction touchpoints of the corresponding channels in each conversion stage of the user's insurance process are collected. By setting dual screening conditions, effective touchpoints of channel interaction are captured. The exposure frequency decay coefficient is introduced to correct the touchpoint weight. Based on the multi-touchpoint time-series attribution algorithm, the contribution ratio is allocated according to the time sequence of touchpoint generation. The direct transaction contribution value and indirect auxiliary contribution value of each online marketing channel are calculated. S4. Retrieve channel placement data and policy surrender data, calculate the customer acquisition cost per customer and surrender risk value for each channel, obtain the contribution value of each online marketing channel based on the direct transaction contribution value and the indirect auxiliary contribution value, and obtain the comprehensive evaluation coefficient of each online marketing channel based on the contribution value, customer acquisition cost per customer, and surrender risk value. S5. With the total marketing budget and the budget for each online marketing channel as dual constraints, and with the goal of maximizing the sum of the comprehensive evaluation coefficients of all online marketing channels, the optimal budget allocation quota for each online marketing channel is output through an iterative algorithm.

[0008] Furthermore, the omni-channel marketing dataset contains full data on uninsured users, insured users, and users who have cancelled their policies.

[0009] Furthermore, relying on the three-level association primary key of user unique identifier, insurance type, and delivery timestamp, the association and matching of heterogeneous data is completed. The specific logic is as follows: By aggregating fragmented data of the same user using a unique user identifier, and classifying business types by insurance product category, the system utilizes campaign timestamps to align marketing activities, channel interactions, and policy purchase and cancellation records across different time periods, thereby achieving the interconnection and integration of multi-source heterogeneous data.

[0010] Furthermore, based on the full-domain marketing dataset, online marketing channels are classified. The specific logic is as follows: Based on the centralized channel source tags, placement attributes, and traffic acquisition methods in the omni-channel marketing dataset, all online marketing channels are categorized into paid channels, content channels, private domain channels, and organic traffic channels. Among them, paid channels refer to promotional channels that exchange financial investment for exposure and traffic; content channels refer to dissemination channels that use images, text, and short videos to attract users; private domain channels refer to dedicated operation channels that repeatedly reach users; and organic traffic channels refer to native traffic channels that do not require financial investment or active operation and are accessed by users through independent search. Based on the user's insurance application process, each online marketing channel is matched to the corresponding conversion stage. The specific logic is as follows: The entire user insurance process is divided into three major conversion stages: customer acquisition and lead generation, intention incubation, and transaction completion. The core of paid channels and organic traffic channels is to reach public domain users and match them to the customer acquisition and lead generation stage. Content channels rely on content output to cultivate user demand and match them to the intention incubation stage. Private domain channels deeply operate existing users and follow up on insurance intentions, matching them to both the intention incubation and transaction completion stages.

[0011] Furthermore, based on the analysis of historical insurance purchase samples for each type of insurance using the full-domain marketing dataset, and adaptively configuring a dedicated backtracking period according to the differentiated consumption decision-making time, the specific logic is as follows: Historical user data for each insurance type is extracted from the omni-channel marketing dataset, and behavioral samples of users who have successfully purchased insurance and those who have abandoned their policies are collected simultaneously. For a single insurance type, the decision-making time consumed by all sample users from the first click or browsing of the marketing channel to the completion of the policy purchase or termination decision is calculated. The decision-making time of all samples is summarized and data cleaning is performed. The speed of user decision-making for different insurance types is determined by combining the difficulty of handling different types of insurance, the complexity of the insurance process, and user consumption characteristics. The average decision-making time of users for the corresponding insurance type is calculated based on the cleaned and valid sample data, and this time is used as the configuration benchmark. Combined with the preset adaptive amplification coefficient, conversion compensation is performed, and finally, a special retrospective period adapted to each insurance type is generated.

[0012] Furthermore, the system collects the original interaction touchpoints of each conversion stage in the entire user insurance application process, and captures effective interaction touchpoints through setting dual filtering conditions. The specific logic is as follows: Using a pre-configured dedicated backtracking period as the time frame, all original interaction touchpoints generated by users with various online marketing channels during each conversion stage are collected. Original interaction touchpoints refer to all user access behaviors within online marketing channels, including ad impressions, page clicks, product browsing, and community viewing. A dual filtering condition is set, consisting of a page dwell time threshold and an interaction behavior type threshold, retaining only touchpoints where both page dwell time and interaction behavior reach the corresponding preset thresholds. Touchpoints that simultaneously meet both filtering conditions are determined to be valid channel interaction touchpoints.

[0013] Furthermore, an exposure frequency attenuation coefficient is introduced to correct the touch point weight, with the specific logic as follows: The time interval between all effective interaction touchpoints and the user's transaction point, as well as the frequency of channel exposure interactions, are statistically analyzed. An exposure frequency decay coefficient is then calculated, and the weight of effective interaction touchpoints is adjusted using the following formula:

[0014]

[0015] in, For the first Exposure frequency decay coefficient for each effective interactive touchpoint For the first The time interval between each effective interactive touchpoint and the user's transaction point. This is a dedicated retrospective period for the current insurance product. For the first Channel exposure and interaction frequency for each effective interaction touchpoint The highest frequency of exposure interaction among all valid interaction touchpoints for the same user. , These are the weighting coefficients for time interval and channel exposure interaction frequency, respectively. In this case, let , For the first Corrected weight coefficients for each valid interactive touchpoint For the first The initial weight coefficients for each valid interactive touchpoint An index for valid interactive touchpoints, , This represents the total number of valid interaction touchpoints within a single user's decision-making cycle. Based on the corrected weighting coefficients, and relying on the multi-touchpoint time-series attribution algorithm, the contribution ratio is allocated according to the time sequence of touchpoint generation. The direct transaction contribution value and indirect auxiliary contribution value of each online marketing channel are calculated using the following formula:

[0016]

[0017] in, For the first The basic contribution percentage of each effective interaction touchpoint For the first The final contribution value after correction of each effective interaction touchpoint This is the total contribution value corresponding to a single insurance transaction, and its value is 1. The contribution types of effective interaction touchpoints are classified and summarized according to the conversion stage in which they are located. The last effective interaction touchpoint before the transaction is completed is designated as the direct conversion touchpoint, and the contribution value of this type of touchpoint is accumulated to obtain the direct transaction contribution value of the corresponding channel. Effective interaction touchpoints used for early customer acquisition and lead generation and intention incubation are designated as auxiliary conversion touchpoints, and the contribution value of this type of touchpoint is accumulated to obtain the indirect auxiliary contribution value of the corresponding channel.

[0018] Furthermore, channel placement data and policy surrender data are retrieved to calculate the customer acquisition cost per customer and surrender risk value for each channel. The specific logic is as follows: Retrieve the total marketing cost, number of effective customers, number of policy transactions and number of surrendered policies for each online marketing channel within the preset statistical period, and calculate the customer acquisition cost per customer and the surrender risk value respectively. The cost per customer acquired for each online marketing channel is obtained by dividing the total marketing cost of each channel by the number of effective customers acquired. The surrender risk value for each online marketing channel is obtained by dividing the number of surrendered policies by the number of policy transactions. Based on contribution value, customer acquisition cost per customer, and policy surrender risk value, a comprehensive evaluation coefficient for each online marketing channel is obtained, using the following formula:

[0019] in, For the first The overall evaluation coefficient of each online marketing channel For the first The contribution value of each online marketing channel For the first The cost of acquiring a customer through each online marketing channel. For the first The risk value of policy cancellation through online marketing channels. As an index for online marketing channels, , The number of online marketing channels, , , These are the weighting coefficients for contribution value, customer acquisition cost per customer, and policy surrender risk value, respectively. , , The specific value is determined based on the analytic hierarchy process (AHP).

[0020] Furthermore, with the optimization objective of maximizing the sum of the comprehensive evaluation coefficients of all online marketing channels, an iterative algorithm is used to output the optimal budget allocation quota for each online marketing channel. The specific logic is as follows: A budget optimization model is built by combining the comprehensive evaluation coefficients of various online marketing channels, and the objective function and constraints corresponding to the budget optimization model are clarified. The objective function is to maximize the sum of the comprehensive evaluation coefficients of all online marketing channels. The constraints include the total marketing budget and the budget for each individual online marketing channel. The formula for the objective function is as follows:

[0021] Constraints:

[0022] in, This is the sum of the comprehensive evaluation coefficients for all online marketing channels. For the first Budget allocation for each online marketing channel remains to be determined. For the total marketing budget, , The first Minimum and maximum thresholds for budget allocation across online marketing channels; By relying on iterative algorithms to solve the budget optimization model repeatedly, the budget allocation schemes for each online marketing channel are continuously updated, gradually approaching the investment state that maximizes the sum of comprehensive evaluation coefficients. When the preset iteration termination condition is met, the calculation stops. The budget allocation quota obtained at this time is the optimal budget allocation quota for each online marketing channel.

[0023] Compared with the prior art, the beneficial effects of the present invention are: This invention leverages a comprehensive marketing dataset to refine online marketing channel types and matches channels to different conversion stages based on the user's insurance application evolution process, clarifying the functional positioning of each channel and fully exploring the differentiated value of channel placement. Simultaneously, it adaptively configures exclusive backtracking cycles based on the decision-making characteristics of different insurance types, coupled with dual touchpoint screening rules and an exposure frequency decay coefficient weight correction mechanism, to capture effective channel interaction touchpoints, improving the quality of touchpoint data collection and attribution accuracy from the source. Furthermore, relying on a multi-touchpoint time-series attribution algorithm, it allocates contribution ratios according to the time sequence of touchpoint generation, calculating the direct transaction contribution value and indirect auxiliary contribution value of the channel, comprehensively quantifying the explicit and implicit value of the channel. Based on this, it constructs a multi-dimensional comprehensive evaluation system combining single-customer acquisition cost and policy surrender risk value, taking into account placement revenue, placement cost, and operational risk. Using the total marketing budget and the placement budget of individual online marketing channels as constraints, it completes iterative optimization solutions, outputting budget allocations adapted to the enterprise's marketing placement model, balancing overall placement revenue and operational risk, and optimizing the overall placement structure of insurance online channels. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the overall method flow of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0026] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0027] Example: Please see Figure 1 The present invention provides a technical solution: A method for attributing traffic and allocating budgets for online insurance channels based on big data analytics, comprising the following steps: S1. Collect fragmented multi-source marketing data of online insurance, and complete the association matching of heterogeneous data based on the user's unique identifier, insurance type, and ad placement timestamp as the three-level association primary key. After standardization processing, a full-domain marketing dataset is obtained. Building upon the aforementioned embodiments, online insurance fragmented multi-source marketing data encompasses diverse and heterogeneous data generated from public domain advertising platforms, content dissemination platforms, private domain operation portals, and natural traffic portals. Specifically, this includes channel placement data, user behavior data, product interaction data, and policy transaction / cancellation data. Channel placement data includes placement amount, placement time, exposure channels, and insurance types. User behavior data includes interaction records such as ad exposure, page clicks, product browsing, and community access. Policy data includes insurance types, purchase time, transaction status, and cancellation records. Through methods such as event tracking, API calls, and log capture, raw fragmented data from four channels—paid channels, content channels, private domain channels, and natural traffic channels—are collected synchronously to achieve batch aggregation of multi-source data.

[0028] Building upon the above embodiments, the omni-channel marketing dataset simultaneously encompasses three categories of user data: uninsured, insured, and surrendered. Uninsured user data is used to assess the channel's customer acquisition and nurturing capabilities; insured user data serves as the core sample for attribution calculations; and surrendered user data is used to evaluate the quality of channel customers and potential risks. By covering the entire user lifecycle and refining the data sample dimensions, the dataset ensures comprehensive and reliable channel attribution and budget assessment results.

[0029] Based on the above embodiments, the association and matching of heterogeneous data is completed by relying on the three-level association primary key of user unique identifier, insurance type, and delivery timestamp. The specific logic is as follows: Online insurance marketing data is fragmented and structurally inconsistent, with user behavior data, campaign data, and policy data collected from different marketing channels operating independently, resulting in numerous isolated data silos. This embodiment employs a three-level primary key linkage binding method to achieve data fusion. First, using the user's unique identifier as the core primary key, it aggregates all fragmented interaction data generated by the same user across different marketing channels and access times. Second, using the insurance type as the secondary classification primary key, it categorizes and splits the aggregated data according to business types such as auto insurance, life insurance, and accident insurance, distinguishing the independent decision-making behaviors of users for different insurance types. Finally, using the campaign timestamp as the time sequence primary key, it unifies the time measurement dimension of all data sources, aligning marketing campaign information, channel browsing interaction information, insurance consultation information, insurance order information, and policy cancellation information for the same user within the same time period. This breaks down data barriers between multi-source heterogeneous data, ultimately achieving the interconnection, integration, and structural unification of multiple types of fragmented data.

[0030] Based on the above embodiments, a full-domain marketing dataset is obtained through standardization. The specific logic is as follows: After completing the association and matching of multi-source heterogeneous data, this embodiment addresses the common problems of messy formats, data redundancy, missing fields, and abnormal data in the original data by standardizing and managing the collected original data. Through data cleaning, abnormal and redundant data are removed, and missing fields are repaired. The data field formats and statistical standards of multiple channels are unified, and sensitive user privacy information is anonymized. After standardized integration of behavioral data, campaign data, and policy data from three types of users, a unified, complete, and clean full-domain marketing dataset is formed, providing a data foundation for subsequent channel classification, touchpoint retrospective analysis, attribution calculation, and budget optimization.

[0031] S2. Based on the full-domain marketing dataset, classify online marketing channels and, in conjunction with the user insurance purchase evolution process, match each online marketing channel to the corresponding conversion stage; Based on the above embodiments, relying on the channel source tags, placement attributes and traffic acquisition methods centralized in the omni-channel marketing dataset, all online marketing channels are classified into categories, including paid channels, content channels, private domain channels and organic traffic channels. Among them, paid channels refer to promotion channels that exchange financial investment for exposure traffic, content channels refer to dissemination channels that rely on text, images and short videos to acquire users, private domain channels refer to exclusive operation channels that repeatedly reach users, and organic traffic channels refer to native traffic channels that do not require financial investment or active operation and are accessed by users through independent search. Based on the user's insurance application process, each online marketing channel is matched to the corresponding conversion stage. The specific logic is as follows: The entire user insurance application process is divided into three major conversion stages: customer acquisition and lead generation, intention incubation, and transaction completion. Among them, the core of paid channels and organic traffic channels is to reach public domain users and match them to the customer acquisition and traffic generation stage; content channels rely on content output to cultivate user demand and match them to the intention incubation stage; private domain channels deeply operate existing users and follow up on insurance intentions, and match them to both the intention incubation and transaction completion stages.

[0032] Based on the above, it should be noted that: By meticulously categorizing online marketing channels and matching them with the entire user insurance purchase conversion process, this approach breaks away from the traditional single-channel classification model. It clarifies the traffic acquisition mechanisms and functional positioning of four types of marketing channels, distinguishing the differentiated roles of public domain customer acquisition, content cultivation, and private domain conversion. Based on the three-tiered conversion process, it breaks down the insurance user's insurance purchase decision-making chain, clarifying the contribution of each channel at different business stages. On the one hand, this facilitates subsequent selection of corresponding interaction touchpoints according to conversion stages, identifying the functional roles of channels in different stages of customer acquisition, intent incubation, and transaction closure. On the other hand, it avoids value bias caused by generalizing overall channel data, providing a layered data foundation for subsequent multi-touchpoint time-series attribution and differentiated weight calculation, thus improving the refinement of channel attribution from a structural perspective.

[0033] S3. Based on the analysis of historical insurance samples of various insurance types using the full-domain marketing dataset, the exclusive retrospective period is adaptively configured according to the differentiated consumption decision-making time. The original interaction touchpoints of the corresponding channels in each conversion stage of the user's insurance process are collected. By setting dual screening conditions, effective touchpoints of channel interaction are captured. The exposure frequency decay coefficient is introduced to correct the touchpoint weight. The multi-touchpoint time-series attribution algorithm is used to calculate the direct transaction contribution value and indirect auxiliary contribution value of each online marketing channel. Based on the above embodiments, the historical insurance purchase samples of each type of insurance are analyzed using the full-domain marketing dataset, and a dedicated retrospective period is adaptively configured according to the differentiated consumption decision-making time. The specific logic is as follows: Insurance business encompasses various types of insurance, including auto insurance, accident insurance, life insurance, and health insurance. The eligibility criteria, document submission requirements, and review process complexity differ significantly among these types, directly impacting the length of time users take to make purchasing decisions. Using a uniform, fixed backtracking period to collect touchpoint data can easily lead to data redundancy for short-decision insurance products and a lack of effective touchpoints for longer-decision insurance products. Therefore, this embodiment configures dedicated backtracking periods independently for different types of insurance.

[0034] In the specific implementation process, the data is first split from the omni-channel marketing dataset according to insurance product category tags. The full historical user data corresponding to each type of insurance is extracted. At the same time, two types of behavioral samples are collected: users who have successfully purchased insurance and users who abandoned their orders midway. This fully covers both complete transaction and non-transaction decision-making scenarios, avoiding the bias of statistical results caused by a single transaction sample. Secondly, for each user sample within a single insurance product, the decision-making start time is defined as the time when the user first touches any marketing channel and generates positive interactive behaviors such as browsing, clicking, and consulting. The decision-making end time is defined as the time when the user completes the policy payment, actively abandons the insurance, or has no interaction behavior for a long period of time. This accurately counts the complete decision-making time of a single user.

[0035] After acquiring all sample decision-making time data, the raw time data is cleaned and denoised to remove extreme values, abnormal blank data, and interference from invalid samples such as malicious data manipulation and misoperation. The user decision-making characteristics are comprehensively determined by combining the business attributes of each insurance type. Based on the difficulty of handling insurance, the steps of online insurance application process, the intensity of risk review, and the user's consumption habits, fast-decision insurance types and slow-decision insurance types are distinguished. Finally, the average user decision-making time for each insurance type is calculated based on the cleaned compliant samples and used as the configuration benchmark. It is then converted by matching a preset adaptive amplification compensation coefficient. That is, by multiplying and increasing the benchmark time, the extra time consumed by user intention incubation and repeated comparison and consultation is compensated. Finally, a dedicated backtracking cycle adapted to the characteristics of the corresponding insurance type and used for touchpoint backtracking is generated.

[0036] The adaptive amplification compensation coefficient is used to compensate for the limitations of the basic decision-making time. The basic decision-making time only includes the hard decision-making time from the user's initial interest to the final insurance purchase, excluding implicit incubation times such as early advertising browsing, multi-channel price comparison, and long-term interest observation. In this embodiment, the amplification compensation coefficient is set differently based on the different decision-making attributes of each insurance type, with an overall value range of 1.1 to 1.8. For quick-decision insurance products such as auto insurance and short-term accident insurance: users have short insurance decision-making cycles and little implicit observation time, so the amplification compensation coefficient is set at 1.1 to 1.3; For medium-decision insurance types such as million-dollar medical insurance and small-amount health insurance: users tend to make simple comparisons, so the amplification compensation coefficient is set at 1.3 to 1.5; For high-value, slow-decision insurance products such as long-term life insurance and critical illness insurance: users are highly cautious and have long incubation periods, so allow sufficient time for observation and comparison, and set the amplification compensation coefficient to 1.5 to 1.8.

[0037] Based on the above embodiments, the original interaction touchpoints of each conversion stage in the entire user insurance application process are collected and then... This embodiment uses the specific backtracking period for each insurance type, determined in the preceding steps, as a time constraint to limit the data collection interval for touchpoints. It comprehensively collects all original interaction touchpoints generated between target users and paid channels, content channels, private domain channels, and organic traffic channels within the three major conversion stages: customer acquisition, intent incubation, and transaction completion. These original interaction touchpoints refer to all user access behaviors within online marketing channels, primarily including basic interactive actions such as ad exposure, page clicks, product detail browsing, and viewing community content, completely recording the user's channel contact trajectory throughout the insurance purchase decision-making process.

[0038] It should be noted that the original interaction touchpoints contain a large number of superficial and invalid behaviors such as accidental touches, instantaneous access, and passive exposure. Such behaviors cannot represent the user's true intention to purchase insurance. If they are directly included in the attribution calculation, it will cause a flood of touchpoint data, weaken the weight of high-quality channel touchpoints, and thus reduce the accuracy of channel contribution measurement.

[0039] Therefore, this embodiment sets dual filtering conditions, namely, a page dwell time threshold and an interaction behavior threshold, to perform dual-layer filtering of the original touchpoints from two dimensions: access depth and behavior richness.

[0040] Regarding threshold settings: Page dwell time thresholds are primarily used to measure the depth of user engagement during a single visit. Combined with differentiated decision-making thresholds for different insurance types, a uniform page dwell time threshold of 3 seconds is set for short-cycle decision-making insurance products, while a uniform threshold of 5 seconds is set for long-cycle decision-making insurance products, filtering out fleeting and invalid visits. Interaction behavior thresholds use the type of interaction behavior as the evaluation criterion, rather than the number of single operations. Users are required to trigger at least two different types of interaction behavior during a single channel visit. This eliminates superficial invalid behaviors such as passive exposure and single accidental clicks, effectively identifying visit samples with genuine browsing intent and avoiding interference from homogeneous, single-operation behaviors in the touchpoint screening results.

[0041] During the screening process, each original interaction touchpoint undergoes dual-metric verification. Only touchpoints with page dwell time and interaction types meeting preset requirements are retained. If any metric fails to reach the preset threshold, the touchpoint is deemed a shallow, invalid touchpoint and is removed. This dual-mechanism linkage screening accurately removes valueless and redundant data, identifying high-quality, effective interaction touchpoints that truly reflect user decision-making preferences. This provides clean and reliable raw sample support for subsequent touchpoint weight correction and multi-touchpoint time-series attribution algorithms.

[0042] Based on the above embodiments, an exposure frequency attenuation coefficient is introduced to correct the touch point weight. The specific logic is as follows: The time interval between all effective interaction touchpoints across all channels and the user's transaction point, as well as the frequency of channel exposure interactions, are statistically analyzed. An exposure frequency decay coefficient is then calculated using the following formula:

[0043] in, For the first The exposure frequency decay coefficient of each effective interactive touchpoint is used to combine time interval, dedicated backtracking period, channel exposure interaction frequency and highest exposure interaction frequency to comprehensively consider the timeliness of effective interactive touchpoints and channel exposure saturation, and correct the weight of effective interactive touchpoints from both time and frequency dimensions. In the formula, For the first Exposure frequency decay coefficient for each effective interactive touchpoint For the first The time interval between each effective interactive touchpoint and the user's transaction point. This is a dedicated retrospective period for the current insurance product. For the first Channel exposure and interaction frequency for each effective interaction touchpoint The highest frequency of exposure interaction among all valid interaction touchpoints for the same user. , These are the weighting coefficients for time interval and channel exposure interaction frequency, respectively. An index for valid interactive touchpoints, , This represents the total number of valid interaction touchpoints within a single user's decision-making cycle. Based on the above, it should be noted that: Assuming a fixed dedicated backtracking period, a longer time interval between an effective interaction touchpoint and the user's transaction point indicates that the touchpoint occurred early in the user's decision-making process. Analyzing decision-making principles, users continuously receive information from various channels throughout their complete insurance purchase decision cycle. Early marketing touchpoints are easily covered and diluted by subsequent new information, gradually weakening their influence on the final insurance purchase decision, and thus reducing their actual contribution value. Conversely, shorter time intervals indicate that the touchpoint occurred towards the end of the decision-making process, with insufficient subsequent information to offset or cover it, resulting in a more significant boost to the transaction outcome. Therefore, the longer the time interval, the lower the overall contribution of the touchpoint. Consequently, the time interval is negatively correlated with the exposure frequency decay coefficient.

[0044] Under the condition of a constant peak exposure and interaction frequency for the same user, and based on the analysis of user insurance behavior principles, insurance belongs to a low-intention, high-prudence consumer product category. A single, superficial contact is insufficient to establish trust in insurance. Continuous and repeated interactive exposure through channels can gradually break down user unfamiliarity, continuously strengthen user awareness and memory of the channel and insurance products, accumulate consumer trust, alleviate insurance concerns, and steadily enhance the potential for this touchpoint to close a deal. That is, within a reasonable range, the higher the exposure and interaction frequency, the more prominent the channel's conversion effect. Therefore, there is a positive correlation between channel exposure and interaction frequency and the exposure frequency decay coefficient.

[0045] in, It adopts a decreasing function form. Combining the user decision-making behavior pattern, users will continuously receive multiple marketing messages throughout the entire insurance purchase process. Early interaction touchpoints are easily covered and diluted by the subsequent massive amount of information, and the decision value will spontaneously decay over time. Adopting the function form of "1-time percentage" can naturally adapt to this decay characteristic, mapping the time interval to the 0~1 range, intuitively reflecting the timeliness value of touchpoints, and matching the user's information memory pattern under the recency effect.

[0046] This is a positively increasing normalized function. Insurance products have a high average transaction value and a high level of expertise required, making it difficult for users to build trust through a single browsing session. By comparing the frequency of channel exposure interactions with the highest frequency of exposure interactions, we can achieve data normalization, unifying the data magnitude of the two components. Furthermore, by representing the degree to which channels accumulate user trust and cultivate insurance intentions through repeated outreach in a relative proportion, we can align with the underlying behavioral logic of the cumulative effect of exposure.

[0047] set up , The algorithm uses a weighted summation of the two representative terms, rather than a simple addition. Since different insurance products have different decision-making focuses—some emphasize short-term decisions and timeliness, while others emphasize long-term incubation and exposure—the weighted structure allows for flexible adjustment of the weighting of the time and frequency dimensions. This adapts to the decision-making characteristics of all types of insurance, including auto insurance, critical illness insurance, and life insurance, thus improving the algorithm's universality.

[0048] Because insurance purchase decisions are highly time-sensitive, information from touchpoints closer to the purchase stage has a greater impact on the final decision than cognitive gains from repeated exposure through various channels. However, excessively high exposure weighting can lead to marginal utility saturation, resulting in inflated weightings for high-frequency, low-efficiency touchpoints. Therefore, this embodiment assigns higher weight to the time dimension and lowers the frequency dimension, prioritizing the recency effect in determining touchpoint value and avoiding the problem of distorted attribution results caused by overestimating the cumulative effect of exposure. In this case, let ; As one implementation method, The value range is 0.5-1. The value range is 0-0.5. The specific value is set by technical personnel according to the actual situation and is not restricted here.

[0049] Based on the above embodiments, the effective interactive touchpoint weight correction is completed using the following formula:

[0050] in, For the first The initial weight coefficients for each valid interactive touchpoint.

[0051] Based on the above, it should be noted that: The initial weighting coefficients are statically assigned based solely on the type of user interaction behavior. Touchpoints of the same type are assigned the same weight value by default. This static assignment method fails to consider the differentiated characteristics such as the timing of touchpoint occurrence and channel exposure frequency, making it impossible to distinguish the varying impacts of the same type of touchpoint on user insurance decisions. This has significant limitations. If the initial weighting coefficients are directly used in attribution calculations, problems arise such as early, inefficient touchpoints having the same weight as high-value touchpoints nearing a transaction, and high-frequency nurturing touchpoints having homogeneous contributions compared to single, shallow touchpoints. This approach is unsuitable for the dynamic and long-term insurance decision-making scenarios. Therefore, this embodiment introduces an exposure frequency decay coefficient to perform a secondary correction on the initial weighting coefficients. This dynamic coefficient fine-tunes the base score of a single touchpoint, overcoming the drawbacks of static, equal weighting, and achieving differentiated and refined allocation of touchpoint weighting coefficients. This ensures that subsequent multi-touchpoint time-series attribution results are objective and accurate.

[0052] Based on the above embodiments, and using the corrected weighting coefficients, the multi-touchpoint time-series attribution algorithm allocates the contribution ratio according to the time sequence of touchpoint generation, and calculates the direct transaction contribution value and indirect auxiliary contribution value corresponding to each online marketing channel. The formula used is as follows:

[0053]

[0054] in, For the first The basic contribution percentage of each effective interaction touchpoint For the first The final contribution value after correction of each effective interaction touchpoint This is the total contribution value corresponding to a single insurance transaction, and its value is 1. The contribution types of effective interaction touchpoints are classified and summarized according to the conversion stage in which they are located. The last effective interaction touchpoint before the transaction is completed is designated as the direct conversion touchpoint, and the contribution value of this type of touchpoint is accumulated to obtain the direct transaction contribution value of the corresponding channel. Effective interaction touchpoints used for early customer acquisition and lead generation and intention incubation are designated as auxiliary conversion touchpoints, and the contribution value of this type of touchpoint is accumulated to obtain the indirect auxiliary contribution value of the corresponding channel.

[0055] Based on the above, it should be noted that: By leveraging the corrected touchpoint weights and combining them with time-series attribution to allocate the contribution percentage of each touchpoint: on the one hand, based on the recency effect of touchpoint occurrence timing aligning with users' insurance purchase decisions, the impact of touchpoints at different times on policy transactions is quantified; on the other hand, exposure frequency correction is used to eliminate measurement biases caused by touchpoint operational quality, avoiding the distortion of results caused by assigning values ​​solely based on the transaction interval; on this basis, direct transaction contributions and indirect auxiliary contributions are separated, accurately distinguishing the value of pre-conversion lead generation and the value of final transaction completion for each channel, improving the shortcomings of traditional attribution that only calculates the final transaction touchpoint and suppresses the contribution of early-stage cultivation channels, and providing quantitative support for subsequent differentiated budget allocation for each channel.

[0056] S4. Retrieve channel placement data and policy surrender data, calculate the customer acquisition cost per customer and surrender risk value for each channel, add the direct transaction contribution value and indirect auxiliary contribution value corresponding to the same channel, and use the sum as the contribution value of the online marketing channel. Based on the contribution value, customer acquisition cost per customer, and surrender risk value, obtain the comprehensive evaluation coefficient of each online marketing channel. Based on the above embodiments, channel deployment data and policy surrender data are retrieved to calculate the customer acquisition cost per customer and surrender risk value for each channel. The specific logic is as follows: Retrieve the total marketing cost, number of effective customers, number of policy transactions and number of surrendered policies for each online marketing channel within the preset statistical period, and calculate the customer acquisition cost per customer and the surrender risk value respectively. Among them, the preset statistical period is configured in combination with the different attributes of the insurance type: the natural month is used as the statistical period for short-term insurance, and the policy cooling-off period is added on the basis of the natural month for long-term insurance, with the policy cooling-off period being 10 to 15 days; The total marketing cost is taken from the back-end billing data of each channel; the channel access records are deduplicated based on the unique user identifier of the full-domain marketing dataset, multiple duplicate behavior data of the same user are removed, and then effective interactive users that meet the dual touchpoint filtering conditions are selected and retained, and the number of effective customer acquisition users of the corresponding channels is summarized; the number of policy transactions and the number of surrendered policies are extracted from the underwriting database and the surrender business database, respectively, for the corresponding period of effective policies and surrendered policies.

[0057] The cost per customer acquired for each online marketing channel is obtained by dividing the total marketing cost of each channel by the number of effective customers acquired. The surrender risk value for each online marketing channel is obtained by dividing the number of surrendered policies by the number of policy transactions.

[0058] Based on the above embodiments, the comprehensive evaluation coefficient of each online marketing channel is obtained according to the contribution value, customer acquisition cost per customer, and policy cancellation risk value. The formula used is as follows:

[0059] in, For the first The comprehensive evaluation coefficient of an online marketing channel is used to evaluate the overall operational performance of the online marketing channel by combining three indicators: contribution value, cost per customer acquisition, and risk of policy cancellation. The higher the comprehensive evaluation coefficient, the better the return on investment of the corresponding channel and the lower the risk of customer acquisition.

[0060] In the formula, For the first The contribution value of each online marketing channel For the first The cost of acquiring a customer through each online marketing channel. For the first The risk value of policy cancellation through online marketing channels. As an index for online marketing channels, , The number of online marketing channels; Based on the above, it should be noted that: The contribution value is the sum of direct sales contribution and indirect incubation contribution from the channel. It represents the policy conversion value created by the channel throughout the entire process of customer acquisition, lead generation, intent incubation, and transaction completion. The higher the channel contribution, the more business revenue it generates, and the higher the comprehensive evaluation coefficient becomes. Therefore, the contribution value and the comprehensive evaluation coefficient are positively correlated.

[0061] Cost per customer (CBC) is a deduction item, representing the financial investment required to acquire a single valid potential customer. Under the same output conditions, a higher CBC indicates greater marketing investment loss per conversion, lower ROI, and a larger cost value, resulting in a greater reduction in the overall evaluation coefficient and thus lowering the overall evaluation coefficient. Therefore, CBC and the overall evaluation coefficient are negatively correlated.

[0062] The surrender risk value is a risk deduction item, representing the percentage of policies surrendered through a particular channel. A higher surrender value indicates that the channel's customer acquisition intentions are not genuine, the customer quality is weak, the probability of lapsed returns on existing policies is high, and the risk of subsequent underwriting losses is greater. The higher the surrender risk value, the higher the hidden operational risks of the channel, and the more the comprehensive assessment coefficient is deducted. Therefore, the surrender risk value and the comprehensive assessment coefficient are negatively correlated.

[0063] In addition, the contribution value measures the revenue generated by the channel, the cost per customer acquisition measures the efficiency of the investment, and the risk value of surrender measures the quality of the customer base and the operational risk. These correspond to three independent dimensions: revenue, cost, and risk control. There is no repetition or nesting of indicators, so a linear structure with weighted summaries of each item is appropriate. The contribution value has a positive effect on channel benefits, while customer acquisition cost and policy surrender risk value are negative losses. Therefore, the contribution value is weighted and accumulated, while the customer acquisition cost and policy surrender risk value are weighted and deducted, which is in line with the inherent positive and negative attributes of the indicator.

[0064] Therefore, the above-mentioned functional form is used to express the functional relationship between the comprehensive evaluation coefficient and the three indicator parameters: combined contribution value, customer acquisition cost per customer, and surrender risk value.

[0065] In the formula, , , These are the weighting coefficients for contribution value, customer acquisition cost per customer, and policy surrender risk value, respectively. , , The specific value is determined based on the Analytic Hierarchy Process (AHP), and the specific logic is as follows: The three evaluation indicators—contribution value, customer acquisition cost per unit, and surrender risk value—are sequentially labeled. The relative importance of each pair of indicators is determined using a nine-scale method, with contribution value labeled as 1, customer acquisition cost per unit as 2, and surrender risk value as 3. A judgment matrix is ​​constructed, and industry experts score the indicators on a scale of 1-9 to create a third-order judgment matrix. The matrix expression is:

[0066] in, For row index, It is a column index, and , , Indicates that the index is The metrics relative to the index are The relative importance of the indicators to the comprehensive evaluation coefficient; Indicates that the index is The metrics compared to the index are In terms of indicators, it is extremely important for the comprehensive evaluation coefficient. Indicates that the index is The metrics compared to the index are In terms of indicators, the comprehensive evaluation coefficient is extremely unimportant; Divide each element value in the judgment matrix by the sum of its columns to obtain a normalized judgment matrix. Calculate the mean of the element values ​​in each row of the normalized judgment matrix. Use the mean of the first row as the proportional coefficient for the contribution value, the mean of the second row as the proportional coefficient for the customer acquisition cost per unit, and the mean of the third row as the proportional coefficient for the surrender risk value. With the constraint that the sum of the scaled values ​​equals 1, scale the contribution value, customer acquisition cost per unit, and surrender risk value proportionally. Use the scaled values ​​as the weights of the corresponding contribution value, customer acquisition cost per unit, and surrender risk value.

[0067] S5. With the total marketing budget and the budget for each online marketing channel as dual constraints, and with the goal of maximizing the sum of the comprehensive evaluation coefficients of all online marketing channels, the optimal budget allocation quota for each online marketing channel is output through an iterative algorithm.

[0068] Based on the above embodiments, with the optimization objective of maximizing the sum of the comprehensive evaluation coefficients of all online marketing channels, the optimal budget allocation quota for each online marketing channel is output through an iterative algorithm. The specific logic is as follows: A budget optimization model is built by combining the comprehensive evaluation coefficients of various online marketing channels, and the objective function and constraints corresponding to the budget optimization model are clarified. The objective function is to maximize the sum of the comprehensive evaluation coefficients of all online marketing channels. The constraints include the total marketing budget and the budget for each individual online marketing channel. The formula for the objective function is as follows:

[0069] Constraints:

[0070] in, This is the sum of the comprehensive evaluation coefficients for all online marketing channels. For the first Budget allocation for each online marketing channel remains to be determined. For the total marketing budget, , The first Minimum and maximum thresholds for budget allocation across online marketing channels; By relying on iterative algorithms to solve the budget optimization model repeatedly, the budget allocation schemes for each online marketing channel are continuously updated, gradually approaching the investment state that maximizes the sum of comprehensive evaluation coefficients. When the preset iteration termination condition is met, that is, when the number of iterations reaches the preset maximum number of iterations, such as 100, the calculation stops. The budget allocation quota obtained at this time is the optimal budget allocation quota for each online marketing channel.

[0071] Among them, the minimum threshold for the budget of each online marketing channel. Maximum threshold The settings are based on the historical deployment and operation data and business capacity of each channel; in, For the first The minimum effective spending amount historically required to maintain normal customer acquisition and traffic generation effects across all online marketing channels. For the first The maximum controllable spending amount corresponding to the customer base and conversion efficiency of each online marketing channel.

[0072] The total marketing budget is a preset fixed constraint value, which is determined comprehensively based on the company's overall marketing strategy plan for the current period, the total scale of online channel investment in the past period, and the upper limit of operating investment. It represents the global upper limit of funds for this round of budget optimization. The total budget amount remains constant during the iterative solution of the budget optimization model, and only the allocation quota of each online marketing channel is optimally decomposed.

[0073] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0074] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by software, electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0075] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0076] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for attributing traffic and allocating budgets for online insurance channels based on big data analytics, characterized in that, The specific steps include: S1. Collect fragmented multi-source marketing data of online insurance, and complete the association matching of heterogeneous data based on the user's unique identifier, insurance type, and ad placement timestamp as the three-level association primary key. After standardization processing, a full-domain marketing dataset is obtained. S2. Based on the full-domain marketing dataset, classify online marketing channels and, in conjunction with the user insurance purchase evolution process, match each online marketing channel to the corresponding conversion stage; S3. Based on the analysis of historical insurance samples of various insurance types using the full-domain marketing dataset, the exclusive retrospective period is adaptively configured according to the differentiated consumption decision-making time. The original interaction touchpoints of the corresponding channels in each conversion stage of the user's insurance process are collected. By setting dual screening conditions, effective touchpoints of channel interaction are captured. The exposure frequency decay coefficient is introduced to correct the touchpoint weight. Based on the multi-touchpoint time-series attribution algorithm, the contribution ratio is allocated according to the time sequence of touchpoint generation. The direct transaction contribution value and indirect auxiliary contribution value of each online marketing channel are calculated. S4. Retrieve channel placement data and policy surrender data, calculate the customer acquisition cost per customer and surrender risk value for each channel, obtain the contribution value of each online marketing channel based on the direct transaction contribution value and the indirect auxiliary contribution value, and obtain the comprehensive evaluation coefficient of each online marketing channel based on the contribution value, customer acquisition cost per customer, and surrender risk value. S5. With the total marketing budget and the budget for each online marketing channel as dual constraints, and with the goal of maximizing the sum of the comprehensive evaluation coefficients of all online marketing channels, the optimal budget allocation quota for each online marketing channel is output through an iterative algorithm.

2. The method for attributing traffic and allocating budget for online insurance channels based on big data analysis as described in claim 1, characterized in that, The omnichannel marketing dataset contains full data on uninsured users, insured users, and users who have cancelled their policies.

3. The method for attributing traffic and allocating budget for online insurance channels based on big data analysis as described in claim 1, characterized in that, Based on the three-level association primary key of user unique identifier, insurance type, and delivery timestamp, heterogeneous data association and matching are completed. The specific logic is as follows: By aggregating fragmented data of the same user using a unique user identifier, and classifying business types by insurance product category, the system utilizes campaign timestamps to align marketing activities, channel interactions, and policy purchase and cancellation records across different time periods, thereby achieving the interconnection and integration of multi-source heterogeneous data.

4. The method for attributing traffic and allocating budgets for online insurance channels based on big data analysis as described in claim 1, characterized in that, The online marketing channels were categorized based on the full-domain marketing dataset. The specific logic is as follows: Based on the centralized channel source tags, placement attributes, and traffic acquisition methods in the omni-channel marketing dataset, all online marketing channels are categorized into paid channels, content channels, private domain channels, and organic traffic channels. Among them, paid channels refer to promotional channels that exchange financial investment for exposure and traffic; content channels refer to dissemination channels that use images, text, and short videos to attract users; private domain channels refer to dedicated operation channels that repeatedly reach users; and organic traffic channels refer to native traffic channels that do not require financial investment or active operation and are accessed by users through independent search. Based on the user's insurance purchase process, each online marketing channel is matched to the corresponding conversion stage. The specific logic is as follows: The entire user insurance process is divided into three major conversion stages: customer acquisition and lead generation, intention incubation, and transaction completion. The core of paid channels and organic traffic channels is to reach public domain users and match them to the customer acquisition and lead generation stage. Content channels rely on content output to cultivate user demand and match them to the intention incubation stage. Private domain channels deeply operate existing users and follow up on insurance intentions, matching them to both the intention incubation and transaction completion stages.

5. The method for attributing traffic and allocating budget for online insurance channels based on big data analysis according to claim 4, characterized in that, Based on the analysis of historical insurance purchase samples for each type of insurance using the full-domain marketing dataset, and adaptively configuring a dedicated backtracking period according to the differentiated consumption decision-making time, the specific logic is as follows: Historical user data for each insurance type is extracted from the omni-channel marketing dataset, and behavioral samples of users who have successfully purchased insurance and those who have abandoned their policies are collected simultaneously. For a single insurance type, the decision-making time consumed by all sample users from the first click or browsing of the marketing channel to the completion of the policy purchase or termination decision is calculated. The decision-making time of all samples is summarized and data cleaning is performed. The speed of user decision-making for different insurance types is determined by combining the difficulty of handling different types of insurance, the complexity of the insurance process, and user consumption characteristics. The average decision-making time of users for the corresponding insurance type is calculated based on the cleaned and valid sample data, and this time is used as the configuration benchmark. Combined with the preset adaptive amplification coefficient, conversion compensation is performed, and finally, a special retrospective period adapted to each insurance type is generated.

6. The method for attributing traffic and allocating budget for online insurance channels based on big data analysis as described in claim 5, characterized in that, The system collects original interaction touchpoints from various channels at each conversion stage throughout the user insurance purchase process, and captures effective interaction touchpoints by setting dual filtering conditions. The specific logic is as follows: Using a pre-configured dedicated backtracking period as the time frame, all original interaction touchpoints generated by users with various online marketing channels during each conversion stage are collected. Original interaction touchpoints refer to all user access behaviors within online marketing channels, including ad impressions, page clicks, product browsing, and community viewing. A dual filtering condition is set, consisting of a page dwell time threshold and an interaction behavior type threshold, retaining only touchpoints where both page dwell time and interaction behavior reach the corresponding preset thresholds. Touchpoints that simultaneously meet both filtering conditions are determined to be valid channel interaction touchpoints.

7. The method for attributing traffic and allocating budget for online insurance channels based on big data analysis as described in claim 6, characterized in that, The exposure frequency attenuation coefficient is introduced to correct the touch point weight. The specific logic is as follows: The time interval between all effective interaction touchpoints and the user's transaction point, as well as the frequency of channel exposure interactions, are statistically analyzed. An exposure frequency decay coefficient is then calculated, and the weight of effective interaction touchpoints is adjusted using the following formula: , ,in, For the first Exposure frequency decay coefficient for each effective interactive touchpoint For the first The time interval between each effective interactive touchpoint and the user's transaction point. This is a dedicated retrospective period for the current insurance product. For the first Channel exposure and interaction frequency for each effective interaction touchpoint The highest frequency of exposure interaction among all valid interaction touchpoints for the same user. , These are the weighting coefficients for time interval and channel exposure interaction frequency, respectively. In this case, let , For the first Corrected weighting coefficients for each valid interactive touchpoint For the first The initial weight coefficients of each valid interactive touchpoint An index for valid interactive touchpoints, , This represents the total number of valid interaction touchpoints within a single user's decision-making cycle. Based on the corrected weighting coefficients, and relying on the multi-touchpoint time-series attribution algorithm, the contribution ratio is allocated according to the time sequence of touchpoint generation. The direct transaction contribution value and indirect auxiliary contribution value of each online marketing channel are calculated using the following formula: , ,in, For the first The basic contribution percentage of each effective interaction touchpoint For the first The final contribution value after correction of each effective interaction touchpoint This is the total contribution value corresponding to a single insurance transaction, and its value is 1. The contribution types of effective interaction touchpoints are classified and summarized according to the conversion stage in which they are located. The last effective interaction touchpoint before the transaction is completed is designated as the direct conversion touchpoint, and the contribution value of this type of touchpoint is accumulated to obtain the direct transaction contribution value of the corresponding channel. Effective interaction touchpoints used for early customer acquisition and lead generation and intention incubation are designated as auxiliary conversion touchpoints, and the contribution value of this type of touchpoint is accumulated to obtain the indirect auxiliary contribution value of the corresponding channel.

8. The method for attributing traffic and allocating budget for online insurance channels based on big data analysis according to claim 1, characterized in that, The system retrieves channel placement data and policy surrender data to calculate the customer acquisition cost per customer and the surrender risk value for each channel. The specific logic is as follows: Retrieve the total marketing cost, number of effective customers, number of policy transactions and number of surrendered policies for each online marketing channel within the preset statistical period, and calculate the customer acquisition cost per customer and the surrender risk value respectively. The cost per customer acquired for each online marketing channel is obtained by dividing the total marketing cost of each channel by the number of effective customers acquired. The surrender risk value for each online marketing channel is obtained by dividing the number of surrendered policies by the number of policy transactions. Based on contribution value, customer acquisition cost per customer, and policy surrender risk value, a comprehensive evaluation coefficient for each online marketing channel is obtained, using the following formula: ,in, For the first The overall evaluation coefficient of each online marketing channel For the first The contribution value of each online marketing channel For the first The cost of acquiring a customer through each online marketing channel. For the first The risk value of policy cancellation through online marketing channels. As an index for online marketing channels, , The number of online marketing channels, , , These are the weighting coefficients for contribution value, customer acquisition cost per customer, and policy surrender risk value, respectively. , , The specific value is determined based on the analytic hierarchy process (AHP).

9. The method for attributing traffic and allocating budget for online insurance channels based on big data analysis as described in claim 8, characterized in that, The optimization objective is to maximize the sum of the comprehensive evaluation coefficients of all online marketing channels. An iterative algorithm is used to output the optimal budget allocation quota for each online marketing channel. The specific logic is as follows: A budget optimization model is built by combining the comprehensive evaluation coefficients of various online marketing channels, and the objective function and constraints corresponding to the budget optimization model are clarified. The objective function is to maximize the sum of the comprehensive evaluation coefficients of all online marketing channels. The constraints include the total marketing budget and the budget for each individual online marketing channel. The formula for the objective function is as follows: Constraints: ,in, This is the sum of the comprehensive evaluation coefficients for all online marketing channels. For the first Budget allocation for each online marketing channel remains to be determined. For the total marketing budget, , The first Minimum and maximum thresholds for budget allocation across online marketing channels; By relying on iterative algorithms to solve the budget optimization model repeatedly, the budget allocation schemes for each online marketing channel are continuously updated, gradually approaching the investment state that maximizes the sum of comprehensive evaluation coefficients. When the preset iteration termination condition is met, the calculation stops. The budget allocation quota obtained at this time is the optimal budget allocation quota for each online marketing channel.

Images