Omnichannel marketing funnel intelligent attribution model and automatic budget distribution system

Abstract

The invention relates to the technical field of digital marketing, and discloses an omni-channel marketing funnel intelligent attribution model and an automatic budget distribution system, and the model comprises a data collection module which is used for collecting data through a contact recording device and a user identification device which are disposed at a marketing terminal, and a behavior tracking device and a channel classification device which are disposed at a user side. Acquiring information such as multi-channel marketing contact interaction data, user behavior data and channel types; the path analysis module realizes data path analysis through time sequence synchronization of a time alignment unit, feature matching of a path association unit and dynamic adjustment of a computing power scheduling unit; the attribution calculation module generates the conversion attribution weight of each channel based on the matching data; and the budget mapping module establishes a dynamic mapping relationship between the weight and the budget and updates the dynamic mapping relationship. The system outputs the attribution result through the model application module, and the budget execution module distributes the budget according to the dynamic mapping, so that the channel contribution can be accurately evaluated, the scientific distribution of the marketing budget is realized, and the marketing efficiency and the rate of return on investment are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital marketing, specifically a full-channel marketing funnel intelligent attribution model and an automatic budget allocation system. BACKGROUND

[0002] In today's digital marketing environment, the marketing channels of enterprises are increasingly diversified, including search engine marketing, social media marketing, email marketing, content marketing, and other forms. Consumers often interact with multiple channels during the purchase decision-making process, forming complex conversion paths. However, traditional marketing attribution methods have many defects and cannot accurately assess the actual contribution of each channel in the conversion process.

[0003] Traditional single-touch attribution models, such as last-click attribution, attribute all conversion credit to the channel that the consumer last contacted. This method is too simple and ignores the influence of other channels on the consumer decision-making process, making it difficult to fully reflect the true value of each channel. Linear attribution models consider multiple touchpoints, but they average the weights of all touchpoints without considering the different roles of different channels at different stages of the conversion process. For example, some channels may be effective in attracting potential customers, while others may be more critical in facilitating the final purchase. Therefore, this model also cannot accurately measure the actual contribution of each channel.

[0004] With the explosive growth of marketing data, traditional attribution methods are also struggling with data processing capabilities. They often cannot efficiently process massive amounts of marketing touchpoint interaction data and user behavior data, making it difficult to collect, organize, and analyze data in real time. Moreover, traditional methods lack precise handling of data time dimensions and cannot accurately synchronize data from different channels over time, leading to deviations when analyzing conversion paths.

[0005] Traditional marketing budget allocation methods are often based on historical experience and subjective judgment, lacking a scientific dynamic connection with attribution results. This makes it difficult for enterprises to allocate marketing budgets based on the actual conversion contribution of each channel, potentially leading to resource waste and failing to maximize marketing effectiveness. For example, some channels that have a high actual contribution may not receive enough budget support, while some ineffective channels may consume too many resources.

[0006] In this context, there is an urgent need for an intelligent attribution model and system that can accurately analyze the effectiveness of multi-channel marketing and reasonably allocate marketing budgets to address the shortcomings of traditional methods in data processing, attribution accuracy, and budget allocation, helping enterprises improve marketing efficiency and return on investment. SUMMARY

[0007] The present application aims to provide a full-channel marketing funnel intelligent attribution model and an automatic budget allocation system to solve the problems raised in the background art.

[0008] To achieve the above-mentioned purpose, the present application provides a full-channel marketing funnel intelligent attribution model, which comprises:

[0009] A data acquisition module, comprising a contact recording device deployed at a marketing terminal and a behavior tracking device arranged at a user side, wherein the contact recording device is used to acquire interaction data of multi-channel marketing contacts, and the behavior tracking device is used to acquire behavior data of a user conversion path;

[0010] A path analysis module, comprising a time alignment unit and a path association unit, wherein the time alignment unit is used to perform time sequence synchronization processing on marketing contact interaction data and user behavior data, and the path association unit is used to perform path matching association on the time sequence synchronized marketing contact interaction data and user behavior data;

[0011] An attribution calculation module, used to generate conversion attribution weights of each marketing channel based on the path matching associated marketing contact interaction data and user behavior data;

[0012] A budget mapping module, used to establish a dynamic mapping relationship between conversion attribution weights and marketing budgets, and update the dynamic mapping relationship based on the conversion attribution weights.

[0013] Preferably, the data acquisition module further comprises a user identification device arranged at the marketing terminal and a channel classification device arranged at the user side, wherein the user identification device is used to record user unique identification information, and the channel classification device is used to mark channel type information of marketing contacts.

[0014] Preferably, the path analysis module comprises a main analysis unit and a plurality of slave analysis units, wherein the main analysis unit is used to coordinate synchronization parameters of the time alignment unit based on time deviation of marketing contact interaction data and user behavior data;

[0015] Each of the slave analysis units corresponds to at least one data acquisition device, and the slave analysis unit is used to perform path matching association operation based on data acquired by the corresponding data acquisition device.

[0016] Preferably, the main analysis unit coordinates synchronization parameters of the time alignment unit based on time deviation of marketing contact interaction data and user behavior data, comprising:

[0017] The time alignment unit performs time labeling on marketing contact interaction data and user behavior data based on an initial synchronization period;

[0018] determine a data delay sub-domain based on a timestamp difference between the marketing touch interaction data and the user behavior data;

[0019] determine a delay collection sub-domain from a plurality of data collection sub-domains based on the data delay sub-domain;

[0020] adjust a synchronization period of a data collection device in the delay collection sub-domain to a target synchronization period, wherein the target synchronization period is less than the initial synchronization period.

[0021] Preferably, the master analysis unit determines a data delay sub-domain based on a timestamp difference between the marketing touch interaction data and the user behavior data, comprising:

[0022] For each data collection device, determine a time delay amount of the data collection device in a plurality of collection time periods based on a data timestamp obtained by the data collection device;

[0023] determine the data delay sub-domain based on the time delay amount of each data collection device in a plurality of collection time periods.

[0024] Preferably, the slave analysis unit performs path matching association operation based on the data obtained by the corresponding data collection device, comprising:

[0025] For each data collection device, the slave analysis unit corresponding to the data collection device performs interference filtering on the marketing touch interaction data and the user behavior data obtained by the data collection device in a plurality of time periods, and the slave analysis unit corresponding to the data collection device extracts a touch feature vector and a behavior feature vector based on the filtered marketing touch interaction data and user behavior data;

[0026] When the matching degree of the touch feature vector and the behavior feature vector is greater than a preset matching degree threshold, determine an association path of the marketing touch interaction data and the user behavior data based on channel positioning information of the data collection device, and complete the path matching association operation based on the association path.

[0027] Preferably, the slave analysis unit performs interference filtering on the marketing touch interaction data and the user behavior data obtained by the data collection device in a plurality of time periods, comprising:

[0028] extract a touch type feature and an interaction frequency feature based on the marketing touch interaction data obtained by the data collection device in a plurality of time periods;

[0029] extract a channel coverage feature and a path length feature based on the user behavior data obtained by the data collection device in a plurality of time periods;

[0030] The marketing touch interaction data and the user behavior data are disturbed and filtered based on the touch type feature, the interaction frequency feature, the channel coverage feature and the path length feature through a multi-source denoising model.

[0031] Preferably, the path analysis module further comprises a computing power scheduling unit, which is configured to adjust the real-time correspondence between the plurality of slave analysis units and the plurality of data collection devices based on the synchronization period of each data collection device.

[0032] Preferably, the computing power scheduling unit adjusts the real-time correspondence between the plurality of slave analysis units and the plurality of data collection devices based on the synchronization period of each data collection device, comprising:

[0033] The data collection device in the delay collection sub-domain is regarded as a high-load collection device, and the slave analysis unit corresponding to the high-load collection device is regarded as a slave analysis unit to be adjusted;

[0034] The real-time correspondence between the plurality of slave analysis units and the plurality of data collection devices is adjusted based on the high-load collection device and the slave analysis unit to be adjusted.

[0035] Preferably, the application further comprises a full-channel marketing funnel automation budget allocation system, which comprises a model application module and a budget execution module, wherein the model application module is configured to call the full-channel marketing funnel intelligent attribution model and output the conversion attribution result, and the budget execution module is configured to allocate marketing budget to each marketing channel based on the dynamic mapping relationship.

[0036] Compared with the prior art, the application has the following advantages:

[0037] In the data collection link, the touch record device deployed in the marketing terminal and the behavior tracking device on the user side can comprehensively and accurately obtain the interaction data of multi-channel marketing touch and the behavior data of the user conversion path. At the same time, the user identification device records the unique identification information of the user, and the channel classification device labels the channel type information of the marketing touch, which lays a solid foundation for subsequent data analysis and processing, and ensures the accuracy and traceability of the data.

[0038] The design of the path analysis module has great advantages. The time alignment unit performs time synchronization processing on the marketing touch interaction data and user behavior data, the main analysis unit can coordinate the synchronization parameters based on the time deviation, and effectively solves the problem of data time synchronization by determining the data delay sub-domain and adjusting the synchronization period, which ensures the consistency of the data in the time dimension. The path correlation unit filters the data from the analysis unit, extracts the touch feature vector and behavior feature vector, and determines the correlation path when the matching degree meets the condition, realizing accurate matching and correlation of the path. The computing power scheduling unit can also adjust the correspondence between the slave analysis unit and the data collection device according to the synchronization period, optimize the allocation of computing resources of the system, and improve the data processing efficiency.

[0039] The attribution calculation module generates the conversion attribution weight of each marketing channel based on the data after path matching and correlation, which can scientifically and objectively evaluate the actual contribution of each channel in the conversion process, change the limitations of traditional attribution methods, and make the attribution results more accurate and reliable.

[0040] The budget mapping module establishes a dynamic mapping relationship between the conversion attribution weight and the marketing budget, and updates the mapping relationship according to the attribution weight, realizing the automatic and scientific allocation of marketing budget. Enterprises can reasonably allocate the budget according to the actual contribution of each channel, invest resources in channels with better effects, improve the utilization efficiency of marketing resources, and maximize the marketing effect.

[0041] The whole system realizes intelligent processing of the whole process from data collection, path analysis, attribution calculation to budget allocation, can efficiently process massive data, accurately analyze multi-channel marketing effect, and provides strong support for enterprise marketing decision-making, helping enterprises improve competitiveness in complex market environment and obtain higher investment return rate. BRIEF DESCRIPTION OF DRAWINGS

[0042] Figure 1 The working principle diagram of the full-channel marketing funnel intelligent attribution model described in the present application;

[0043] Figure 2 The process diagram of the master-slave unit cooperation of the path analysis module;

[0044] Figure 3 The flowchart of the time alignment unit synchronization parameter adjustment;

[0045] Figure 4 The flowchart of the path matching and correlation of the slave analysis unit;

[0046] Figure 5 The flowchart of interference filtering. DETAILED DESCRIPTION

[0047] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work are within the protection scope of the present application.

[0048] Please refer to Figures 1-5 The present application provides a full-channel marketing funnel intelligent attribution model, which includes.

[0049] The data collection module is deployed in the contact recording device of the marketing terminal and the behavior tracking device on the user side. The contact recording device obtains interaction data of multi-channel marketing contacts, such as interaction information of users and marketing content in different marketing channels (such as search engines, social media, emails, etc.), including click time, interaction type, content display duration, etc. The behavior tracking device obtains behavior data of the user conversion path, covering all operation behaviors of the user from first contact with the marketing contact to finally complete the conversion (such as purchase, registration, etc.) process, such as page browsing sequence, dwell time, jump record, etc.

[0050] The path analysis module includes a time alignment unit and a path association unit. The time alignment unit performs time sequence synchronization processing on the marketing contact interaction data and the user behavior data, and ensures the time consistency of different source data through a unified time reference. The path association unit matches and associates the data after time sequence synchronization, analyzes the corresponding relationship between user behavior and marketing contact, and constructs a complete conversion path.

[0051] The attribution calculation module generates conversion attribution weights of each marketing channel based on the data after path matching and association, and quantifies the contribution degree of each channel in the conversion process through a specific algorithm (such as a rule-based, data-driven or hybrid model).

[0052] The budget mapping module establishes a dynamic mapping relationship between the conversion attribution weight and the marketing budget, adjusts the budget allocation proportion of each channel according to the real-time updated attribution weight, and realizes the dynamic optimization of the budget.

[0053] Embodiment 1:

[0054] The data collection module includes, in addition to the contact recording device deployed at the marketing terminal and the behavior tracking device arranged at the user side, a user identification device arranged at the marketing terminal and a channel classification device arranged at the user side. The user identification device records the unique identification information of the user. The device can achieve the unique identification of the user in multiple ways. For example, when the user registers an account, the system generates a unique user ID, which is associated with the user's registration information. When the user operates at different marketing terminals, the user identification device reads the user's registered account information and converts it into unique identification information. Alternatively, the user identification device can extract the identifier of the user's device, such as the IMEI number and MAC address, which are unique and assigned when the device is manufactured. In actual application, the user identification device monitors the user's operation behavior at the marketing terminal in real time. When the user logs in, the device immediately obtains the account information and generates the corresponding unique identification information. At the same time, when the user performs interactive operations such as browsing and clicking, the identification information is bound to the interactive data, so that the user's interactive behavior at different times and different marketing touchpoints can be accurately identified.

[0055] The channel classification device labels the channel type information of the marketing touchpoint. The device needs to accurately classify various types of marketing touchpoints. For example, for marketing advertisements in search engines, the channel classification device will label them as search engine channels according to the search engine platform (such as Baidu, Google, etc.) where the advertisements are placed and the form of the advertisements (such as keyword advertisements, display advertisements, etc.). For marketing content on social media, such as WeChat public account tweets and microblog advertisements, the device will classify them as social media channels according to the type of social media platform (WeChat, microblog, TikTok, etc.) and the way the content is spread. In specific implementation, the channel classification device may have a database of channel types preset inside, which contains the characteristics and classification standards of various common marketing channels. When the contact recording device obtains the interactive data of the marketing touchpoint, it transmits the data to the channel classification device. The device analyzes the characteristics of the touchpoint source, the placement platform, and the content form in the data, and then matches them with the classification standards in the database to determine the channel type information of the marketing touchpoint and add the corresponding label.

[0056] During data collection, the user identification device and the channel classification device need to work in coordination with the touch point recording device and the behavior tracking device. When a user contacts a certain marketing touch point, the touch point recording device will obtain the interaction data of the touch point, and at the same time, the channel classification device will label the channel type of the touch point. The user identification device will record the unique identification information of the user and associate it with the interaction data of the touch point and the subsequent user behavior data. For example, the user clicks on an advertisement of a WeChat public account on a mobile phone. At this time, the touch point recording device will record the interaction data such as the time of clicking and the content of the advertisement, the channel classification device will label the touch point as a WeChat public account type in the social media channel, and the user identification device will generate a unique user identification information by obtaining the device identifier of the mobile phone or the user's WeChat account information, and bind all these data together.

[0057] When the user performs a series of behavior operations in the conversion path, the behavior tracking device will obtain these behavior data, such as the time and content of page browsing, adding to shopping cart, payment, etc. At the same time, the user identification device will ensure that these behavior data and the previous marketing touch point interaction data belong to the same user, and associate them through the unique identification information. The channel classification device will label the channel type of other marketing touch points contacted by the user in the conversion path, so as to analyze each channel in the entire conversion path in the future.

[0058] Through the setting of the user identification device and the channel classification device, accurate identification and classification of users and their contacted marketing touch point channels can be realized, providing a clear and accurate data basis for subsequent path analysis, attribution calculation, etc. The user identification device ensures the consistency of data, so that the user interaction behavior at different times and in different places can be accurately attributed to the same user, avoiding data confusion. The channel classification device makes the channel type of the marketing touch point more clear, which is convenient for subsequent analysis of the role and contribution of different channels in the conversion process.

[0059] In actual application scenarios, for example, in the full-channel marketing activities of e-commerce platforms, users may first search for related goods on search engines, then see advertisements for the goods on social media, and finally complete purchases on the e-commerce platform's APP. In this process, the user identification device will associate all the operation behaviors of the user on the search engine, social media, and e-commerce APP through the user's account information or device identifier, and determine the conversion path as the same user. The channel classification device will respectively mark the search behavior of the search engine as the search engine channel, the advertisement click of the social media as the social media channel, and the purchase behavior of the e-commerce APP as the mobile APP channel. In this way, in subsequent data analysis, the user's conversion path between different channels and the role of each channel in the conversion process can be clearly seen, providing accurate data support for attribution calculation and budget allocation.

[0060] The various devices in the data collection module interact through data transmission interfaces to ensure the real-time and accuracy of the data. After obtaining the unique identification information of the user, the user identification device will immediately transmit this information to the touch point recording device and the behavior tracking device, so that the subsequent collected interaction data and behavior data can be bound to this identification information. After marking the channel type of the marketing touch point, the channel classification device will also transmit the marking results to other devices, so that the channel type can be accurately identified in the data processing process.

[0061] The design of this data collection module can comprehensively and accurately collect interaction data of multi-channel marketing touch points and behavior data of user conversion paths, and through user identification and channel classification, the data is more structured and standardized, providing a solid data foundation for the operation of the entire full-channel marketing funnel intelligent attribution model. Whether it is for complex marketing activities of large enterprises or precise marketing promotion of small and medium-sized enterprises, such a data collection method can meet their needs for user behavior and channel effect analysis, helping enterprises better understand user needs and channel performance, and thus optimizing marketing strategies and budget allocation.

[0062] Embodiment 2:

[0063] The path analysis module includes a main analysis unit and multiple slave analysis units. The main analysis unit is configured to coordinate the synchronization parameters of the time alignment unit based on the time deviation between the marketing touch interaction data and the user behavior data. Specifically, the time alignment unit first time-stamps the data according to an initial synchronization period, which can be set to a fixed time length, such as 10 minutes or 15 minutes, and can be determined according to the processing capacity of the system and the frequency of data collection in actual application. After the time alignment unit adds time stamps to the marketing touch interaction data and the user behavior data, the main analysis unit starts to analyze the time stamp differences of the data to determine the data delay sub-domain.

[0064] In the process of determining the data delay sub-domain, the main analysis unit calculates the time delay of each data collection device in multiple collection time periods according to the time stamps of the data collected by the device. The multiple collection time periods can be different hours of the day or different date periods in a week, and the specific division method depends on the data collection period and the system analysis requirements. For example, for a data collection device, the main analysis unit calculates the time delay of the device in different time periods, such as morning, afternoon, and evening, i.e. the difference between the actual occurrence time and the collection time of the data. Then, the main analysis unit determines the data delay sub-domain based on the time delay of each data collection device in the multiple collection time periods. The data delay sub-domain refers to the data collection area or time period with significant time delay, such as the time delay of some data collection devices in a particular time period is generally larger, and these devices and time periods will be divided into the data delay sub-domain.

[0065] After determining the data delay sub-domain, the main analysis unit finds out the delay collection sub-domain from the multiple data collection sub-domains, i.e. the sub-domain to which the data collection device with time delay exceeding the preset threshold belongs. The preset threshold can be set according to the real-time requirement of the system for the data, such as 5 minutes or 8 minutes, and when the time delay of the data collection device exceeds the threshold, it will be identified as belonging to the delay collection sub-domain. Then, the main analysis unit adjusts the synchronization period of the data collection device in the delay collection sub-domain to the target synchronization period, and the target synchronization period is smaller than the initial synchronization period. For example, the initial synchronization period is 10 minutes, and the target synchronization period can be adjusted to 5 minutes or shorter. The purpose of this is to improve the real-time performance of data collection by shortening the synchronization period, reduce the time deviation, and make the subsequent data processing more accurately reflect the actual behavior of the user and the interaction of the marketing touch.

[0066] In the path resolution module, each slave analysis unit corresponds to at least one data collection device, and its function is to perform path matching and association operations based on the data obtained by the corresponding data collection device. The correspondence between the slave analysis unit and the data collection device can be fixed, or it can be dynamically adjusted according to the running situation of the system to ensure the efficiency and accuracy of data processing. When the data collection device obtains marketing touch interaction data and user behavior data, it will transmit these data to the corresponding slave analysis unit, and the slave analysis unit will start processing the data to realize path matching and association.

[0067] For example, in actual marketing scenarios, there may be multiple data collection devices working at the same time, each responsible for collecting data of different marketing channels or different user groups. The master analysis unit will analyze the time deviation of the data collected by these devices, find out the devices with delay, i.e. the devices belonging to the delayed collection sub-domain, and adjust their synchronization period. The corresponding slave analysis unit will process the data transmitted by these devices and perform path matching and association operations.

[0068] Suppose a data collection device is responsible for collecting marketing touch interaction data of social media channels and user behavior data in that channel. Due to network transmission and other reasons, the data collected by this device in the afternoon has a large time delay, which exceeds the preset threshold and is divided into the delayed collection sub-domain. At this time, the master analysis unit will adjust the synchronization period of this device from the initial 10 minutes to 5 minutes to speed up the data collection frequency and reduce the time delay. When the corresponding slave analysis unit receives the data collected by this device after adjusting the synchronization period, it will process these data and perform path matching and association operations.

[0069] When the slave analysis unit performs path matching and association operations, it will first process the data obtained by the corresponding data collection device in multiple time periods to ensure the accuracy and effectiveness of the data and provide reliable data support for subsequent path matching and association. In this process, the master analysis unit adjusts the synchronization parameters of the time alignment unit to make the data collected by different data collection devices more consistent in time, creating conditions for the slave analysis unit to accurately perform path matching and association operations. Multiple slave analysis units processing data from different data collection devices at the same time can improve the processing efficiency of the entire path resolution module and ensure that the system can still quickly and accurately complete path resolution work under a large amount of data.

[0070] The path analysis module design containing a master analysis unit and multiple slave analysis units can effectively handle the time synchronization problem of multi-channel and multi-source data. By dynamically adjusting the synchronization period of the data collection device, it reduces the time deviation and improves the time consistency of the data, thereby laying the foundation for the subsequent attribution calculation module to generate accurate conversion attribution weights. At the same time, the parallel processing mode of multiple slave analysis units also improves the system's processing capacity for large-scale data, making the entire full-channel marketing funnel intelligent attribution model better adapt to complex marketing scenarios and large data processing needs.

[0071] In practical applications, this structural design of the path analysis module can be applied to various industries and marketing scenarios. For example, in the e-commerce industry, users may conduct shopping behavior through multiple channels such as search engines, social media, e-commerce platforms, etc., and each channel has corresponding data collection devices to collect data. The master analysis unit of the path analysis module will perform time synchronization processing on the data collected by these devices, adjusting the synchronization period of the devices with delays, while the slave analysis units will perform path matching and association on the processed data, constructing the user's complete conversion path, and providing accurate data support for e-commerce enterprises to analyze the conversion effect of each channel.

[0072] In this embodiment, the master analysis unit and multiple slave analysis units of the path analysis module work together to achieve time synchronization processing and path matching association of marketing touch interaction data and user behavior data. The specific workflow and processing method can ensure the time consistency of the data and the accuracy of the path analysis, providing important support for the effective operation of the entire intelligent attribution model.

[0073] Embodiment 3:

[0074] Before the path matching correlation operation is performed by the analysis unit based on the data obtained by the corresponding data collection device, interference filtering needs to be performed on the marketing touch interaction data and user behavior data obtained by the device at multiple time periods. Specifically, the analysis unit will perform feature extraction on the data collected by the data collection device at different time periods (such as early, mid, and late periods in a day, weekdays and weekend periods in a week, etc.). For marketing touch interaction data, the extracted touch type features include the distribution of different interaction behaviors of the user on the touch, such as clicking, browsing, collecting, downloading, etc.; the interaction frequency feature is the number of interactions between the user and the marketing touch within a unit time, such as the number of clicks per hour, the browsing frequency per day, etc. As for user behavior data, the channel coverage feature refers to the number and type of different marketing channels that the user contacts in the conversion path, such as whether the user has contacted search engines, social media, email marketing, etc.; the path length feature is the number of behavior steps included in the conversion path, such as the number of page jumps and operation steps experienced from the first click on the advertisement to the final purchase.

[0075] After extracting the above features, the analysis unit performs interference filtering on the data through a multi-source denoising model. The multi-source denoising model can be a machine learning algorithm model or a preset rule filtering system. Taking the machine learning model as an example, the model takes the touch type feature, the interaction frequency feature, the channel coverage feature, and the path length feature as input, and uses the trained model parameters to identify abnormal values, repeated data, or interference data irrelevant to the conversion path in the data. For example, if the interaction frequency of a certain marketing touch interaction data is abnormally high within a short period of time, and the corresponding channel coverage feature does not match the path length feature in the user behavior data, the model may determine it as interference data and filter it out; for example, if the channel coverage feature of a certain step in the user behavior data does not match the channel type of the actual marketing touch, it may also be identified as interference information and excluded by the model.

[0076] After interference filtering is completed, the analysis unit extracts the touch feature vector and the behavior feature vector based on the filtered data. The touch feature vector integrates various key information of the touch, including touch type, interaction time, channel type, interaction duration, etc. Each dimension of the feature is represented in the form of a numerical value or a classification label. The behavior feature vector includes user behavior type, time sequence, path node, operation duration, etc. The same is presented in a structured manner. For example, the touch feature vector may be represented as [click, 14:30, search engine, 30 seconds], and the behavior feature vector may be represented as [add to shopping cart, 14:35, e-commerce platform page, 20 seconds]. These vectors can more clearly reflect the essential features of the data through abstract processing of the data.

[0077] When the matching degree of the touch point feature vector and the behavior feature vector is greater than a preset matching degree threshold, the association path is determined based on the channel positioning information of the data collection device from the analysis unit. The calculation of the matching degree can be realized by common vector similarity algorithms such as cosine similarity, Euclidean distance, etc. The preset matching degree threshold (such as 0.7 or 0.8) is set according to actual business requirements and data characteristics. For example, if the touch point feature vector corresponds to the click behavior of the search engine channel, the behavior feature vector corresponds to the browsing behavior of the e-commerce platform after half an hour, and the feature vector matching degree of the two exceeds the threshold, and the channel positioning information of the data collection device shows that the touch point belongs to the search engine channel, then the analysis unit from the analysis unit considers that there is an association between the two data, and determines the association path from the search engine click to the e-commerce platform browsing.

[0078] In actual application scenarios, taking the conversion path of an e-commerce user as an example, the data collection device may collect that the user clicked on an advertisement of a social media channel (touch point interaction data) at 10:00 am, then browsed the product detail page of an e-commerce platform (behavior data) at 10:05, added the shopping cart (behavior data) at 10:10, and completed the payment (behavior data) at 10:15. When processing these data, the analysis unit from the analysis unit first extracts the touch point type feature (click) of the touch point interaction data, the interaction frequency feature (the number of clicks on the advertisement in the morning), and the channel coverage feature (social media and e-commerce platform) and path length feature (4 steps) of the behavior data. After filtering out possible abnormal data (such as repeated click records of the same user within a short period of time) by the multi-source denoising model, the touch point feature vector (click, 10:00, social media, 15 seconds) and each behavior feature vector (browse, 10:05, e-commerce platform, 30 seconds; add shopping cart, 10:10, e-commerce platform, 25 seconds; payment, 10:15, e-commerce platform, 40 seconds) are extracted. Then the matching degree of the touch point feature vector and each behavior feature vector is calculated, and when it is found that the matching degree of the click behavior and the browsing behavior exceeds the threshold, the association path of “social media click→e-commerce platform browse” is determined in combination with the positioning information of the data collection device to the social media channel, and the subsequent behavior path association is determined in turn by matching degree calculation, and finally a complete conversion path is constructed.

[0079] From the parsing unit in the execution of the above operation, the need to maintain real-time data interaction with the data acquisition device, to ensure that the latest acquisition data. At the same time, multiple slave analysis units can handle different data acquisition device data in parallel to improve overall processing efficiency. For example, when there are multiple data acquisition devices in the system, respectively responsible for searching engine, social media, email marketing and other data collection of different channels, each slave analysis unit processes the data of one or more devices, through the respective interference filtering, feature extraction and matching correlation process, respectively constructing the user conversion path segment of each channel, and finally integrating into a complete full-channel conversion path by the master parsing unit or other modules.

[0080] The working mechanism of such slave parsing unit can effectively exclude interference factors in the data, ensure the accuracy of the correlation path, and thus provide reliable path data for the subsequent attribution calculation module. Through the matching calculation of feature vectors, the correlation degree between different data can be quantified, avoiding the error caused by subjective judgment, making the path matching and correlation operation more objective and operable. In the complex full-channel marketing environment, the user's conversion path often involves multiple channels and a large number of interaction behaviors, and the processing mode of the slave parsing unit can accurately identify the correlation between each channel touch point and user behavior, providing key path basis for enterprise analysis of each channel conversion contribution.

[0081] In addition, the interference filtering process of the slave parsing unit is not completed at one time, but continues to be carried out with the continuous collection and update of data, so as to adapt to the dynamic changes of the marketing scene. For example, when a new activity form is launched in a certain marketing channel, the touch point type feature may change, and the multi-source denoising model can adjust the filtering rules through regular update or real-time learning, to ensure the effectiveness of interference filtering. Similarly, the preset matching degree threshold can also be adjusted according to the actual business effect, to optimize the precision and recall rate of path matching and correlation.

[0082] The slave parsing unit in the embodiment realizes the path matching and correlation operation based on the data of the data acquisition device through interference filtering, feature extraction, vector matching and other steps, and the specific implementation mode can effectively process the interference information in multi-source data, accurately construct the correlation path between marketing touch points and user behaviors, provide solid path data support for attribution analysis of full-channel marketing funnel, and enable enterprises to more clearly understand the channel influence mechanism in the user conversion path, providing an important basis for subsequent budget allocation optimization.

[0083] Embodiment 4:

[0084] The path analysis module includes a computing power scheduling unit, which has the core function of adjusting the real-time correspondence between multiple slave analysis units and multiple data collection devices based on the synchronization period of each data collection device. Specifically, when the master analysis unit determines the delay collection sub-domain, the computing power scheduling unit will identify the data collection devices in this sub-domain as high-load collection devices. This is because the synchronization period of these devices is adjusted to a shorter target synchronization period (such as from 10 minutes to 5 minutes), resulting in an increase in the amount of data collected per unit time and an increase in data processing load. The slave analysis units corresponding to the high-load collection devices will be listed as slave analysis units to be adjusted at this time, and their processing tasks need to be redistributed.

[0085] Taking an actual scenario as an example, assume that in the full-channel marketing system of an e-commerce enterprise, there are 10 data collection devices, each responsible for collecting data from different channels such as search engines, social media, emails, and mobile APPs. Among them, 3 data collection devices responsible for social media channels are divided into the delay collection sub-domain due to network delay and other reasons, and the synchronization period is adjusted from the initial 10 minutes to 5 minutes. These 3 devices become high-load collection devices, and the 2 slave analysis units corresponding to them originally responsible for processing the data of these 3 devices now have their processing load significantly increased due to the doubling of data volume, which may result in processing delays or resource shortages.

[0086] At this time, the computing power scheduling unit will start the adjustment mechanism. First, it will identify the 3 high-load collection devices and the 2 slave analysis units to be adjusted corresponding to them. Then, the computing power scheduling unit will analyze the load of other slave analysis units in the system. Assume that there are 3 slave analysis units in the system responsible for processing data from the search engine channel, and the current load is low (because the data collection synchronization period of the search engine channel has not been adjusted, the data volume is stable). The computing power scheduling unit will redistribute 1 or 2 of the high-load collection devices to the slave analysis units with lower load. For example, it will assign 1 data collection device from the social media channel to a slave analysis unit responsible for the search engine channel, so that the 2 slave analysis units originally processing 3 devices will be reduced to processing 2 devices, and the newly added slave analysis unit will share the data processing task of 1 device.

[0087] During the adjustment process, the computing power scheduling unit will ensure the continuity and accuracy of data processing. It will record the historical data processing of each data collection device and synchronize the historical data processing context of the device to the new slave analysis unit when it is redistributed, avoiding data processing interruption or errors caused by task transfer. For example, when a social media data collection device is redistributed to a new slave analysis unit, the computing power scheduling unit will transfer the device's previous time synchronization parameters, processed data range, and other information to the new slave analysis unit, allowing it to seamlessly continue subsequent data processing work.

[0088] In addition, the adjustment of the computing power scheduling unit is not a one-time operation, but a dynamic optimization according to real-time load conditions. For example, during the e-commerce promotion period, the data collection volume of the mobile APP channel may suddenly increase, causing the data collection device of this channel to be shortened in synchronization period and become a new high-load collection device. At this time, the computing power scheduling unit will re-evaluate the load of all slave analysis units and allocate part of the tasks to other slave analysis units with lower load, such as the slave analysis unit corresponding to the email channel (if its current data volume is small).

[0089] In another example, assume that in the marketing system of a certain bank, data collection device A (responsible for the SMS marketing channel) and device B (responsible for the APP push channel) become high-load collection devices due to synchronization period adjustment, and the corresponding slave analysis units C and D have too high load. The computing power scheduling unit finds that slave analysis unit E (responsible for the web advertisement channel) is currently processing a small amount of data, and the synchronization period of this channel has not been adjusted. At this time, the computing power scheduling unit will allocate part of the data processing tasks of device A to slave analysis unit E, and adjust the task range of slave analysis unit C so that it only processes the remaining data of device A and the data of device B. In this way, the load of slave analysis units C and D is relieved, the resources of slave analysis unit E are fully utilized, and the computing power resources of the entire system achieve dynamic balance.

[0090] When adjusting the corresponding relationship, the computing power scheduling unit also considers the processing capacity differences of the slave analysis units. Different slave analysis units may have different data processing efficiencies due to factors such as hardware configuration and software version. The computing power scheduling unit will set a processing capacity parameter for each slave analysis unit based on historical performance data, and when allocating tasks, it will preferentially allocate the data of high-load collection devices to slave analysis units with higher processing capacity, in order to improve overall processing efficiency. For example, if the processing capacity of slave analysis unit F is 30% higher than that of slave analysis unit G, when a new high-load collection device needs to be allocated, the computing power scheduling unit will preferentially allocate it to slave analysis unit F to ensure that the data can be processed quickly.

[0091] The advantage of this computing power scheduling mechanism is that it can adapt to the dynamic changes in data volume in the full-channel marketing scenario. When the data collection volume of certain channels suddenly increases due to marketing activities, changes in user behavior, etc., the computing power scheduling unit can timely adjust the task allocation of the slave analysis units to avoid processing delays caused by excessive load in some slave analysis units, and ensure the overall operation efficiency of the path analysis module. At the same time, it can also avoid the idle resources of some slave analysis units due to insufficient data volume, and improve the resource utilization of the system.

[0092] In practical applications, the computing power scheduling unit needs to maintain real-time communication with the master analysis unit and the slave analysis unit. After determining the delay collection sub-domain and adjusting the synchronization period of the data collection device, the master analysis unit will transmit the relevant information to the computing power scheduling unit in real time, so that the computing power scheduling unit can timely start the adjustment mechanism. The slave analysis unit will feed back its load situation (such as the current data volume being processed, the remaining processing capacity, etc.) to the computing power scheduling unit in real time, providing a basis for task allocation by the computing power scheduling unit.

[0093] For example, the master analysis unit adjusts the synchronization period of data collection device X from 15 minutes to 8 minutes at a certain time, making it a high-load collection device. The master analysis unit transmits this information to the computing power scheduling unit, which immediately queries the load situation of slave analysis unit Y corresponding to device X and finds that it is already in a high-load state. At this time, the computing power scheduling unit will scan the load states of other slave analysis units and find a slave analysis unit Z with lower load, and allocate part of the data processing tasks of device X to slave analysis unit Z, while updating the task lists of slave analysis units Y and Z.

[0094] The real-time adjustment capability of the computing power scheduling unit enables the path analysis module to maintain stable and efficient operation in a complex full-channel marketing data environment. Whether it is daily marketing data processing or dealing with the surge of data brought by sudden large-scale marketing activities, the computing power scheduling unit can dynamically adjust the real-time correspondence between the slave analysis unit and the data collection device, ensuring the timeliness and accuracy of data processing, and providing reliable support for subsequent attribution calculation and budget allocation.

[0095] In this embodiment, the computing power scheduling unit adjusts the real-time correspondence between the slave analysis unit and the data collection device based on the synchronization period of the data collection device, and realizes the optimal allocation of system computing power resources through steps such as identifying high-load collection devices, evaluating slave analysis unit load, and dynamically allocating processing tasks. This implementation can effectively cope with the dynamic changes in data volume in the full-channel marketing scenario, avoid waste and overload of computing power resources, and ensure the efficient operation of the path analysis module, thereby providing important support for the stable operation of the entire full-channel marketing funnel intelligent attribution model.

[0096] Embodiment 5:

[0097] The full-channel marketing funnel automatic budget allocation system includes a model application module and a budget execution module. The function of the model application module is to call the full-channel marketing funnel intelligent attribution model and output the conversion attribution result, and the budget execution module allocates marketing budgets to each marketing channel based on the dynamic mapping relationship.

[0098] Taking the marketing scenario of a certain clothing e-commerce enterprise as an example, the enterprise carries out marketing activities in multiple channels, including search engine advertising, social media promotion, email marketing, and push of its own APP. The model application module will first access the various types of data obtained by the data collection module, such as user keyword search and click data on search engines, advertising interaction data on social media, opening email records, and browsing, adding, and purchasing behavior data within the APP.

[0099] After the model application module calls the intelligent attribution model, the data will be processed by the data collection module, the path analysis module, the attribution calculation module, and the budget mapping module in turn. The touch record device and the behavior tracking device in the data collection module obtain marketing touch interaction data and user conversion path behavior data, respectively, and the user identification device and the channel classification device identify and classify the data to ensure that the data can be accurately matched to each user and the corresponding channel.

[0100] The main analysis unit and the slave analysis unit of the path analysis module will synchronize the data in time and match and associate the paths. Suppose a user's conversion path is: first search for "autumn women's clothing" on the search engine and click on the enterprise's advertisement, then see the enterprise's new product push on social media and like it, then view the coupon information in the opened email, and finally complete the purchase in the APP using the coupon. The path analysis module will synchronize and process these touch interaction data and behavior data from different channels in chronological order to match and associate them into a complete conversion path.

[0101] The attribution calculation module generates the conversion attribution weight of each marketing channel based on the path-matched and associated data. For example, through a certain algorithm, it is calculated that the attribution weight of the search engine channel is 40%, the social media channel is 25%, the email channel is 20%, and the APP push channel is 15%. The model application module outputs these conversion attribution results, i.e., the conversion attribution weight of each channel.

[0102] The budget execution module allocates marketing budgets based on the dynamic mapping relationship established by the budget mapping module. Suppose the enterprise's total marketing budget for this month is 1 million yuan, and the budget mapping module will determine the allocation proportion according to the conversion attribution weight of each channel. According to the above weights, the search engine channel will get 400,000 yuan (1 million x 40%), the social media channel 250,000 yuan, the email channel 200,000 yuan, and the APP push channel 150,000 yuan.

[0103] The budget execution module allocates the budget by investing funds into specific marketing activities for each channel. For example, in the search engine channel, 40,000 yuan can be used to purchase keyword ads, optimize search rankings, etc.; in the social media channel, 25,000 yuan can be used to launch information flow ads, cooperate with net red promotion, etc.; 20,000 yuan in the email channel can be used to design and send personalized coupon emails, maintain customer relationships; and 15,000 yuan in the APP push channel can be used to optimize the display of ads in the APP, push accurate product recommendations, etc.

[0104] During the budget execution process, the system monitors the budget usage and conversion effect of each channel in real time. For example, when the budget of the search engine channel is used up to 30,000 yuan, the model application module recalculates the conversion attribution weight of this channel according to the newly collected data. Assuming that due to the recent increase in click-through rate and conversion rate of search engine ads, its attribution weight rises to 45%, and the weights of other channels are adjusted accordingly. At this time, the budget execution module will adjust the allocation of the remaining budget according to the new dynamic mapping relationship. The remaining total budget is 100-30 (search engine has been used)-15 (social media has been used)-10 (email has been used)-8 (APP has been used)=37 million yuan. According to the new weight, the search engine channel can obtain 37 million x 45% ≈ 16.65 million yuan, the social media channel obtains 37 million x 23% ≈ 8.51 million yuan, the email channel obtains 37 million x 19% ≈ 7.03 million yuan, and the APP push channel obtains 37 million x 13% ≈ 4.81 million yuan.

[0105] The budget execution module will reallocate the remaining budget to each channel according to the adjusted allocation scheme, ensuring that the budget allocation can reflect the conversion contribution of each channel in real time. For example, the search engine channel will use the additional 16.65 million yuan to increase the intensity of keyword ad placement, and the social media channel will use the 8.51 million yuan to increase new ad creative testing, etc.

[0106] Another example is an education and training enterprise, whose marketing channels include Baidu promotion, WeChat public number promotion, Douyin short video ads, and offline activities. When allocating the budget at the beginning of the month, the model application module calculates the conversion attribution weight of each channel as Baidu promotion 35%, WeChat public number 25%, Douyin 30%, and offline activities 10%, with a total budget of 500,000 yuan. The budget execution module allocates 17.5 million yuan to Baidu promotion, 12.5 million yuan to WeChat public number, 15 million yuan to Douyin, and 5 million yuan to offline activities.

[0107] In the middle of the month, the company launched a new series of short video ads on Douyin, attracting a large number of users to inquire about enrollment, and the conversion attribution weight of the Douyin channel rose to 35%, while the Baidu promotion dropped to 30%, the WeChat public number to 20%, and the offline activity to 15%. At this time, the total budget has been used 300,000 yuan, leaving 200,000 yuan. The budget execution module reallocates the remaining budget according to the new weight, with Baidu promotion obtaining 20 million x 30% = 6 million yuan, WeChat public number 20 million x 20% = 4 million yuan, Douyin 20 million x 35% = 7 million yuan, and offline activity 20 million x 15% = 3 million yuan. The budget execution module will use this part of the funds for the expansion of the new ad series on Douyin, the expansion of offline activities, etc.

[0108] Through the collaborative work of the model application module and the budget execution module, the enterprise can dynamically adjust the budget allocation according to the actual conversion contribution of each marketing channel, making the budget resources more effectively utilized. The model application module continuously calls the intelligent attribution model, constantly updating the conversion attribution results, providing accurate basis for budget allocation. The budget execution module adjusts the budget allocation scheme in real time according to these bases, and accurately invests funds in channels with good conversion effect, improving the return on marketing investment.

[0109] In actual application, frequent data interaction is needed between the model application module and the budget execution module. The conversion attribution results output by the model application module are transmitted to the budget execution module in real time, and the budget execution module calculates and adjusts the budget allocation amount of each channel according to these results. At the same time, the budget execution module also feeds back the budget usage of each channel to the model application module, so that the model can consider the influence of budget input on conversion effect when calculating attribution weight.

[0110] For example, when a channel increases its budget input, its conversion volume may increase accordingly. The model application module will reflect this change when calculating the attribution weight next time, thereby affecting the budget allocation of the budget execution module to that channel. This two-way data interaction ensures that the budget allocation system can continuously optimize according to the actual situation, forming a closed-loop dynamic adjustment mechanism.

[0111] The all-channel marketing funnel automatic budget allocation system in this embodiment realizes dynamic budget allocation based on conversion attribution weight through the cooperation of the model application module and the budget execution module. Through specific examples, it can be seen that this system can adjust the budget allocation in real time according to the actual conversion contribution of each channel, making the enterprise's marketing budget more reasonably utilized and improving the effectiveness and return on investment of marketing activities.

[0112] It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting; it is not intended to exclude myriad other embodiments of the present application that other inventors can develop based on the same general inventive concepts embodied by the described embodiments. That is, although the present application is described in terms of particular embodiments and illustrative figures, it should be apparent that the scope of the present application is not limited to these specific embodiments.

[0113] While the embodiments of the application have been shown and described herein, it will be understood by those skilled in the art that many changes, modifications, substitutions and alterations to these embodiments can be made without departing from the principles and spirits of the application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. Omni-channel marketing funnel intelligent attribution model, characterized by: include: A data collection module, comprising a contact recording device deployed on a marketing terminal and a behavior tracking device provided on the user side, wherein the contact recording device is used to obtain interaction data of multi-channel marketing contacts, and the behavior tracking device is used to obtain behavior data of user conversion paths; A path parsing module, comprising a time alignment unit and a path association unit. The time alignment unit is used to perform time-series synchronization processing on the marketing touchpoint interaction data and the user behavior data. The path association unit is used to perform path matching and association on the time-series synchronized marketing touchpoint interaction data and the user behavior data. The attribution calculation module is used to generate conversion attribution weights for each marketing channel based on marketing touchpoint interaction data and user behavior data after path matching association; The budget mapping module is used to establish a dynamic mapping relationship between conversion attribution weights and marketing budgets, and update the dynamic mapping relationship based on the conversion attribution weights.

2. The omni-channel marketing funnel intelligent attribution model according to claim 1 is characterized in that: The data collection module also includes a user identification device provided on the marketing terminal and a channel classification device provided on the user side, wherein the user identification device is used to record the user's unique identification information, and the channel classification device is used to mark the channel type information of the marketing touchpoints.

3. The omni-channel marketing funnel intelligent attribution model according to claim 2 is characterized in that: The path parsing module includes a master parsing unit and multiple slave parsing units, wherein the master parsing unit is used to coordinate the synchronization parameters of the time alignment unit based on the time deviation between the marketing contact interaction data and the user behavior data; Each of the slave parsing units corresponds to at least one data acquisition device, and the slave parsing unit is used to perform a path matching association operation based on the data acquired by the corresponding data acquisition device.

4. The omni-channel marketing funnel intelligent attribution model according to claim 3 is characterized in that: The main parsing unit coordinates the synchronization parameters of the time alignment unit based on the time deviation between the marketing contact interaction data and the user behavior data, including: The time alignment unit performs time stamping on the marketing contact interaction data and the user behavior data based on the initial synchronization period; Determine the data delay subdomain based on the timestamp differences between marketing touchpoint interaction data and user behavior data; determining a delayed acquisition subfield from a plurality of data acquisition subfields based on the data delay subfield; The synchronization period of the data acquisition device in the delayed acquisition subdomain is adjusted to a target synchronization period, wherein the target synchronization period is smaller than the initial synchronization period.

5. The omni-channel marketing funnel intelligent attribution model according to claim 4 is characterized in that: The main parsing unit determines the data delay subdomain based on the timestamp difference between the marketing touchpoint interaction data and the user behavior data, including: For each of the data acquisition devices, determining a time delay of the data acquisition device in a plurality of acquisition time periods based on a timestamp of data acquired by the data acquisition device; The data delay sub-range is determined based on the time delay amount of each of the data acquisition devices in multiple acquisition time periods.

6. The omni-channel marketing funnel intelligent attribution model according to claim 4 is characterized in that: The slave parsing unit performs a path matching association operation based on the data acquired by the corresponding data acquisition device, including: For each of the data acquisition devices, the secondary parsing unit corresponding to the data acquisition device performs interference filtering on the marketing contact interaction data and user behavior data acquired by the data acquisition device in multiple time periods, and the secondary parsing unit corresponding to the data acquisition device extracts a contact feature vector and a behavior feature vector based on the filtered marketing contact interaction data and user behavior data; When the matching degree between the contact feature vector and the behavior feature vector is greater than a preset matching degree threshold, the association path between the marketing contact interaction data and the user behavior data is determined based on the channel positioning information of the data acquisition device, and the path matching association operation is completed based on the association path.

7. The omni-channel marketing funnel intelligent attribution model according to claim 6, characterized in that: The secondary parsing unit performs interference filtering on the marketing contact interaction data and user behavior data acquired by the data acquisition device in multiple time periods, including: Extracting contact type features and interaction frequency features based on the marketing contact interaction data acquired by the data acquisition device in multiple time periods; Extracting channel coverage features and path length features based on user behavior data acquired by the data acquisition device over multiple time periods; The marketing contact interaction data and user behavior data are interfered with and filtered out by a multi-source denoising model based on the contact type characteristics, interaction frequency characteristics, channel coverage characteristics, and path length characteristics.

8. The omni-channel marketing funnel intelligent attribution model according to any one of claims 4 to 7, characterized in that: The path parsing module further includes a computing power scheduling unit, which is used to adjust the real-time correspondence between the multiple slave parsing units and the multiple data acquisition devices based on the synchronization period of each of the data acquisition devices.

9. The omni-channel marketing funnel intelligent attribution model according to claim 8, characterized in that: The computing power scheduling unit adjusts the real-time correspondence between the plurality of slave parsing units and the plurality of data acquisition devices based on the synchronization period of each of the data acquisition devices, including: The data acquisition device in the delayed acquisition subdomain is used as a high-load acquisition device, and the slave parsing unit corresponding to the high-load acquisition device is used as the slave parsing unit to be adjusted; Based on the high-load acquisition device and the slave parsing unit to be adjusted, the real-time correspondence between the plurality of slave parsing units and the plurality of data acquisition devices is adjusted.

10. An omni-channel marketing funnel automated budget allocation system, characterized by: The system includes a model application module and a budget execution module, wherein the model application module is used to call the omni-channel marketing funnel intelligent attribution model described in claims 1-9 and output the conversion attribution results, and the budget execution module is used to allocate the marketing budget to each marketing channel based on the dynamic mapping relationship.

Images