Customer acquisition tracking method and system based on multi-channel fusion and computer device

Abstract

The present application belongs to the technical field of data analysis, and particularly relates to a customer acquisition tracking method and system based on multi-channel fusion and a computer device. The method of the present application comprises: obtaining user behavior data; constructing a connected graph, extracting historical touchpoints of each user in all channels in time sequence to form a behavior trajectory sequence, and identifying a preset key conversion node in the behavior trajectory sequence; when the user behavior matches the key conversion node, calculating a contribution weight for each touchpoint in the sequence; and determining a customer acquisition tracking result of each channel based on the contribution weight of each touchpoint. The present application effectively solves the problems of data isolation, attribution difficulty and evaluation lag generally existing in the existing customer acquisition tracking scheme.

Description

Technical Field

[0001] This invention belongs to the technical field of data analysis, specifically relating to a customer acquisition tracking method, system, and computer equipment based on multi-channel integration. Background Technology

[0002] Currently, when companies engage in multi-channel marketing to acquire customers, they typically utilize multiple channels simultaneously, including search engine advertising, social media promotion, e-commerce platform placements, offline store events, and trade show marketing. To evaluate the suitability of each channel, companies usually track customer acquisition based on the characteristics of each channel in order to better expand their customer base.

[0003] However, due to the different data collection methods, user identification systems, data formats, and reporting times of various channels, user behavior data from different channels are isolated from each other. For example, the same user may browse products on a mobile app, register on a PC website, consult customer service in a WeChat mini program, or complete a purchase in an offline store. However, because different channels use different user identifiers (such as CookieID, OpenID, mobile phone number, device ID, etc.), enterprises cannot obtain a cross-channel, full-cycle view of customer behavior, making it difficult to analyze the complete path of users from their first contact to their final conversion.

[0004] Meanwhile, customer conversion paths typically involve multiple touchpoints (e.g., search engine ad clicks → website visits → customer service inquiries → trial applications → email follow-ups → order payments). Traditional attribution methods, such as "last click attribution," attribute all conversion value to the last touchpoint, ignoring the "supporting" role of earlier touchpoints; conversely, "first click attribution" ignores the facilitating value of later touchpoints.

[0005] In addition, existing ROI (Return on Investment) analysis mostly adopts offline batch processing mode, usually summarizing and calculating data on a daily (T+1) or longer time granularity. This makes it impossible for enterprises to monitor changes in customer acquisition costs, conversion rates and ROI of each channel in real time, and makes it difficult to detect abnormal channel performance in a timely manner (such as sudden increase in costs or sharp drop in conversion rates). Summary of the Invention

[0006] To address the aforementioned problems in existing technologies, namely data isolation, attribution difficulties, and evaluation lag in current customer acquisition tracking solutions, the first aspect of this invention provides a customer acquisition tracking method based on multi-channel integration, comprising:

[0007] Acquire user behavior data from multiple channels; A connected graph is constructed based on the user behavior data, and the identity identifiers belonging to the same connected component in the connected graph are mapped to unique global user identifiers. Using the global user identifier as an index, historical touchpoints of each user across all channels are extracted in chronological order to form a behavioral trajectory sequence, and key conversion nodes preset in the behavioral trajectory sequence are identified. When user behavior matches the key conversion node, the behavior trajectory sequence is traced back, and a contribution weight is calculated for each touchpoint in the sequence. The contribution weight is determined based on the touchpoint's time decay factor, position factor, and channel quality factor. The time decay factor is negatively correlated with the interval between the touchpoint occurrence time and the conversion completion time. The position factor is set such that the weight of the first and last touchpoints in the behavior trajectory sequence is greater than that of the middle touchpoints. The channel quality factor is dynamically adjusted based on the historical conversion rate of the corresponding channel. Based on the contribution weight of each touchpoint, the customer acquisition tracking results for each channel are determined.

[0008] In some preferred embodiments, the method further includes, before constructing the connectivity graph based on the user behavior data: The user behavior data is standardized, deduplicated, and quality checked. Extract the identity identifier fields reported from each channel. The identity identifier includes at least one of the following: phone number hash value, email address hash value, device ID, and IP address. The identity identifiers are clustered using the DBSCAN density clustering algorithm, and noise points are filtered out. The similarity between different identity identifiers after clustering is calculated, and the similarity is calculated using cosine similarity, edit distance or Jaccard coefficient. Based on the similarity, a confidence assessment is performed, and identities belonging to the same entity are merged to generate nodes for constructing a connected graph.

[0009] In some preferred embodiments, constructing the connected graph includes: Using identity identifiers from different entities as nodes, and the associations between identifiers caused by the same event or the same entity as edges, when different identity identifier nodes form a connected component, all identity identifiers within the connected component are mapped to the same global user identifier.

[0010] In some preferred embodiments, the time decay factor is calculated using the following formula: ; in, The time interval between the occurrence time of contact Ti and the conversion time is the preset attenuation coefficient; The formula for calculating the position factor is: ; Where n is the total length of the contact sequence; The formula for calculating the contribution weight is as follows: ; The attribution revenue allocated to each channel is calculated using the following formula: ; ; in, The time decay factor, For position factors, For channel quality factors, The overall weight of a single touch point. The contribution ratio of each touchpoint The attribution revenue allocated to channel c.

[0011] In some preferred embodiments, the key conversion nodes include at least lead nodes, intention nodes, and transaction nodes; Among them, the clue node corresponds to the registration success or form submission event, the intention node corresponds to the add to cart or request demo event, and the transaction node corresponds to the payment success or contract signing event; The time point that triggers the calculation of the contribution weight is when the user behavior matches the transaction node.

[0012] In some preferred embodiments, the method further includes: The conversion value corresponding to the key conversion nodes is allocated to each channel according to the contribution weight of each touchpoint. The return on investment of each channel is calculated based on the attribution revenue allocated to each channel and the real-time investment cost obtained. Decision recommendations are generated based on the assessment results of the aforementioned rate of return on investment.

[0013] In some preferred embodiments, calculating the return on investment for each channel includes: The total cost of current channels is aggregated, including advertising expenditure costs, personnel costs, platform costs, and other costs. The calculation formula is as follows: ; in, The current advertising cost for these channels, For personnel costs, For platform costs, Other costs; The formula for aggregating direct conversion revenue and attributable revenue from current channels is as follows: ; in, For the direct conversion revenue of the current channels, This is the primary revenue attributed to the current channel. The formula for calculating the return on investment (ROI) of the current channel is as follows: .

[0014] In some preferred embodiments, the method further includes: Determine the ROI values ​​of the current channel at the most recent N time points, and use the least squares method to perform linear regression analysis. Determine whether the current channel is in an upward or downward trend based on the sign of the slope, where N is a positive integer greater than 0. If the return on investment is higher than the preset threshold and the trend is upward, then it is recommended to increase the budget; if the conversion rate is lower than the preset threshold, then it is recommended to optimize the landing page or suspend the campaign. Record the changes in indicators after the strategy is executed, adjust the preset threshold through reinforcement learning feedback, and after generating a global user identifier, parse the UTM parameters, reference page or channel-specific QR code identifier in the collected data, bind the parsed channel identifier with the current global user identifier, and record the associated valid time period to determine the channel affiliation of the user within a specific time period.

[0015] In a second aspect, the present invention proposes a customer acquisition tracking system based on multi-channel integration, characterized in that it includes: The data acquisition module is configured to acquire user behavior data from multiple channels; The data mapping module is configured to construct a connected graph based on the user behavior data, and map the identity identifiers belonging to the same connected component in the connected graph to a unique global user identifier; The data extraction module is configured to use the global user identifier as an index to extract the historical touchpoints of each user across all channels in chronological order to form a behavior trajectory sequence, and to identify the preset key conversion nodes in the behavior trajectory sequence; The weight calculation module is configured to backtrack the behavior trajectory sequence when the user behavior matches the key conversion node, and calculate the contribution weight for each touchpoint in the sequence. The contribution weight is determined based on the time decay factor, position factor and channel quality factor of the touchpoint. The time decay factor is negatively correlated with the interval between the time of touchpoint occurrence and the time of conversion completion. The position factor is set such that the weight of the first and last touchpoints in the behavior trajectory sequence is greater than that of the middle touchpoints. The channel quality factor is dynamically adjusted based on the historical conversion rate of the corresponding channel. The data output module is configured to determine the customer acquisition tracking results for each channel based on the contribution weight of each touchpoint.

[0016] In a third aspect, the present invention provides a computer device comprising: At least one processor; and a memory communicatively connected to at least one of the processors; The memory stores instructions that can be executed by the processor to implement the multi-channel fusion-based customer acquisition tracking method as described in the first aspect.

[0017] The beneficial effects of this invention are: On one hand, this invention maps identity identifiers belonging to the same connected component to a unique global user identifier by constructing a connected graph. Before constructing the connected graph, user behavior data is standardized, deduplicated, and quality checked. Identity identifiers are fused using hashing, DBSCAN density clustering, and multi-dimensional similarity calculation (cosine similarity, edit distance, or Jaccard coefficient). This enables fragmented identity identifiers from different channels to be accurately associated with the same user, thereby eliminating data silos and forming a unified customer view across channels.

[0018] On the other hand, this invention uses a combination of time decay factor, location factor and channel quality factor to calculate the contribution weight of each touchpoint. The time decay factor uses an exponential decay function, the location factor gives greater weight to the first and last touchpoints, and the channel quality factor is dynamically adjusted according to historical conversion rates. This approach can more comprehensively evaluate the true contribution of each touchpoint in the customer behavior trajectory, avoid the underestimation of the value of early touchpoints by traditional attribution methods, and make the channel contribution assessment more scientific and reasonable.

[0019] Furthermore, this invention calculates ROI by aggregating the full-link costs and attributable benefits of each channel in real time, and uses a sliding time window combined with the 3-Sigma principle for anomaly detection. It can identify abnormal changes in conversion rate or ROI within a small delay, realize real-time performance evaluation and alerts, and support timely adjustments to marketing strategies. Attached Figure Description

[0020] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart illustrating a customer acquisition and tracking method based on multi-channel integration provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the framework of a customer acquisition tracking system based on multi-channel integration provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a computer system that implements the methods, systems, and computer equipment embodiments of this application. Detailed Implementation

[0021] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0022] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0023] To more clearly explain the customer acquisition tracking method based on multi-channel integration provided by the present invention, the following will be combined with... Figure 1 The steps in the embodiments of the present invention will be described in detail below.

[0024] The first embodiment of the present invention provides a customer acquisition tracking method based on multi-channel integration, including steps S10-S50, each step of which is described in detail below: Step S10: Obtain user behavior data from multiple channels; Step S20: Construct a connected graph based on the user behavior data, and map the identity identifiers belonging to the same connected component in the connected graph to a unique global user identifier; Step S30: Using the global user identifier as an index, extract the historical touchpoints of each user across all channels in chronological order to form a behavior trajectory sequence, and identify the preset key conversion nodes in the behavior trajectory sequence; Step S40: When user behavior matches the key conversion node, backtrack the behavior trajectory sequence and calculate the contribution weight for each touchpoint in the sequence. The contribution weight is determined based on the time decay factor, position factor, and channel quality factor of the touchpoint. The time decay factor is negatively correlated with the interval between the touchpoint occurrence time and the conversion completion time. The position factor is set such that the weight of the first and last touchpoints in the behavior trajectory sequence is greater than that of the middle touchpoints. The channel quality factor is dynamically adjusted based on the historical conversion rate of the corresponding channel. Step S50: Determine the customer acquisition tracking results for each channel based on the contribution weight of each touchpoint.

[0025] The customer acquisition tracking method based on multi-channel integration provided in this embodiment can be applied to scenarios such as e-commerce platforms, financial services, online education, and SaaS services that require evaluating the effectiveness of multi-channel customer acquisition and optimizing marketing decisions.

[0026] In this embodiment, user behavior data is obtained through data collection components pre-deployed across multiple channels.

[0027] Specifically, for online channels, a JavaScript SDK can be embedded on the web, the corresponding mobile SDK can be integrated on mobile devices (iOS / Android), and a log collection agent can be deployed on the server. This collection component captures user behavior events such as clicks, browsing, form submissions, and payments in real time and reports the event data to the data access layer.

[0028] For offline channels, such as physical stores and exhibitions, deploy Wi-Fi probes or QR code scanning devices. When users enter a store or scan a QR code specific to an exhibition, collect information such as the physical address of their device, geographical location, and scanning time, which will also be reported as user behavior data.

[0029] For third-party channels, such as partners, agents, and recommendation systems, user behavior data (such as registration or purchase events brought about by recommendations) is received through API interfaces.

[0030] Generally, to avoid data congestion, a high-concurrency asynchronous processing mechanism can be adopted: configure an independent task executor, set the core thread count and maximum thread count (e.g., 10 core threads, 50 maximum threads), and use a bounded queue for buffering. In this way, all reported behavioral data first enters the asynchronous processing queue, and the main request thread returns immediately, thus supporting the real-time access of large amounts of data.

[0031] Based on this, this example extracts various identity fields from the obtained user behavior data. These identity fields may include: hash values ​​of phone numbers, hash values ​​of email addresses, device IDs (such as IDFA, OAID), CookieID, OpenID (WeChat Mini Programs, etc.), IP addresses, etc.

[0032] Then, these identity identifiers are used as nodes, and the relationships between identifiers arising from the same event or the same entity are used as edges to construct a connected graph. For example, when a user logs in to an app, the device ID is associated with the phone number hash; when a user binds their phone number in a WeChat mini-program, the OpenID is associated with the phone number hash. Through this association, a path is formed between the three nodes: device ID, OpenID, and phone number hash, which belong to the same connected component.

[0033] When a connected component is detected to contain multiple identity nodes, a unique global user identifier (Global_UID) is generated for that connected component. Thereafter, all actions of that user across different channels and devices are associated with this Global_UID.

[0034] Furthermore, after obtaining the Global_UID, this identifier can be used as an index to retrieve all historical touchpoints of the user from the time-series database.

[0035] In this embodiment, a touchpoint refers to an interaction between a user and a marketing channel, such as: clicking an ad, browsing a webpage, contacting customer service, applying for a trial, or making an order payment. These touchpoints are arranged in ascending order of their occurrence time to form a complete sequence of behavioral trajectories.

[0036] As an example, this embodiment presupposes three types of key conversion nodes: Lead Node (L1): Corresponds to events such as successful registration and form submission, marking anonymous visitors as potential leads; Intent node (L2): Corresponds to events such as adding to cart and requesting a demo, marking the confirmation of purchase intention; Transaction Node (L3): This corresponds to events such as successful payment and contract signing, marking the completion of the sales conversion.

[0037] After generating the behavior trajectory sequence, by automatically traversing each event in the sequence and matching it with the aforementioned preset nodes, it is possible to identify which touchpoints belong to key conversion nodes. When user behavior matches a key conversion node, the behavior trajectory sequence is backtracked, and the contribution weight of each touchpoint is calculated. Specifically, this embodiment can be configured with an event stream processing engine (e.g., based on Apache Flink). When the engine detects that a user’s behavior matches a key conversion node (especially a transaction node), it immediately triggers attribution calculation, traces back all touchpoints in the user’s current behavior trajectory sequence, and calculates three factors for each touchpoint Ti: time decay factor, location factor, and channel quality factor.

[0038] The formula for calculating the time decay factor is: ; in, The time interval between the occurrence time of contact Ti and the conversion time is the preset attenuation coefficient; The formula for calculating the position factor is: ; Where n is the total length of the contact sequence; The formula for calculating the contribution weight is as follows: ; The attribution revenue allocated to each channel is calculated using the following formula: ; ; in, The time decay factor, For position factors, For channel quality factors, The overall weight of a single touch point. The contribution ratio of each touchpoint The attribution revenue allocated to channel c.

[0039] As a feasible implementation method, the channel quality factor is dynamically adjusted based on the historical conversion rate of the corresponding channel. The process can be as follows: Suppose there are M marketing channels. For any channel c, define its conversion rate in the t-th time window (e.g., the past 7 days): CVR_c(t) = (Number of transaction nodes brought by this channel) / (Total number of touchpoints or total number of clicks brought by this channel); At the same time, calculate the average conversion rate across all channels within the same time window: CVR_avg(t) = (Σ CVR_c(t)) / M; The quality factor of channel c within the time window t is then: f_qual_c(t) = CVR_c(t) / CVR_avg(t); To avoid the denominator being zero, when CVR_avg(t)=0, f_qual_c(t) is set to 1.

[0040] In addition, to avoid drastic changes in factors due to short-term random fluctuations, an exponentially weighted moving average (EWMA) can be used to smooth the quality factors: f_qual_smoothed(t)=α×f_qual_c(t)+(1-α)×f_qual_smoothed(t-1); Where α is the smoothing coefficient (e.g., 0.3). The initial value of f_qual_smoothed(0) is set to 1.

[0041] Specifically, in a real-world application example, the user's behavioral trajectory sequence includes 5 touchpoints: T1 (clicking a search engine ad), T2 (browsing the official website), T3 (customer service inquiry), T4 (applying for a trial), and T5 (order payment).

[0042] The conversion completion time corresponds to T5, with Δt being 5 days, 3 days, 1 day, 0.5 days, and 0, respectively. Taking λ=0.1, the time decay factors are 0.6065, 0.7408, 0.9048, 0.9512, and 1, respectively.

[0043] Location factor: n=5, T1 is the first touchpoint → 0.4, T5 is the last touchpoint → 0.4, and the three middle touchpoints are each allocated 0.2 / 3≈0.0667. For the channel quality factor: assuming the historical conversion rate of the search engine is 1.2 times the average, then f_qual(T1)=1.2; the official website channel is 0.9, the customer service channel is 1.0, the trial channel is 1.1, and the payment channel is 1.0.

[0044] Calculate the overall weight of each touchpoint, then normalize it to obtain the contribution ratio. Finally, allocate the order amount of 1000 yuan proportionally: T1 receives approximately 380 yuan, T5 receives approximately 350 yuan, and the three middle touchpoints receive a total of approximately 270 yuan.

[0045] Accordingly, this embodiment, through the above implementation method, enables early ad clicks to obtain reasonable attribution benefits, avoiding the problem of traditional last click attribution ignoring them.

[0046] In this embodiment, the customer acquisition tracking result refers to the data set that is finally output after the contribution weight of each touchpoint is calculated, which is used to characterize the channel contribution distribution. Its essence is the mapping result of touchpoint attribution weight in the channel dimension.

[0047] For example, customer acquisition tracking results may include channel attribution weights, which are the weights of all user touchpoint contributions by channel, to obtain the weight coefficient of each channel in the total transaction amount.

[0048] For example, if the total transaction amount in a certain month is 1 million yuan, after weighting, the attribution weight of the search engine channel is 32.5%, which means that the weight allocation value of this channel is 325,000 yuan.

[0049] In addition, customer acquisition tracking results may also include detailed records of user touchpoints, that is, attribution data records corresponding to each user and each conversion, such as the user's behavior trajectory sequence, the weight coefficient of each touchpoint in the sequence, and the node type identifier that triggered the attribution.

[0050] In addition, customer acquisition tracking results may also include derivative metrics, such as average channel touchpoint weight, which reflects the statistical mean of the weight coefficient of a single touchpoint in that channel; channel assist frequency ratio, which reflects the proportion of the channel appearing as a non-final touchpoint in the conversion path; channel first touch ratio and last touch ratio, which reflect the frequency of the channel appearing at the beginning and end positions of the behavior sequence, respectively; and channel conversion time distribution, which reflects the statistical characteristics of the time window from the first touch to conversion for users associated with that channel.

[0051] Furthermore, the method in this embodiment also includes: The conversion value corresponding to the key conversion nodes is allocated to each channel according to the contribution weight of each touchpoint. The return on investment of each channel is calculated based on the attribution revenue allocated to each channel and the real-time investment cost obtained. Decision recommendations are generated based on the ROI assessment results. In this embodiment, the investment costs for each channel are obtained in real-time from the financial module or advertising API. Costs include advertising expenditures (such as search engine bidding fees and social media advertising costs), personnel costs (such as salaries of marketing personnel responsible for that channel), platform costs (such as e-commerce platform commissions), and other miscellaneous costs. All costs for the same channel within the time window T are summed to obtain the total cost, Cost_total.

[0052] At the same time, all revenue attributed to this channel (including direct conversion revenue and assist attribution revenue) is aggregated to obtain the total revenue Revenue_total.

[0053] The calculation formula is as follows: ; in, The current advertising cost for these channels, For personnel costs, For platform costs, Other costs; The formula for aggregating direct conversion revenue and attributable revenue from current channels is as follows: ; in, For the direct conversion revenue of the current channels, This is the primary revenue attributed to the current channel. The formula for calculating the return on investment (ROI) of the current channel is as follows:

[0054] The calculated ROI is compared with a preset threshold (e.g., ROI threshold = 2.0). Simultaneously, the ROI values ​​of the most recent N time points (N = 10) of this channel are selected, and linear regression analysis is performed using the least squares method to fit the straight line y = kx + b. The trend is determined based on the sign of the slope k: k > 0 indicates an upward trend, and k < 0 indicates a downward trend.

[0055] For example, the decision rule is as follows: If ROI > the threshold and the trend is upward, a recommendation to increase the budget for that channel is generated.

[0056] If the conversion rate (calculated from attribution revenue and number of touchpoints) is lower than a preset threshold, suggestions to optimize the landing page or pause the campaign will be generated.

[0057] Display the above decision-making suggestions in a visual form on the monitoring panel, and at the same time push them to the marketing automation platform through the API. In addition, record the metric changes after the strategy execution, and dynamically adjust the preset threshold through reinforcement learning (such as Q-learning) to form a closed-loop optimization.

[0058] So far, this embodiment has fully implemented a complete customer acquisition tracking method from multi-channel data collection, unified identity recognition, full-link behavior tracking, multi-touch attribution, real-time ROI calculation to intelligent decision-making suggestions.

[0059] As a feasible implementation, before constructing the connectivity graph based on the user behavior data, the method further includes: Perform standardization processing, duplicate removal processing and quality inspection on the user behavior data; Extract the identity identification fields reported by each channel, and the identity identification includes at least one of a telephone number hash value, an email address hash value, a device ID, and an IP address; Use the DBSCAN density clustering algorithm to cluster the identity identifications, filter out the noise points, and calculate the similarity between different identity identifications after clustering. The similarity is calculated using cosine similarity, edit distance or Jaccard coefficient; Perform confidence evaluation according to the similarity, and merge the identity identifications belonging to the same subject to generate nodes for constructing the connectivity graph.

[0060] Specifically, before constructing the connectivity graph, first preprocess the original user behavior data. For example, for data from different channels, unify the timestamp format (all converted to Unix millisecond timestamps) and unify the channel identification encoding (map different writings such as "baidu_sem" and "Baidu Search" to the standard Channel_ID).

[0061] The duplicate removal processing uses a combination key based on (user identification, event type, timestamp), retains the first record within a sliding window (such as 1 second), and discards duplicate records. Quality inspection includes verifying required fields (such as the original identity identification field before the generation of Global_UID cannot be empty) and filtering out obviously abnormal data (such as timestamps being future times or earlier than the online time).

[0062] After that, extract the identity identification fields, perform SHA-256 hashing on the mobile phone number and email to protect privacy. Use the DBSCAN algorithm to cluster the identity identifications, set the neighborhood radius ε = 0.5, the minimum number of samples MinPts = 3, group the density-connected identifications into one cluster, and mark and filter out the isolated noise points. For different identifications within the cluster after clustering, calculate the cosine similarity (encoding the identifications as vectors), edit distance (applicable to short strings) or Jaccard coefficient (applicable to set features).

[0063] Matches with a similarity score (e.g., above 0.8) are considered high-confidence matches. The identity identifiers are then merged, and the merged identifiers are used as nodes in the subsequent construction of the connected graph. This method significantly reduces the number of nodes and false edges in the connected graph, improving the accuracy of identity mapping.

[0064] As one possible implementation method, the construction of the connected graph includes: Using identity identifiers from different entities as nodes, and the associations between identifiers caused by the same event or the same entity as edges, when different identity identifier nodes form a connected component, all identity identifiers within the connected component are mapped to the same global user identifier.

[0065] Specifically, this embodiment can use a graph database (such as Neo4j) or an in-memory graph computing framework (such as GraphX) to construct a connected graph.

[0066] Assume that from the preprocessing stage, we obtain node A (device ID: ABC123), node B (phone number hash: 5e884898da28...), and node C (OpenID: oUpF8uMuAJO_M2pxb2Q9...). When event 1 is detected: device ABC123 logged in with the account-bound mobile phone number hash 5e884898da28..., then an edge is established between A and B; Event 2: If OpenID oUpF8... is bound to the same phone number hash in a WeChat mini program, then an edge is established between B and C. At this time, nodes A, B, and C are connected by two edges, forming a connected component.

[0067] Traverse the entire graph and use the Union-Find algorithm to identify all connected components. Generate a unique Global_UID (e.g., UUID v4) for each connected component. Then, map all identities within the connected component to the Global_UID and store them in a mapping table for quick lookup during subsequent behavior trajectory extraction.

[0068] As a feasible implementation method, the key conversion nodes include at least lead nodes, intention nodes, and transaction nodes; Among them, the clue node corresponds to the registration success or form submission event, the intention node corresponds to the add to cart or request demo event, and the transaction node corresponds to the payment success or contract signing event; The time point that triggers the calculation of the contribution weight is when the user behavior matches the transaction node.

[0069] Specifically, this embodiment maintains an event type mapping table within the scene.

[0070] For example: Clue node: The event name is either "user_register" (registration successful) or "form_submit" (form submitted); Intent node: The event name is "add_to_cart" (add to cart) or "demo_request" (request demo); Transaction milestone: The event name is either "payment_success" (payment successful) or "contract_signed" (contract signed).

[0071] An event stream processing engine (such as Flink) continuously consumes user behavior data streams. For each event, the engine checks if its event name is in the aforementioned mapping table. If a transaction node is matched, attribution calculation is immediately triggered, tracing back the user's behavioral trajectory sequence (from the first touchpoint to the end of the transaction node), and performing the weight calculation and revenue allocation as described in claim 4. If a lead node or intention node is matched, only the node information is recorded, but attribution calculation is not triggered, thereby avoiding unnecessary computational overhead.

[0072] As one possible implementation method, the anomaly detection includes: Construct a sliding time window and use the 3-Sigma principle to identify abnormal data mutations within the window. When the current value of the indicator deviates from the mean within the sliding time window by more than three times the standard deviation, it is judged as an abnormal mutation.

[0073] Specifically, in this embodiment, a sliding time window is maintained for the conversion rate and return on investment (ROI) of each channel. For example, the window width is set to 60 minutes, the sliding step is 5 minutes, and each window contains a set of all data points within that time period. The mean μ and standard deviation σ of the indicators within the window are calculated in real time.

[0074] When the value of the latest data point x satisfies |x-μ|>3σ, it is determined to be an abnormal mutation. At the same time, an alarm message is generated. Generally, the alarm message can be pushed in real time via email, DingTalk / WeChat Work robot, or monitoring panel.

[0075] As one possible implementation, the method further includes: Determine the ROI values ​​of the current channel at the most recent N time points, and use the least squares method to perform linear regression analysis. Determine whether the current channel is in an upward or downward trend based on the sign of the slope, where N is a positive integer greater than 0. If the return on investment is higher than the preset threshold and the trend is upward, a suggestion to increase the budget is determined. If the conversion rate is lower than the preset threshold, suggestions to optimize the landing page or pause the delivery are generated; Record the changes in metrics after the strategy is executed. Adjust the preset threshold through reinforcement learning feedback. After generating the global user identifier, parse the UTM parameters, referral page, or channel-specific QR code identifier in the collected data, bind the parsed channel identifier to the current global user identifier, and record the associated valid time period to determine the user's channel attribution within a specific time period.

[0076] In this embodiment, for each channel, retain the ROI values of the most recent N time points (e.g., N = 12, each time point is 5 minutes apart, for a total of 1 hour), and use the least squares method to calculate the slope k of the fitted line y = kx + b. k > 0 indicates an upward trend in ROI, k < 0 indicates a downward trend, and k ≈ 0 indicates stability.

[0077] Secondly, assume the preset ROI threshold T_ROI = 2.0 and the conversion rate threshold T_CVR = 0.05; When ROI > T_ROI and k > 0, output "Suggest to increase the budget by 10%"; when CVR < T_CVR, output suggestions to optimize the landing page A / B test or suggest pausing the delivery for 24 hours for observation.

[0078] In this embodiment, record the execution results of each decision suggestion (e.g., the change in ROI in the next time window after increasing the budget). Use the ROI change as the reward signal and update the preset thresholds T_ROI and T_CVR using the Q-learning algorithm. For example, if the ROI further improves after increasing the budget, it is a positive reward and the threshold can be appropriately increased; otherwise, it is a negative reward and the threshold is decreased.

[0079] After generating the global user identifier, parse the UTM parameters (utm_source, utm_medium, utm_campaign), HTTP Referrer, or channel-specific QR code ID carried in the user behavior data. Bind the parsed Channel_ID to the Global_UID and record the start and end times of the binding in the time series database.

[0080] For example, if a user first enters by scanning the QR code of Baidu SEM, bind the "baidu_sem" channel, and set the validity period to 30 days. Within these 30 days, all conversions of this user (even if entering again through other channels) will be recorded as assisted attribution, so as to more accurately calculate the long-term value of each channel.

[0081] Furthermore, please refer to Figure 2The second embodiment of the present invention proposes a customer acquisition tracking system based on multi-channel integration, the system comprising: Data acquisition module 210 is configured to acquire user behavior data from multiple channels; The data mapping module 220 is configured to construct a connected graph based on the user behavior data, and map the identity identifiers belonging to the same connected component in the connected graph to a unique global user identifier; The data extraction module 230 is configured to use the global user identifier as an index to extract the historical touchpoints of each user across all channels in chronological order to form a behavior trajectory sequence, and to identify the preset key conversion nodes in the behavior trajectory sequence. The weight calculation module 240 is configured to backtrack the behavior trajectory sequence when the user behavior matches the key conversion node, and calculate the contribution weight for each touchpoint in the sequence. The contribution weight is determined based on the time decay factor, position factor and channel quality factor of the touchpoint. The time decay factor is negatively correlated with the interval between the time of the touchpoint occurrence and the time of conversion completion. The position factor is set such that the weight of the first and last touchpoints in the behavior trajectory sequence is greater than that of the middle touchpoints. The channel quality factor is dynamically adjusted based on the historical conversion rate of the corresponding channel. The data output module 260 is configured to determine the customer acquisition tracking results of each channel based on the contribution weight of each touchpoint.

[0082] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process and related descriptions of the system described above can be found in the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0083] It should be noted that the customer acquisition tracking system based on multi-channel integration provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the modules or steps in the embodiments of the present invention can be further decomposed or combined. For example, the modules in the above embodiments can be merged into one module, or further divided into multiple sub-modules to complete all or part of the functions described above. The names of the modules and steps involved in the embodiments of the present invention are only for distinguishing the various modules or steps and are not considered as an improper limitation of the present invention.

[0084] An electronic device according to a third embodiment of the present invention includes: At least one processor; and a memory communicatively connected to at least one of the processors; The memory stores instructions that can be executed by the processor to implement the aforementioned customer acquisition tracking method based on multi-channel integration.

[0085] A computer-readable storage medium according to a fourth embodiment of the present invention stores computer instructions, which are executed by the computer to implement the above-described customer acquisition tracking method based on multi-channel integration.

[0086] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process and related descriptions of the storage device and processing device described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0087] The following is for reference. Figure 3 It shows a schematic diagram of the structure of a computer system used to implement the methods and system embodiments of this application. Figure 3 The server shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0088] like Figure 3 As shown, the computer system includes a Central Processing Unit (CPU) 301, which can perform various appropriate actions and processes based on programs stored in Read Only Memory (ROM) 302 or programs loaded from storage section 308 into Random Access Memory (RAM) 303. The RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An Input / Output (I / O) interface 305 is also connected to the bus 304.

[0089] The following components are connected to I / O interface 305: an input section 306 including a keyboard, mouse, etc.; an output section 307 including a cathode ray tube (CRT), liquid crystal display (LCD), and speakers, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card and a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to I / O interface 305 as needed. Removable media 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0090] Specifically, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by central processing unit (CPU) 301, it performs the functions defined in the methods of this application. It should be noted that the computer-readable medium described above in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on a computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0091] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0092] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0093] The terms “first”, “second”, etc., are used to distinguish similar objects, not to describe or indicate a specific order or sequence.

[0094] The term "comprising" or any other similar term is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus / device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent in such process, method, article, or apparatus / device.

[0095] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A customer acquisition tracking method based on multi-channel integration, characterized in that, include: Acquire user behavior data from multiple channels; A connected graph is constructed based on the user behavior data, and the identity identifiers belonging to the same connected component in the connected graph are mapped to unique global user identifiers. Using the global user identifier as an index, historical touchpoints of each user across all channels are extracted in chronological order to form a behavioral trajectory sequence, and key conversion nodes preset in the behavioral trajectory sequence are identified. When user behavior matches the key conversion node, the behavior trajectory sequence is traced back, and a contribution weight is calculated for each touchpoint in the sequence. The contribution weight is determined based on the touchpoint's time decay factor, position factor, and channel quality factor. The time decay factor is negatively correlated with the interval between the touchpoint occurrence time and the conversion completion time. The position factor is set such that the weight of the first and last touchpoints in the behavior trajectory sequence is greater than that of the middle touchpoints. The channel quality factor is dynamically adjusted based on the historical conversion rate of the corresponding channel. Based on the contribution weight of each touchpoint, the customer acquisition tracking results for each channel are determined.

2. The method according to claim 1, characterized in that, Before constructing the connectivity graph based on the user behavior data, the method further includes: The user behavior data is standardized, deduplicated, and quality checked. Extract the identity identifier fields reported from each channel. The identity identifier includes at least one of the following: phone number hash value, email address hash value, device ID, and IP address. The identity identifiers are clustered using the DBSCAN density clustering algorithm, and noise points are filtered out. The similarity between different identity identifiers after clustering is calculated, and the similarity is calculated using cosine similarity, edit distance or Jaccard coefficient. Based on the similarity, a confidence assessment is performed, and identity identifiers belonging to the same subject are merged to generate nodes for constructing a connected graph.

3. The method according to claim 1, characterized in that, The construction of the connected graph includes: Using identity identifiers from different entities as nodes, and the associations between identifiers caused by the same event or the same entity as edges, when different identity identifier nodes form a connected component, all identity identifiers within the connected component are mapped to the same global user identifier.

4. The method according to claim 1, characterized in that, The formula for calculating the time decay factor is: ； in, The time interval between the occurrence of contact Ti and the transition time. The preset attenuation coefficient; The formula for calculating the position factor is: ； Where n is the total length of the contact sequence; The formula for calculating the contribution weight is as follows: ； The attribution revenue allocated to each channel is calculated using the following formula: ； ； in, The time decay factor, For position factors, For channel quality factors, The overall weight of a single touch point. The contribution ratio of each touchpoint The attribution revenue allocated to channel c.

5. The method according to claim 1, characterized in that, The key conversion nodes include at least the lead node, the intention node, and the transaction node; Among them, the clue node corresponds to the registration success or form submission event, the intention node corresponds to the add to cart or request demo event, and the transaction node corresponds to the payment success or contract signing event; The time point that triggers the calculation of the contribution weight is when the user behavior matches the transaction node.

6. The method according to claim 1, characterized in that, The method further includes: The conversion value corresponding to the key conversion nodes is allocated to each channel according to the contribution weight of each touchpoint. The return on investment of each channel is calculated based on the attribution revenue allocated to each channel and the real-time investment cost obtained. Decision recommendations are generated based on the assessment results of the aforementioned rate of return on investment.

7. The method according to claim 6, characterized in that, The calculation of the return on investment for each channel includes: The total cost of current channels is aggregated, including advertising expenditure costs, personnel costs, platform costs, and other costs. The calculation formula is as follows: ； in, The current advertising cost for these channels, For personnel costs, For platform costs, Other costs; The formula for aggregating direct conversion revenue and attributable revenue from current channels is as follows: ； in, For the direct conversion revenue of the current channels, This is the primary revenue attributed to the current channel. The formula for calculating the return on investment (ROI) of the current channel is as follows: 。 8. The method according to claim 7, characterized in that, The method further includes: Determine the ROI values ​​of the current channel at the most recent N time points, and use the least squares method to perform linear regression analysis. Determine whether the current channel is in an upward or downward trend based on the sign of the slope, where N is a positive integer greater than 0. If the return on investment is higher than the preset threshold and the trend is upward, then it is recommended to increase the budget; if the conversion rate is lower than the preset threshold, then it is recommended to optimize the landing page or suspend the campaign. Record the changes in indicators after the strategy is executed, adjust the preset threshold through reinforcement learning feedback, and after generating a global user identifier, parse the UTM parameters, reference page or channel-specific QR code identifier in the collected data, bind the parsed channel identifier with the current global user identifier, and record the associated valid time period to determine the channel affiliation of the user within a specific time period.

9. A customer acquisition tracking system based on multi-channel integration, characterized in that, include: The data acquisition module is configured to acquire user behavior data from multiple channels; The data mapping module is configured to construct a connected graph based on the user behavior data, and map the identity identifiers belonging to the same connected component in the connected graph to a unique global user identifier; The data extraction module is configured to use the global user identifier as an index to extract the historical touchpoints of each user across all channels in chronological order to form a behavior trajectory sequence, and to identify the preset key conversion nodes in the behavior trajectory sequence; The weight calculation module is configured to backtrack the behavior trajectory sequence when the user behavior matches the key conversion node, and calculate the contribution weight for each touchpoint in the sequence. The contribution weight is determined based on the time decay factor, position factor and channel quality factor of the touchpoint. The time decay factor is negatively correlated with the interval between the time of touchpoint occurrence and the time of conversion completion. The position factor is set such that the weight of the first and last touchpoints in the behavior trajectory sequence is greater than that of the middle touchpoints. The channel quality factor is dynamically adjusted based on the historical conversion rate of the corresponding channel. The data output module is configured to determine the customer acquisition tracking results for each channel based on the contribution weight of each touchpoint.

10. A computer device, characterized in that, include: At least one processor; and a memory communicatively connected to at least one of the processors; The memory stores instructions that can be executed by the processor to implement the customer acquisition tracking method, system, and computer device based on multi-channel fusion as described in any one of claims 1-8.

Images