Digital intelligence marketing decision-making platform

Abstract

The invention relates to the technical field of digital intelligence marketing decision support, and discloses a digital intelligence marketing decision platform. Constructing a customer and channel index, and forming a page view and trading volume matrix; dynamically truncating according to the mean value and the standard deviation, filling the missing, and cleaning the original data; calculating a flow-value coupling degree after global normalization, and introducing channel complementation; candidate combinations are adaptively screened according to the comprehensive score, and budgets are allocated in proportion; and dynamically adjusting the putting times in combination with the residual budget and the minimum putting unit price. According to the method, data format inconsistency is solved through unified indexing, data quality is guaranteed through statistics of truncation and interpolation, dimensional difference is eliminated through global normalization, coupling complementation gives consideration to a core and a long tail, resource waste is avoided through dynamic allocation, decision automation and refinement are achieved, and return on investment and fund utilization efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of digital marketing decision support technology, specifically a digital marketing decision platform. Background Technology

[0002] With the widespread adoption of the internet and mobile devices, countless customers leave behind rich behavioral traces on various online channels (such as search engines, social media, and e-commerce platforms), generating a large amount of transaction data. Traditional marketing decision-making methods often optimize for a single channel or a single metric (such as page views, click-through rate, or conversion rate), making it difficult to consider the traffic and value performance of customers on different channels. This leads to a situation where, in actual campaigns, marketing budgets are often concentrated on channels with high traffic but limited value returns, or conversely, budgets are biased towards channels with high transaction value but narrow coverage, making it difficult for overall marketing investment to generate the optimal return on investment (ROI).

[0003] Many solutions focus solely on customer traffic or sales volume on a particular channel, or simply add the two together before ranking, ignoring the "synergistic effect" of customers possessing both high traffic and high sales value on the same channel; or they emphasize value while ignoring volume, leading to insufficient coverage or wasted resources. Existing technologies, when segmenting customers or optimizing channels, often use predefined thresholds (e.g., traffic ≥ X times, sales volume ≥ Y yuan) for hard grouping, lacking dynamic adjustment mechanisms and failing to adapt to rapid market changes and real-time fluctuations in customer behavior. The mutual influence between channels is often overlooked; existing methods aim for single-point optimization, lacking horizontal complementary compensation for other customers or resources on the same channel or other channels, failing to improve campaign efficiency from an overall perspective. Many platforms still manually adjust campaigns based on fixed weekly / monthly budgets, or use simple proportional allocation models, lacking the ability to continuously and dynamically reallocate fine-grained factors such as remaining budget, minimum cost per campaign, and number of campaigns. Existing algorithms and models typically rely on complex machine learning or optimization processes, requiring significant investment in data labeling, model training, and manual intervention, and are difficult to deploy quickly into actual marketing systems, hindering the application and promotion by SMEs.

[0004] Therefore, this case aims to propose a digital marketing decision-making platform that organically integrates scattered heterogeneous data such as traffic and transaction volume from multiple sources into quantifiable marketing metrics through standardized processing and coupled calculation. Based on dynamic budgets and campaign costs, it can then form the optimal resource allocation plan for each customer-channel pair. Summary of the Invention

[0005] This invention provides a digital marketing decision-making platform, which helps to solve the problems mentioned in the background art.

[0006] This invention provides the following technical solution: a digital marketing decision-making method, comprising:

[0007] Build customer index sets and channel index sets, obtain the original visit volume and transaction volume data of each customer in each channel, and form the original data matrix;

[0008] Calculate the mean and standard deviation of page views and transaction volume, set effective upper and lower limits, and clean the raw data based on these limits;

[0009] Calculate the total number of visits and transaction volume for all customers and channels, and process it according to the normalization formula;

[0010] Configure the coupling degree of traffic and value for each customer across all channels;

[0011] Calculate the total coupling degree of all customers and set the complementarity degree for each customer in each channel;

[0012] Calculate the final composite score for each customer-channel pair, determine the candidate score threshold, and select candidate pairs based on the composite score;

[0013] Based on the combined scores of all candidate pairs, calculate the budget allocated to each customer in each channel;

[0014] The number of campaigns for each client and channel is dynamically adjusted based on the remaining budget, minimum bid price, and budget already used.

[0015] Optionally, the construction of the customer index set and channel index set, obtaining the original visit volume and transaction amount data of each customer on each channel, and forming an original data matrix, specifically includes:

[0016] Construct a customer index set I = {1, 2, ..., N}; where i ∈ I is the customer index, each i represents a unique customer; N is the total number of customers;

[0017] Construct a channel index set J = {1, 2, ..., M}; where j ∈ J is the channel index, and each j represents a unique channel; M is the total number of channels;

[0018] Obtain the raw visit volume data of the i-th customer on the j-th channel, denoted as . And form the original visit volume matrix

[0019] Obtain the original transaction amount data of the i-th customer in the j-th channel, denoted as . And form the original transaction volume matrix

[0020] Obtain the overall budget, denoted as B. tot Let P be the cost per placement for channel j.j .

[0021] Optionally, the calculation of the mean and standard deviation of visit volume and transaction amount, setting effective upper and lower limits, and cleaning the raw data according to these upper and lower limits, specifically includes:

[0022] Calculate the mean of original visits from all customers and channels.

[0023] Calculate the standard deviation of all raw visits.

[0024] Construct the maximum access limit U separately F And the lower limit of the number of visits truncation L F :

[0025] U F =μ F +2σ F L F =max(μ F -2σ F ,0);

[0026] Clean the traffic from the i-th customer on the j-th channel:

[0027] Among them, F ij Let i be the number of visits to the i-th customer after cleaning in the j-th channel;

[0028] Set the valid visitor index set and count: in, For the customer index set on channel j that has no missing visits; The number of valid customers based on the original traffic on channel j;

[0029] Calculate the sum of all valid customer visits on channel j.

[0030] like If missing, then

[0031] Calculate the average transaction amount μ separately V With standard deviation σ V ;

[0032] Construct the truncated upper limit of transaction amount U respectively V and the lower limit of transaction volume cutoff L V ;

[0033] Clean the transaction amount of the i-th customer in the j-th channel:

[0034] Among them, Vij Let i be the transaction amount of the i-th customer after cleaning in the j-th channel;

[0035] Set up a valid customer index set based on transaction volume.

[0036] Number of valid customers with transaction volume on channel j

[0037] Calculate the sum of transaction amounts of all valid customers on channel j.

[0038] like If missing, then

[0039] Optionally, the calculation of the total number of visits and transaction amounts for all customers and channels, and the processing according to the normalization formula, specifically includes:

[0040] Calculate the total number of visits from all customers and channels.

[0041] Calculate the total transaction volume across all customers and channels.

[0042] Normalized visit count

[0043] Normalized transaction volume

[0044] If S F >0, then Otherwise all

[0045] If S V >0, then Otherwise all

[0046] Optionally, setting the coupling degree of traffic and value for each customer across channels specifically includes:

[0047] Let the coupling degree of traffic value of the i-th customer in the j-th channel be .

[0048] Optionally, the calculation of the sum of coupling degrees for all customers and the setting of complementarity degrees for each customer in each channel specifically includes:

[0049] Calculate the sum of coupling degrees of all customers in channel j.

[0050] On acquisition channel j, the average coupling degree h of all customers except the i-th customer. ij :

[0051]

[0052] Optionally, the step of calculating a final comprehensive score for each customer-channel pair, determining a candidate score threshold, and selecting candidate pairs based on the comprehensive score specifically includes:

[0053] Calculate the final overall score M for the i-th customer in the j-th channel. ij =C ij +H ij ;

[0054] Obtain the maximum composite score for all customer-channel pairs.

[0055] Set candidate score threshold

[0056] Building the selected customer-channel pair set

[0057] Calculate the number of candidate pairs

[0058] Optionally, the step of calculating the budget allocated to each customer in each channel based on the combined score of all candidate pairs specifically includes:

[0059] Calculate the sum of the overall scores for all candidate pairs.

[0060] Calculate the budget B allocated to the i-th customer in the j-th channel. ij :

[0061]

[0062] Optionally, the step of dynamically adjusting the number of campaigns for each customer-channel based on the remaining budget, minimum bid price, and used budget specifically includes:

[0063] Set the initial number of deliveries for the i-th customer on the j-th channel as follows:

[0064] Get used budget

[0065] Get the remaining budget B rem =B tot -B used ;

[0066] Get the minimum delivery cost within the candidate set

[0067] When B rem ≥P min At that time, steps S801 to S804 are executed repeatedly until B.rem <P min :

[0068] S801. Construct the candidate pair index set Q that can be deployed with the current remaining budget:

[0069]

[0070] S802, if Terminate and exit the loop;

[0071] S803, Select (i * ,j * )∈Q, such that If there are multiple values, then take the minimum value i, and then the minimum value j;

[0072] S804, Update:

[0073]

[0074] List {(i,j),C ij H ij M ij B ij N ij}

[0075] A platform for implementing the aforementioned digital marketing decision-making method includes:

[0076] The data acquisition and identification module is used to collect index information about customers and channels;

[0077] The data cleaning module is used to clean and preprocess the collected raw data;

[0078] The normalization module is used to perform global normalization on the data;

[0079] The traffic-value coupling calculation module is used to calculate the traffic-value coupling degree of each customer in each channel.

[0080] The channel complementarity module is used to calculate the complementarity between channels;

[0081] The overall score calculation module is used to calculate the overall score for each customer-channel pair;

[0082] The budget allocation module is used to allocate the budget based on the overall score;

[0083] The campaign redistribution module is used to redistribute campaign frequency and budget based on the remaining budget.

[0084] The present invention has the following beneficial effects:

[0085] 1. By constructing a "customer index set" and a "channel index set," each customer and channel is mapped to a unified numbering system, breaking down the problems of inconsistent data formats and misaligned records from multiple sources. Compared to existing technologies that rely on manual comparison or piecemeal system imports, this solution automatically generates row and column indexes, achieving matrix processing of raw visit volume and transaction amount data, ensuring consistency and scalability in subsequent data operations. This approach not only significantly reduces the manpower cost of data preparation but also lays the foundation for batch computing and large-scale parallel processing through a unified index format, effectively improving deployment speed and maintenance convenience.

[0086] 2. By statistically analyzing the central tendency and dispersion of website traffic and transaction volume, upper and lower limits are dynamically set to truncate extreme values ​​exceeding the normal range and impute missing data with averages, avoiding calculation biases caused by anomalies or null values. Unlike traditional methods that simply remove anomalies or directly fill in fixed values, this solution uses threshold settings based on global statistical characteristics, making data cleaning more adaptive and robust. This method not only eliminates noise from outliers but also ensures data integrity, providing high-quality input for subsequent normalization and coupled calculations, thus contributing to improved stability and predictability of the marketing strategy.

[0087] 3. For visitor volume and transaction volume data with different dimensions, a unified mapping of the overall sum normalizes all data to the same range, thus eliminating comparison barriers caused by differences in dimensions. Compared to local or channel-specific normalization methods, this solution adopts a global perspective, ensuring that all customer-channel pairs are evaluated under the same standard, resulting in higher comparability and fairness in coupling and overall scores. This method effectively avoids model bias caused by excessive differences in the scale of various channels, making traffic and value indicators more balanced across the overall sample set, providing a reliable foundation for subsequent strategy formulation.

[0088] 4. By synthesizing the normalized visitor traffic and transaction volume using a geometric mean, this approach extracts customer-channel interaction pairs that simultaneously possess high traffic and high value, thus forming a "coupling" metric. Unlike traditional methods that rely solely on single-dimensional ranking or simple weighting, this solution emphasizes the coexistence of both factors, enabling more precise targeting of key combinations that possess both substantial potential traffic and high commercial value. This innovative coupling strategy effectively avoids the biases caused by a single metric, focusing decisions on user-channel pairs with genuine conversion potential and improving the ROI of marketing investment.

[0089] 5. Considering that a single customer-channel pair might be overlooked due to insufficient coupling, a "lateral compensation" mechanism is provided for weakly coupled pairs by calculating the average coupling level of other customers within the same channel, enhancing the diversity and coverage across different channels. Compared to existing technologies that simply eliminate low-scoring items or increase resource allocation for high-scoring items, this solution maintains the core high-potential combination while not abandoning potential long-tail opportunities, promoting omnichannel synergy and risk diversification. This channel complementarity strategy makes the marketing layout more comprehensive, helps identify "potential stocks," and improves the overall robustness of the solution.

[0090] 6. The results of traffic-value coupling and channel horizontal compensation are organically combined to form a final comprehensive score, and candidate customer-channel pairs are selected based on adaptively set thresholds. Unlike traditional fixed thresholds or manual adjustment strategies, this solution dynamically adjusts the selection criteria according to the sample distribution, ensuring that the candidate set is both high-quality and not too sparse. This method improves the accuracy and coverage of candidate pairs, provides a more representative decision-making space for subsequent budget allocation, effectively balances efficiency and fairness, ensures that excellent combinations can stand out, and guarantees that long-tail combinations have a chance to be discovered.

[0091] 7. Based on the overall score ratio of all candidate pairs, the overall budget is allocated in priority order to ensure that high-scoring combinations receive more resources, while low-scoring combinations receive correspondingly fewer resources. Compared with existing common average allocation or subjective allocation methods, this scheme achieves objective, data-driven budget allocation. Through proportional allocation after score normalization, it reflects performance priority and avoids resource waste. This approach not only improves the efficiency of fund utilization but also continuously optimizes the scoring model based on feedback from the allocation results, achieving closed-loop optimization and truly ensuring that funds are "spent wisely."

[0092] 8. After the initial budget allocation, this step further quantifies the budget into executable campaign runs and dynamically adjusts the frequency of each candidate pair based on the minimum bid price and remaining funds until the budget is exhausted. Compared to the traditional static approach of allocating funds once and then not adjusting them, this solution establishes a real-time redistribution mechanism that can flexibly replenish resources based on actual consumption and remaining funds, thereby maximizing the utilization of every budget unit. This dynamic campaign strategy effectively prevents budget waste or premature exhaustion, improves the controllability and flexibility of the overall campaign performance, and provides strong support for the refined management of marketing activities. Attached Figure Description

[0093] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0094] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0095] Example, refer to Figure 1 A digital marketing decision-making method, comprising:

[0096] Build customer index sets and channel index sets, obtain the original visit volume and transaction volume data of each customer in each channel, and form the original data matrix;

[0097] Calculate the mean and standard deviation of page views and transaction volume, set effective upper and lower limits, and clean the raw data based on these limits;

[0098] Calculate the total number of visits and transaction volume for all customers and channels, and process it according to the normalization formula;

[0099] Configure the coupling degree of traffic and value for each customer across all channels;

[0100] Calculate the total coupling degree of all customers and set the complementarity degree for each customer in each channel;

[0101] Calculate the final composite score for each customer-channel pair, determine the candidate score threshold, and select candidate pairs based on the composite score;

[0102] Based on the combined scores of all candidate pairs, calculate the budget allocated to each customer in each channel;

[0103] The number of campaigns for each client and channel is dynamically adjusted based on the remaining budget, minimum bid price, and budget already used.

[0104] By uniformly indexing all customers and channels and constructing the original visit volume and transaction volume matrix, the problems of inconsistent data formats and misaligned records from multiple sources were completely resolved, ensuring that subsequent calculations were based on the same data foundation. By applying statistical feature-based upper and lower limit truncation and missing value imputation to the original matrix data, noise and missing values ​​were eliminated, guaranteeing data quality. Global normalization mapped the two different data types—visit volume and transaction volume—to the same standard interval, avoiding evaluation biases caused by single-dimensional or local normalization, making the coupling degree and comprehensive score comparable. The introduction of overcoupling extraction and channel complementarity mechanisms can accurately identify core combinations that "have both traffic and value," while also providing compensatory assessments for weakly coupled or long-tail opportunities, thus balancing a diverse layout across both primary and potential markets. Dynamically selecting candidate combinations through comprehensive scoring ensures that high-scoring combinations are prioritized, while adaptively adjusting thresholds to balance the size and quality of the candidate set. Finally, budgets are allocated proportionally based on candidate scores, and dynamic redistribution of campaign runs is implemented during actual campaign execution based on minimum single-run cost and remaining budget, ensuring that every budget is neither wasted through static allocation nor prematurely exhausted, preventing missed opportunities. Overall, this method systematically connects all stages from data preparation to campaign execution, automating, refining, and dynamizing the decision-making process, significantly improving marketing ROI and capital utilization efficiency.

[0105] The construction of customer index sets and channel index sets, obtaining the original visit volume and transaction amount data of each customer on each channel, forming an original data matrix, specifically includes:

[0106] Construct a customer index set I = {1, 2, ..., N}; where i ∈ I is the customer index, and each i represents a unique customer; n is the total number of customers; define the ID range of all customers to be marketed, and establish row indexes for subsequent matrix operations;

[0107] Construct a channel index set J = {1, 2, ..., M}; where j ∈ J is the channel index, and each j represents a unique channel; M is the total number of channels; define the number range of all distribution channels to establish column indexes for subsequent matrix operations;

[0108] Obtain the raw visit volume data of the i-th customer on the j-th channel, denoted as . And form the original visit volume matrix Collect initial, unprocessed traffic data to provide input for the cleaning process;

[0109] Obtain the original transaction amount data of the i-th customer in the j-th channel, denoted as . And form the original transaction volume matrix Collect initial, unprocessed transaction data to provide input for the cleaning process;

[0110] Obtain the overall budget, denoted as B. tot Let P be the cost per placement for channel j. j Clearly define the total available funds and the consumption of each channel, for subsequent budget allocation and calculation of the number of deployments.

[0111] By indexing customers and channels in the initial stage and assigning each customer and channel a unique number, a raw matrix of visit volume and transaction volume is constructed. This solves the alignment problems caused by inconsistent data formats and information silos across different systems, avoiding the high costs and error-prone risks of manual compilation and cross-platform import. The matrix organization method centrally manages massive, multi-source visit and transaction data, providing a unified and standardized input for subsequent batch calculation algorithms, greatly improving data processing efficiency and stability. Simultaneously, obtaining the overall budget and channel single-campaign cost at the same stage clearly defines the total available funds and resource consumption benchmarks for each channel, laying a solid foundation for subsequent budget allocation and quantification of campaign frequency. Overall, this stage reduces the time and manpower required for initial data preparation, eliminates data format mismatches and missing cost parameters, and provides reliable data support and cost control capabilities for the efficient implementation of the entire digital marketing decision-making method.

[0112] The calculation of the mean and standard deviation of visit volume and transaction amount, setting effective upper and lower limits, and cleaning the raw data based on these limits specifically includes:

[0113] Calculate the mean of original visits from all customers and channels.

[0114] Calculate the standard deviation of all raw visits.

[0115] Construct the maximum access limit U separately F And the lower limit of the number of visits truncation L F :

[0116] U F =μ F +2σ F L F =max(μ F -2σ F ,0); Identify the upper and lower boundaries of outlier access volumes to provide a threshold for truncation operations;

[0117] Clean the traffic from the i-th customer on the j-th channel:

[0118] Among them, F ijLet i be the number of visits to the i-th customer after cleaning in the j-th channel; limit extreme values ​​to a reasonable range to prevent outliers from having too much impact on subsequent calculations.

[0119] Set the valid visitor index set and count: in, For the customer index set on channel j that has no missing visits; The number of valid customers based on the original traffic on channel j;

[0120] Calculate the sum of all valid customer visits on channel j.

[0121] like If missing, then Missing data are imputed with average values ​​to ensure that no null values ​​are included in the subsequent normalization.

[0122] Calculate the average transaction amount μ separately V With standard deviation σ V ;

[0123] Construct the truncated upper limit of transaction amount U respectively V and the lower limit of transaction volume cutoff L V ;

[0124] Clean the transaction amount of the i-th customer in the j-th channel:

[0125] Among them, V ij Let i be the transaction amount of the i-th customer after cleaning in the j-th channel;

[0126] Set up a valid customer index set based on transaction volume.

[0127] Number of valid customers with transaction volume on channel j

[0128] Calculate the sum of transaction amounts of all valid customers on channel j.

[0129] like If missing, then

[0130] The transaction amount data is truncated and interpolated to eliminate anomalies and gaps.

[0131] By implementing statistical cleaning strategies on all visit volume and transaction volume data, the overall mean and fluctuation range are first calculated, and a truncation boundary is adaptively set accordingly. This ensures that outlier data deviating from the normal range is reasonably removed or limited, fundamentally avoiding the misleading influence of extreme values ​​on subsequent model judgments. Furthermore, missing or invalid data is imputed using the mean of other valid samples from the same channel, restoring data integrity and ensuring consistency between the imputed values ​​and the actual data distribution. Compared to traditional methods of simple removal or fixed-value filling, this approach retains most of the true information while eliminating noise and gaps, providing high-quality input for global normalization and coupling degree calculation. This effectively improves the robustness and stability of the model, reduces marketing decision-making biases caused by data anomalies, and enhances the usability and credibility of the entire decision-making process in complex and ever-changing market environments.

[0132] The calculation of the total number of visits and transaction amounts for all customers and channels, and the processing according to the normalization formula, specifically includes:

[0133] Calculate the total number of visits from all customers and channels.

[0134] Calculate the total transaction volume across all customers and channels.

[0135] Obtain the overall scale for subsequent normalization;

[0136] Normalized visit count

[0137] Normalized transaction volume

[0138] Mapping data of different dimensions to the same [0,1] interval ensures the comparability of subsequent coupled operations;

[0139] If S F >0, then Otherwise all

[0140] If S V >0, then Otherwise all

[0141] By normalizing the cleaned data according to its overall scale, the varying dimensions of visit volume and transaction amount are mapped to the same standard range, fundamentally solving the comparability problem caused by scale differences among different indicators. This consistent mapping from a global perspective replaces the narrow approach of normalization only within local or channel-specific areas, enabling fair comparison of different customer-channel combinations under the same evaluation benchmark and avoiding interference from channel scale differences or abnormally high values ​​in the overall ranking. The normalization results can be directly input into subsequent calculations such as coupling degree and comprehensive score, eliminating the need for repeated adjustments to model parameters for different dimensions, thus simplifying algorithm design and maintenance. This step ensures a uniform distribution of indicators across the entire scope, providing a solid data foundation for accurate identification of core combinations and dynamic resource allocation, improving system operating efficiency and the interpretability of decision-making results.

[0142] The method of setting the coupling degree between traffic and value for each customer across various channels specifically includes:

[0143] Let the coupling degree of traffic value of the i-th customer in the j-th channel be .

[0144] Coupling degree, range [0,1], measures the geometric mean of normalized traffic and value; extracts the core metric of customer-channel pairs with both high traffic and high value.

[0145] By integrating traffic and transaction volume, a traffic-value coupling index is formed, which solves the drawback of relying on a single indicator and potentially overlooking comprehensive potential. Specifically, this coupling index takes into account the dual requirements of high traffic and high transaction volume within the same measurement framework, avoiding misleading decisions based on combinations of high traffic but low conversion rates or high transaction volume but scarce traffic. With this comprehensive index, key customer-channel combinations that provide sufficient exposure opportunities and generate actual revenue can be accurately identified, providing a more valuable basis for subsequent resource allocation. At the same time, this innovative coupling method enhances the model's ability to capture real business value, enabling marketing resources to be prioritized for the most promising and profitable targets, thereby improving overall ROI and marketing efficiency.

[0146] The calculation of the sum of coupling degrees for all customers and the setting of complementarity degrees for each customer on each channel specifically includes:

[0147] Calculate the sum of coupling degrees of all customers in channel j. Measuring the overall coupling strength of the channel;

[0148] On acquisition channel j, the average coupling degree H of all customers except the i-th customer. ij :

[0149] After excluding itself from the same channel, the average coupling degree of other customers; to compensate for weak coupling pairs through the "group effect" and improve decision-making diversity.

[0150] By statistically analyzing the average coupling level of all customers on the same channel, excluding the target customer-channel combination, and using this as a complementary reference for the target combination, an innovative horizontal channel complementarity mechanism was established. This step effectively overcomes the limitations of simply focusing on highly coupled combinations while neglecting long-tail or emerging opportunities. It ensures the concentration of the main battlefield while providing compensatory evaluation for potential, high-growth combinations that are still in a "weakly coupled" state. As a result, the marketing strategy can achieve both "focusing on the optimal" and "strategizing the overall picture," making resource allocation more balanced and forward-looking. At the same time, this complementary mechanism helps to diversify the risk of deployment, optimize the overall channel structure, and ensure that there are corresponding alternative combinations available for use in different market environments, thereby enhancing the resilience and coverage of the marketing strategy.

[0151] The process of calculating a final comprehensive score for each customer-channel pair, determining a candidate score threshold, and selecting candidate pairs based on the comprehensive score specifically includes:

[0152] Calculate the final overall score M for the i-th customer in the j-th channel. ij =C ij +H ij The ranking criteria are formed by combining self-coupling and horizontal compensation.

[0153] Obtain the maximum composite score for all customer-channel pairs.

[0154] Set candidate score threshold

[0155] Automatically set the candidate set threshold;

[0156] Building the selected customer-channel pair set

[0157] Calculate the number of candidate pairs

[0158] Filter out customers – channel pairs – whose scores are above a threshold.

[0159] By organically combining coupling degree and channel complementarity, the system calculates a comprehensive evaluation score for each customer-channel combination and dynamically adjusts the screening threshold based on the score distribution of all combinations to form a candidate set, ensuring that high-potential combinations are selected first. This step overcomes the inflexibility of traditional fixed thresholds or manual adjustments, enabling the system to adaptively optimize screening criteria based on real-time data scale and distribution characteristics, ensuring both the quality of the candidate set and sufficient scale coverage. The comprehensive score reflects both the intrinsic value of the combination and the overall performance of the channel, helping decision-makers accurately distinguish between high, medium, and low priority targets. At the same time, the dynamic and visualized screening process improves decision-making transparency, facilitating operators to monitor and adjust strategies at any time, achieving intelligent and efficient screening.

[0160] The calculation of the budget allocated to each customer in each channel based on the combined score of all candidate pairs specifically includes:

[0161] Calculate the sum of the overall scores for all candidate pairs. Normalized denominator, used for proportional allocation;

[0162] Calculate the budget B allocated to the i-th customer in the j-th channel. ij :

[0163] Given a customer-channel pair budget, funds are dynamically allocated based on scores.

[0164] By normalizing the overall scores of candidate combinations and automatically allocating budgets based on their score proportions, this mechanism solves the problem of resource waste or missed opportunities caused by traditional manual subjective allocation or average allocation. Data-driven at its core, this mechanism ensures that high-scoring combinations receive more funding, enhancing their conversion efficiency. Simultaneously, it provides corresponding resources to medium or low-scoring combinations, avoiding market blind spots caused by a "one-size-fits-all" allocation. The budget allocation process is simple and transparent, allowing real-time tracking of budget usage for each combination, providing a clear basis for subsequent campaign execution and performance evaluation, thereby improving fund utilization efficiency and overall return on investment.

[0165] The method of dynamically adjusting the number of campaigns for each customer and channel based on the remaining budget, minimum bid price, and used budget includes:

[0166] Set the initial number of deliveries for the i-th customer on the j-th channel as follows: Quantify the budget into the number of actionable deployment actions;

[0167] Get used budget

[0168] Get the remaining budget B rem =B tot -Bused ;

[0169] Track funds already spent and funds yet to be allocated;

[0170] Get the minimum delivery cost within the candidate set

[0171] When B rem ≥P min At that time, steps S801 to S804 are executed repeatedly until B. rem <P min :

[0172] S801. Construct the candidate pair index set Q that can be deployed with the current remaining budget:

[0173]

[0174] S802, if Terminate and exit the loop;

[0175] S803, Select (i * ,j * )∈Q, such that If there are multiple values, then take the minimum value i, and then the minimum value j;

[0176] S804, Update:

[0177]

[0178] Use the entire budget to ensure that every penny is spent on the most cost-effective and highest-scoring investments.

[0179] List {(i,j),C ij H ij M ij B ij N ij Provide complete and actionable deployment plans and evaluation metrics.

[0180] By quantifying the budget allocated to each combination into initial executable runs, and comparing the latest remaining budget with the lowest single-run cost after each run, the most promising combination's runs are replenished sequentially, repeating this cycle until the budget is exhausted, thus achieving closed-loop management of resource usage throughout the entire process. This dynamic adjustment mechanism completely overcomes the drawbacks of static one-time allocation that cannot respond to actual consumption. While ensuring that the optimal combination is executed first, it can also flexibly increase or decrease runs based on real-time budget conditions, avoiding the risk of idle funds or premature exhaustion. This design not only improves the flexibility and responsiveness of runs but also ensures that every penny of the budget is accurately used at the most valuable points, improving the overall effectiveness and return on investment of marketing campaigns.

[0181] This embodiment also provides a platform for a digital marketing decision-making method, including:

[0182] The data acquisition and identification module is used to collect index information about customers and channels;

[0183] The data cleaning module is used to clean and preprocess the collected raw data;

[0184] The normalization module is used to perform global normalization on the data;

[0185] The traffic-value coupling calculation module is used to calculate the traffic-value coupling degree of each customer in each channel.

[0186] The channel complementarity module is used to calculate the complementarity between channels;

[0187] The overall score calculation module is used to calculate the overall score for each customer-channel pair;

[0188] The budget allocation module is used to allocate the budget based on the overall score;

[0189] The campaign redistribution module is used to redistribute campaign frequency and budget based on the remaining budget.

[0190] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0191] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A digital marketing decision-making method, characterized in that, include: Build customer index sets and channel index sets, obtain the original visit volume and transaction volume data of each customer in each channel, and form the original data matrix; Calculate the mean and standard deviation of page views and transaction volume, set effective upper and lower limits, and clean the raw data based on these limits; Calculate the total number of visits and transaction volume for all customers and channels, and process it according to the normalization formula; Configure the coupling degree of traffic and value for each customer across all channels; Calculate the total coupling degree of all customers and set the complementarity degree for each customer in each channel; Calculate the final composite score for each customer-channel pair, determine the candidate score threshold, and select candidate pairs based on the composite score; Based on the combined scores of all candidate pairs, calculate the budget allocated to each customer in each channel; The number of campaigns for each client and channel is dynamically adjusted based on the remaining budget, minimum bid price, and budget already used.

2. The digital marketing decision-making method according to claim 1, characterized in that, The construction of customer index sets and channel index sets, obtaining the original visit volume and transaction amount data of each customer on each channel, forming an original data matrix, specifically includes: Construct a customer index set I = {1, 2, ..., N}; where i ∈ I is the customer index, each i represents a unique customer; N is the total number of customers; Construct a channel index set J = {1, 2, ..., M}; where j ∈ J is the channel index, and each j represents a unique channel; M is the total number of channels; Obtain the raw visit volume data of the i-th customer on the j-th channel, denoted as . And form the original visit volume matrix Obtain the original transaction amount data of the i-th customer in the j-th channel, denoted as . And form the original transaction volume matrix Obtain the overall budget, denoted as B. tot Let P be the cost per placement for channel j. j .

3. The digital marketing decision-making method according to claim 2, characterized in that, The calculation of the mean and standard deviation of visit volume and transaction amount, setting effective upper and lower limits, and cleaning the raw data based on these limits specifically includes: Calculate the mean of original visits from all customers and channels. Calculate the standard deviation of all raw visits. Construct the maximum access limit U separately F And the lower limit of the number of visits truncation L F : U F =μ F +2s F ,L F =max(μ F -2s F ,0); Clean the traffic from the i-th customer on the j-th channel: Among them, F ij Let i be the number of visits to the i-th customer after cleaning in the j-th channel; Set the valid visitor index set and count: Non-missing}, in, For the customer index set on channel j that has no missing visits; The number of valid customers based on the original traffic on channel j; Calculate the sum of all valid customer visits on channel j. like If missing, then Calculate the average transaction amount μ separately V With standard deviation σ V ; Construct the truncated upper limit of transaction amount U respectively V and the lower limit of transaction volume cutoff L V ; Clean the transaction amount of the i-th customer in the j-th channel: Among them, V ij Let i be the transaction amount of the i-th customer after cleaning in the j-th channel; Set up a valid customer index set based on transaction volume. Non-missing}; Number of valid customers with transaction volume on channel j Calculate the sum of transaction amounts of all valid customers on channel j. like If missing, then 4. The digital marketing decision-making method according to claim 3, characterized in that, The calculation of the total number of visits and transaction amounts for all customers and channels, and the processing according to the normalization formula, specifically includes: Calculate the total number of visits from all customers and channels. Calculate the total transaction volume across all customers and channels. Normalized visit count Normalized transaction volume If S F >0, then Otherwise all If S V >0, then Otherwise all 5. The digital marketing decision-making method according to claim 4, characterized in that, The method of setting the coupling degree between traffic and value for each customer across various channels specifically includes: Let the coupling degree of traffic value of the i-th customer in the j-th channel be .

6. The digital marketing decision-making method according to claim 5, characterized in that, The calculation of the sum of coupling degrees for all customers and the setting of complementarity degrees for each customer on each channel specifically includes: Calculate the sum of coupling degrees of all customers in channel j. On acquisition channel j, the average coupling degree H of all customers except the i-th customer. ij :

7. The digital marketing decision-making method according to claim 6, characterized in that, The process of calculating a final comprehensive score for each customer-channel pair, determining a candidate score threshold, and selecting candidate pairs based on the comprehensive score specifically includes: Calculate the final overall score M for the i-th customer in the j-th channel. ij =C ij +H ij ; Obtain the maximum composite score for all customer-channel pairs. Set candidate score threshold Building the selected customer-channel pair set Calculate the number of candidate pairs 8. The digital marketing decision-making method according to claim 7, characterized in that, The calculation of the budget allocated to each customer in each channel based on the combined score of all candidate pairs specifically includes: Calculate the sum of the overall scores for all candidate pairs. Calculate the budget B allocated to the i-th customer in the j-th channel. ij :

9. The digital marketing decision-making method according to claim 8, characterized in that, The method of dynamically adjusting the number of campaigns for each customer and channel based on the remaining budget, minimum bid price, and used budget includes: Set the initial number of deliveries for the i-th customer on the j-th channel as follows: Get used budget Get the remaining budget B rem =B tot -B used ; Get the minimum delivery cost within the candidate set When B rem ≥P min At that time, steps S801 to S804 are executed repeatedly until B. rem <P min : S801. Construct the candidate pair index set Q that can be deployed with the current remaining budget: S802, if Terminate and exit the loop; S803, Select (i * ,j * )∈Q, such that If there are multiple values, then take the minimum value i, and then the minimum value j; S804, Update: List {(i,j),C ij H ij M ij B ij N ij } 10. A platform employing the digital marketing decision-making method of claim 9, characterized in that, include: The data acquisition and identification module is used to collect index information about customers and channels; The data cleaning module is used to clean and preprocess the collected raw data; The normalization module is used to perform global normalization on the data; The traffic-value coupling calculation module is used to calculate the traffic-value coupling degree of each customer in each channel. The channel complementarity module is used to calculate the complementarity between channels; The overall score calculation module is used to calculate the overall score for each customer-channel pair; The budget allocation module is used to allocate the budget based on the overall score; The campaign redistribution module is used to redistribute campaign frequency and budget based on the remaining budget.

Images