Information display efficiency conversion evaluation system for promotional information

Abstract

This invention discloses a system for evaluating the efficiency of promotional information display, relating to the field of promotional information conversion evaluation technology. The system collects basic user behavior data and device data during batch promotion periods. Users are categorized through primary classification of account characteristic data and secondary classification of device characteristic data to obtain promotional reach data. A natural conversion tendency score is calculated based on behavioral tendency characteristics. Users are segmented according to this score, and filtering rules are constructed by combining the baseline conversion probability before promotion, the conversion probability after reach, and the conversion probability after promotion interruption for each segment. Users whose conversion is driven by promotional information are then selected, and the promotional conversion rate is calculated. This invention solves the problems of cross-device and cross-scenario data fragmentation and weakened user-level tracking in existing technologies, while also avoiding the defects of traditional attribution methods that misjudge natural conversion. This ensures that the calculated promotional conversion rate truly and accurately reflects the core driving effect of the promotional information.

Description

Technical Field

[0001] This invention belongs to the field of publicity information conversion evaluation technology, specifically a publicity information display efficiency conversion evaluation system. Background Technology

[0002] Existing promotional information conversion evaluation technologies mostly rely on single-dimensional data collection and superficial statistical analysis. They typically collect basic data such as exposure and click-through rate from scattered terminals, and then directly calculate conversion-related indicators after simple classification to measure the effectiveness of the promotion.

[0003] Existing technologies have significant shortcomings: On the one hand, the problem of data fragmentation across devices and scenarios is prominent. Due to restrictions on privacy policies and insufficient device permissions, user behavior data from multiple online devices and offline scenarios cannot be effectively linked, resulting in weakened user-level tracking capabilities, insufficient data closure rate for promotional outreach, and difficulty in comprehensively reflecting the full picture of user interaction with promotional information. On the other hand, there is a lack of scientific user classification and natural conversion interference removal mechanisms. Traditional attribution methods are prone to misjudging naturally converting users who would convert without promotion as promotion-driven conversions. At the same time, single data analysis and filtering logic cannot accurately distinguish between real promotional effects and interference data, leading to distorted calculations of the final promotional conversion rate. The evaluation results cannot truly reflect the core driving value of promotional information and cannot provide reliable data support for optimizing promotional strategies.

[0004] This invention provides an information display efficiency conversion evaluation system for promotional information to solve the above-mentioned technical problems. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes an information display efficiency conversion evaluation system for promotional information.

[0006] To achieve the above objectives, a first aspect of the present invention provides an information display efficiency conversion evaluation system for promotional information, including a cloud and several edge terminals connected thereto;

[0007] Edge: Used to collect promotional exposure data during batch promotion periods; where batch promotion periods are set up to verify the conversion efficiency of promotional information, and promotional exposure data includes behavioral basic data and device basic data;

[0008] Cloud-based: Used to categorize users based on advertising exposure data, obtaining user advertising reach data; using advertising reach data as the basis for statistical analysis, filtering out users whose conversions are driven by advertising information, and integrating the advertising reach data of the filtered users into an advertising conversion dataset; and,

[0009] Used to calculate the conversion rate of promotional information based on the promotional conversion dataset.

[0010] In one possible implementation, advertising exposure data is collected during the batch advertising period, including:

[0011] Set up batch publicity periods for promotional information; these batch publicity periods include publicity placement periods and publicity interruption periods.

[0012] Collect advertising exposure data of users reached during the advertising period; the advertising exposure data includes device characteristic data and behavioral characteristic data.

[0013] In one possible implementation, the advertising exposure data is categorized by user, including:

[0014] Extract basic device data from several publicity and exposure data; among which, basic device data includes account characteristic data and device characteristic data;

[0015] Based on the device's basic data, several types of promotional exposure data are categorized, and the promotional exposure data of the same category are sorted in chronological order according to the timestamp to obtain the user's promotional reach data.

[0016] In one possible implementation, several types of advertising and exposure data are categorized based on basic device data, including:

[0017] By using account feature data, several promotional exposure data are classified to obtain several initial categories;

[0018] The unclassified publicity and exposure data is reclassified using device feature data. Based on the results of the reclassification, several initial categories are filled in to complete the classification of the publicity and exposure data.

[0019] In one possible implementation, statistical analysis is based on advertising reach data to identify users whose conversions are driven by the advertising message, including:

[0020] The campaign reached data was used to assess users' natural conversion tendency, and the results of the natural conversion tendency assessment were used to segment several users.

[0021] Statistical analysis was performed on the promotional reach data of users in each tier to obtain the reach conversion probability; user screening rules were constructed based on the reach conversion probability and the baseline conversion probability before promotion; among which, the reach conversion probability includes the conversion probability after reach and the conversion probability after promotion is interrupted;

[0022] Users who were triggered by the promotional information were selected based on user filtering rules, and the user's promotional reach data was integrated into a promotional conversion dataset.

[0023] In one possible implementation, user organic conversion propensity is assessed through advertising reach data, including:

[0024] Extract user behavioral characteristics from advertising outreach data; these characteristics include static and dynamic features. Static features include spending power tags and interaction frequency, while dynamic features are extracted from behavioral data prior to advertising outreach.

[0025] The behavioral tendency features are input into a pre-trained natural conversion tendency model to obtain the user's natural conversion tendency score; the value range of the natural conversion tendency score is [0,10].

[0026] In one possible implementation, statistical analysis is performed on the promotional reach data of users in each tier, including:

[0027] Extract promotional outreach data from each tier of users;

[0028] The conversion probability after reaching the target audience is calculated based on the statistical data of the promotional reach; where the conversion probability after reaching the target audience = the number of users who converted during the promotional period after the promotional reach / the total number of users reached in this batch;

[0029] The conversion probability after a campaign interruption is calculated based on the campaign reach data; where the conversion probability after a campaign interruption = the number of users who converted during the campaign interruption period / the total number of users in that campaign.

[0030] In one possible implementation, user filtering rules are constructed based on the reach conversion probability and the baseline conversion probability before advertising; wherein, the user filtering rules include natural conversion filtering rules and advertising conversion filtering rules;

[0031] The natural conversion screening rules are as follows: natural conversion tendency score ≥ 7 points, the difference between the conversion probability after reaching the target audience and the baseline conversion probability before the promotion is ≤ 10%, and the difference between the conversion probability after the promotion is interrupted and the baseline conversion probability before the promotion is ≤ 5%.

[0032] The conversion screening rules are as follows: the difference between the conversion probability after reaching the target audience and the baseline conversion probability before the promotion is ≥15%, and the difference between the conversion probability after the promotion is interrupted and the baseline conversion probability before the promotion is ≥25%.

[0033] In one possible implementation, the conversion rate of promotional information is calculated based on the promotional conversion dataset, including:

[0034] Based on the advertising conversion dataset, we can count the number of users who converted through advertising and the number of users reached by advertising.

[0035] The conversion rate is the ratio of the number of users converted by the campaign to the number of users reached by the campaign.

[0036] In one possible implementation, a negative impact factor is calculated based on the advertising conversion dataset to correct the advertising conversion rate, including:

[0037] Determine whether there are negative behavioral scenarios in the advertising outreach data of the advertising conversion dataset; if yes, mark the users associated with the advertising outreach data as negative scenario users; otherwise, do not mark them.

[0038] The negative impact factor is calculated based on the number of users in negative scenarios and the number of users converted from the promotion; the negative impact factor is then used to adjust the promotion conversion rate.

[0039] In one possible implementation, a negative impact factor is calculated based on the number of users experiencing negative scenarios and the number of users converted through advertising, including:

[0040] Extract the weight coefficients of negative behavior scenarios corresponding to users in negative scenarios;

[0041] The negative impact factor is obtained by weighting and summing the ratio of the number of users in negative scenarios to the number of users converted from promotional activities, along with a weighting coefficient.

[0042] Compared with the prior art, the beneficial effects of the present invention are:

[0043] 1. This invention collects basic user behavior data and device data during the batch promotion period; it categorizes users through primary classification of account feature data and secondary classification of device feature data to obtain promotion reach data; it calculates a natural conversion tendency score based on behavioral tendency characteristics, segments users according to the natural conversion tendency score, and constructs screening rules by combining the baseline conversion probability before promotion, the conversion probability after reach, and the conversion probability after promotion interruption for each layer, and then calculates the promotion conversion rate after screening out users whose conversion is driven by promotional information; this invention not only solves the problems of cross-device and cross-scenario data fragmentation and weakened user-level tracking in the prior art, but also avoids the defects of traditional attribution methods in misjudging natural conversion, so that the calculated promotion conversion rate can truly and accurately reflect the core driving effect of promotional information.

[0044] 2. Based on the promotional conversion dataset, this invention determines whether there are negative behavioral scenarios in the promotional outreach data and marks users in negative scenarios; it extracts the weight coefficients of negative behavioral scenarios, and obtains the negative impact factor by weighting and summing the ratio of the number of users in negative scenarios to the number of users converted from promotion, thereby correcting the promotional conversion rate; this invention overcomes the one-sidedness of traditional evaluation that only focuses on positive conversions, and incorporates and quantifies the implicit negative impacts such as order cancellations after conversion and churn due to excessive outreach into the evaluation system, making the corrected promotional conversion rate more in line with the actual business scenario, effectively improving the comprehensiveness, accuracy and practical reference value of the evaluation results. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a schematic diagram of the method flow for evaluating the efficiency of information display in Embodiment 1 of the present invention;

[0047] Figure 2 This is a schematic diagram of the method for filtering promotional information to drive user conversion in Embodiment 1 of the present invention;

[0048] Figure 3 This is a schematic diagram of the method for correcting the advertising conversion rate using negative impact factors in Embodiment 2 of the present invention. Detailed Implementation

[0049] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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. Example 1:

[0050] Please see Figure 1 The first aspect of the present invention provides an information display efficiency conversion evaluation system for promotional information, including a cloud and several edge terminals connected thereto;

[0051] Edge: Used to collect promotional exposure data during batch promotion periods; where batch promotion periods are set up to verify the conversion efficiency of promotional information, and promotional exposure data includes behavioral basic data and device basic data;

[0052] Cloud-based: Used to categorize users based on advertising exposure data to obtain user advertising reach data; using advertising reach data as the basis for statistical analysis, it filters out users whose conversions are driven by advertising information, integrates the advertising reach data of the filtered users into an advertising conversion dataset; and, used to calculate the advertising conversion rate of advertising information based on the advertising conversion dataset.

[0053] To improve the accuracy of evaluating the conversion efficiency of promotional information, it is necessary to collect data from multiple channels as supporting data, and to ensure the correlation between promotional exposure data and promotional information. Therefore, when collecting promotional exposure data within a batch of promotional periods, the batch of promotional periods should be set first, and promotional exposure data of users reached by the promotion within that batch of promotional periods should be collected.

[0054] The promotion period includes the promotion period and the promotion interruption period. The promotion period refers to the time range from the start of the promotion to the interruption of the promotion. The promotion interruption period refers to the time range after the promotion is interrupted.

[0055] It is worth noting that the duration of the promotional campaign and the interruption period are equal, and the promotional campaign and the interruption period are consecutive. In addition to the time range from the launch of the promotional information to its interruption, the promotional campaign also includes a buffer period after the interruption, which is 24 hours by default.

[0056] In one example, assuming the promotional information is launched at 8:00 AM on January 1st, interrupted at 8:00 AM on January 9th, and has a buffer period of 24 hours, then the promotional period is from 8:00 AM on January 1st to 8:00 AM on January 10th.

[0057] Promotional exposure data refers to basic behavioral data (such as user behavior data or user interaction data) and device data (such as exposure device data or interaction device data) of users before and after being reached by the promotional information during the batch promotion period. The time range corresponding to promotional exposure data includes the batch promotion period. Generally, it is also necessary to collect basic behavioral data and device data before the promotion is reached in order to provide data support for excluding users who have converted organically. Promotional exposure data is collected through edge devices, which include online and offline devices. Online devices include terminals such as mobile phones and computers, while offline devices include advertising screens and self-service terminals. The edge devices send the collected promotional exposure data to the cloud, where the cloud tags the promotional exposure data and assigns a unique ID.

[0058] Online, once the promotional information has been exposed, user behavior data and corresponding exposure device data are collected. User behavior data includes exposure records, click logs, dwell time, scrolling time, add-to-cart / place-order / cancellation, etc. Exposure device data includes device anonymity identifiers, timestamps, scene tags, etc.

[0059] Offline, once an interaction with a user is completed, user interaction data and corresponding interactive device data are collected. User interaction data includes QR code scanning records and contact records, while interactive device data includes device anonymity identifiers and timestamps.

[0060] It should be noted that the promotional period for each batch is linked to the promotional information. Due to differences in channels or devices between online and offline platforms, there are no restrictions on the format of the promotional information, as long as the content remains consistent. The collection of promotional exposure data requires user authorization, and the data has been anonymized; the promotional exposure data is used solely for evaluating the conversion rate of the promotional information.

[0061] Because promotional exposure data collected from edge devices comes from a wide range of sources, independently evaluating the conversion efficiency of promotional information based on the promotional exposure data from each edge device could lead to repeated evaluations of the same user's promotional exposure data, compromising the accuracy of the conversion efficiency assessment. Therefore, it is necessary to associate promotional exposure data with corresponding users in chronological order to obtain user promotional reach data, which can then be used to analyze the conversion efficiency of promotional information.

[0062] After receiving the promotional exposure data of users reached during the batch promotion phase, the cloud needs to classify the promotional exposure data by user, that is, to associate the promotional exposure data generated by the same user in multiple platforms and scenarios with that user to form the user's promotional reach data.

[0063] When categorizing advertising and exposure data by user, basic device data is extracted from several advertising and exposure data sets. Based on the basic device data, several advertising and exposure data sets are categorized, and the advertising and exposure data sets in the same category are sorted according to the order of timestamps to obtain the user's advertising reach data.

[0064] When classifying a number of promotional exposure data based on device basic data, the account feature data is used to classify the promotional exposure data in the first stage to obtain several initial categories; the device feature data is used to classify the unclassified promotional exposure data in the second stage, and the initial categories are filled in according to the results of the second classification to complete the classification of promotional exposure data.

[0065] In the process of user classification, the use of account feature data for primary classification solves the problem of promotional information reaching users on multiple platforms; the use of device feature data for secondary classification solves the problem of promotional information reaching users on multiple devices even when the account is not logged in.

[0066] Ultimately, promotional exposure data within the same category belongs to the same user. This data is then sorted by timestamp and assigned a unique identifier to obtain the user's promotional reach data. Furthermore, the conversion rate evaluation of promotional information does not require explicit user identification; a unique identifier suffices, thus avoiding any involvement of user privacy data.

[0067] In one example, the cloud receives several promotional exposure data points within a batch of promotional periods. First, it extracts the basic device data from these data. Then, using the account characteristic data within the basic device data, promotional exposure data corresponding to the same account are grouped into one category, resulting in several initial categories. If uncategorized promotional exposure data still exists, it may be due to users not logging in. The device characteristic data of these uncategorized data is matched with the device characteristic data of the promotional exposure data in each initial category. If a match is found, the uncategorized promotional exposure data is added to the initial category.

[0068] This invention addresses the problems in existing technologies, such as susceptibility to restrictions from privacy policies and insufficient device permissions, resulting in weakened user-level tracking capabilities and the inability to effectively connect online multi-device and offline behavioral data. By classifying the device-based data of promotional exposure data multiple times, this invention ensures that the obtained user promotional reach data covers multiple platforms and scenarios, thereby improving user-level tracking capabilities.

[0069] If the cloud platform directly uses the promotional reach data of each user to evaluate the conversion rate of promotional messages, the evaluation results will be unreliable. This is because the data includes not only users whose conversions are driven by promotional messages, but also users whose conversions are driven by natural conversions and other users whose conversions are not driven by promotional messages. If users whose conversions are not driven by promotional messages are included in the evaluation, the evaluation results will obviously fail to measure the conversion rate of promotional messages.

[0070] When evaluating the conversion rate of promotional information, we first use the promotional reach data received from the cloud as a foundation. By analyzing the behavioral data within this data, focusing on users whose conversions are not driven by promotional information, we can ensure the reliability of the evaluation results. The specific screening process is as follows: We assess users' natural conversion tendency using promotional reach data; based on the assessment results of natural conversion tendency, we stratify several users; we statistically analyze the promotional reach data of users in each stratum to obtain the reach-to-conversion probability; we construct user screening rules based on the reach-to-conversion probability and the baseline conversion probability before promotion; we then filter users triggered by the promotional information based on the user screening rules, and integrate the users' promotional reach data into a promotional conversion dataset.

[0071] To assess a user's natural conversion tendency through advertising outreach data, behavioral tendency features of the user are first extracted from the advertising outreach data. These behavioral tendency features include static and dynamic features. Static features include spending power tags and interaction frequency, while dynamic features are extracted from behavioral baseline data prior to advertising outreach. The behavioral tendency features are then input into a pre-trained natural conversion tendency model to obtain the user's natural conversion tendency score. The natural conversion tendency score ranges from 0 to 10.

[0072] The system extracts promotional outreach data from the cloud, extracts behavioral tendency features from each data point, and inputs these features into a pre-built natural conversion tendency model to obtain the user's natural conversion tendency score. Behavioral tendency features include static and dynamic features. Static features include spending power tags, interaction frequency with similar products in promotional information, and product prices. Dynamic features include user behavior related to similar products in the promotional information N days (e.g., N=30 days) prior to the promotional information's exposure.

[0073] It should be noted that the aforementioned collection of promotional exposure data during the batch promotion period only indicates that if the promotion reaches users during the batch promotion period, then the user's promotional exposure data will be collected. The time range corresponding to this promotional exposure data is longer than the batch promotion period. For example, the data selected for extracting behavioral tendency characteristics is user behavior data before the promotion reaches the user.

[0074] In one example, among the static features: the spending power tag can be quantified into 1-5 levels based on the spending amount / frequency over the past 6 months, reflecting the user's spending power. The interaction frequency can be the cumulative number of browsing, adding to cart, favorites, and inquiries about the target category during the past 6 months without promotional periods, reflecting long-term demand tendencies.

[0075] Among the dynamic features are: target product browsing time / number of views in the 30 days prior to the promotion, add-to-cart dwell time (retention time from add-to-cart to first contact), and search keyword matching degree (the degree of fit with the target product / category, quantified in the range of 0-1), etc.

[0076] The natural conversion tendency model can be built based on a three-layer backpropagation (BP) neural network model. First, a standard training set is constructed using a large amount of historical data. Each training sample in the standard training set includes behavioral tendency features and corresponding natural conversion tendency scores. The BP neural network model is then trained using this standard training set and labeled as the natural conversion tendency model. The BP neural network model includes an input layer, a single hidden layer, and an output layer. The number of nodes is set according to the specific training samples. Default parameters can be used for the model parameters. The training process will not be detailed here.

[0077] After obtaining users' natural conversion tendency scores, users are segmented based on these scores. For example, users can be divided into three tiers according to their natural conversion tendency scores: high tendency: 7-10 points, medium tendency: 4-6 points, and low tendency: 0-3 points.

[0078] After segmenting users based on their natural conversion tendency scores, it is necessary to construct natural conversion filtering rules and promotional conversion filtering rules based on the baseline conversion probability before promotion, the conversion probability after reach, and the conversion probability after promotion interruption. The promotional reach data is then filtered sequentially using the natural conversion filtering rules and the promotional conversion filtering rules to obtain the real conversion data triggered by the promotional information. The real conversion data is then integrated into a promotional conversion dataset.

[0079] Based on the baseline conversion probability before promotion, the conversion probability after reaching the target audience, and the conversion probability after promotion interruption, natural conversion filtering rules and promotional conversion filtering rules are constructed. Promotional reach data is filtered sequentially using natural conversion filtering rules and promotional conversion filtering rules to obtain real conversion data triggered by promotional information. The real conversion data is then integrated into a promotional conversion dataset.

[0080] Pre-promotion baseline conversion probability (P0): Directly retrieved from the tiered baseline data, extracting the corresponding probability value according to the user's tier. It represents the conversion probability of users in that tier during the same period without promotional outreach. The calculation formula is: P0 = Number of users in that tier who converted during the same period without promotion / Total number of users in that tier during the same period without promotion. For example, if 30 out of 100 users in the high-propensity tier converted during the same period without promotion, then P0 for that tier would be uniformly set to 30%. It should be noted that the tiered baseline data is obtained by statistically analyzing user conversion rates during promotional periods for the same product category without promotional information. The statistical results can be directly accessed, and the tiered logic corresponding to the tiered baseline data is the same.

[0081] Conversion probability after reaching (P1): Calculated based on the promotional reach data, according to the actual conversion rate after reaching a single user. The calculation formula is: P1 = Number of users converted during the promotional period / Total number of users reached in this batch.

[0082] Conversion probability after advertising interruption (P2): Calculates the conversion probability of the batch of users reached during the advertising interruption period. The formula is: P2 = Number of users who converted during the advertising interruption period / Total number of users reached in this batch.

[0083] In a preferred embodiment, user screening rules are constructed based on the reach conversion probability and the baseline conversion probability before advertising; wherein, the user screening rules include natural conversion screening rules and advertising conversion screening rules; the natural conversion screening rules are: natural conversion tendency score ≥ 7 points, the difference between the conversion probability after reach and the baseline conversion probability before advertising ≤ 10%, and the difference between the conversion probability after advertising interruption and the baseline conversion probability before advertising ≤ 5%; the advertising conversion screening rules are: the difference between the conversion probability after reach and the baseline conversion probability before advertising ≥ 15%, and the difference between the conversion probability after advertising interruption and the baseline conversion probability before advertising ≥ 25%.

[0084] Natural conversion filtering rules and promotional conversion filtering rules are constructed based on the baseline conversion probability before promotion (P0), the conversion probability after reaching (P1), and the conversion probability after promotion interruption (P2).

[0085] Natural conversion screening rules: Natural conversion tendency score ≥ 7 points, the difference between the conversion probability after reaching (P1) and the baseline conversion probability before promotion (P0) ≤ 10%, and the difference between the conversion probability after promotion interruption (P2) and the baseline conversion probability before promotion (P0) ≤ 5%.

[0086] The natural conversion filtering rule is designed to identify users who would convert even if the promotional message was not directly relevant to their conversion. Users are matched within a high-propensity stratum based on this rule. Matched users are identified as naturally converting users, are marked, and their promotional reach data is not used to calculate the conversion efficiency of the promotional message.

[0087] Promotion conversion screening rules: The difference between the conversion probability after reaching the target audience (P1) and the baseline conversion probability before promotion (P0) is ≥15%, and the difference between the conversion probability after promotion is interrupted (P2) and the baseline conversion probability before promotion (P0) is ≥25%.

[0088] The conversion filtering rules for advertising are designed to identify genuine conversions resulting from advertising message reach. Users are matched across different tiers based on these rules; the matched users are identified as those whose conversions were driven by the advertising message, and this data forms the core metric for calculating the conversion efficiency of advertising messages.

[0089] In one example, User A's natural conversion tendency score is 8 (≥7, belonging to the high tendency tier). The baseline conversion probability before promotion for this tier is P0 = 30%, and the conversion probability after reaching this tier is P1 = 33%. Therefore: the difference between P1 and P0 is 33% - 30% = 3% (≤10%, satisfying the rule condition); the conversion probability after the promotion for this tier is interrupted is P2 = 31%, and the difference between P2 and P0 is 31% - 30% = 1% (≤5%, satisfying the rule condition). Therefore, User A meets all the conditions of the natural conversion screening rule: ① Natural conversion tendency score = 8 ≥ 7; ② P1 - P0 = 3% ≤ 10%; ③ P2 - P0 = 1% ≤ 5%, and is judged as a natural conversion user.

[0090] User B's natural conversion tendency score is 5 points. The baseline conversion probability before promotion in this stratum is P0=10%; the conversion probability after reaching this stratum is P1=30%, and the difference between P1 and P0 is 30%-10%=20% (≥15%, satisfying the rule condition); the conversion probability after the promotion in this stratum is interrupted is P2=35%, and the difference between P2 and P0 is 35%-10%=25% (≥25%, satisfying the rule condition). Therefore, User B meets all the conditions of the promotion conversion screening rule: ①P1-P0=20%≥15%; ②P2-P0=25%≥25%, and is determined to be a user who converts through promotion. Their promotion reach data is included in the promotion conversion dataset.

[0091] After removing users whose conversions during the batch promotion period in the cloud were not driven by promotional information, the remaining users are those whose conversions were driven by promotional information. The promotional reach data of these users is then integrated into a promotional conversion dataset. The evaluation is completed by statistically analyzing the user conversion rates corresponding to this dataset. Specifically, the number of users who converted through promotion and the number of users reached through promotion are counted based on the promotional conversion dataset; the ratio of the number of users who converted through promotion to the number of users reached through promotion is taken as the promotional conversion rate.

[0092] Considering only the impact of promotional information on conversion, we obtain a dataset of users who have converted through promotion; these users are the ones who have converted through promotion. Simultaneously, based on promotional exposure data, we determine the users reached by the promotion, and the ratio of the number of users who have converted through promotion to the number of users reached by the promotion is taken as the promotional conversion rate. Example 2:

[0093] Besides positively driving conversions, promotional information can also have negative conversion impacts. These negative impacts can be used to correct promotional conversion rates and improve accuracy. First, determine if negative behavioral scenarios exist in the promotional reach data within the conversion dataset. If so, mark the users associated with the promotional reach data as negative scenario users; otherwise, do not mark them. Calculate a negative impact factor based on the number of negative scenario users and the number of users who converted through promotion. The negative impact factor is then used to correct the promotional conversion rate. When calculating the negative impact factor, extract the weight coefficients of the negative behavioral scenarios corresponding to the negative scenario users. The ratio of the number of negative scenario users to the number of users who converted through promotion is weighted and summed with the weight coefficients to obtain the negative impact factor. The weight coefficients are preset.

[0094] In one example, negative behavior scenarios are extracted from the advertising conversion dataset based on a negative impact scenario library. Negative behavior scenarios include: order cancellation after conversion, such as users who have placed an order / added it to their cart canceling their order 1-24 hours after being exposed to the advertising information; and order cancellation due to excessive reach, such as situations where users cancel their orders after placing them during multiple outreach sessions.

[0095] Extract the number of users in each negative scenario and calculate the negative impact factor = (number of users in negative scenarios / number of users converted from the campaign) × corresponding weight coefficient. For example, the weight coefficient for order cancellation after conversion can be set to 0.5, and the weight coefficient for order cancellation due to excessive reach can also be set to 0.5; if there are 200 users converted from the campaign, 10 will cancel their orders after conversion, and 5 will be lost due to negative feelings caused by excessive reach, then the negative impact factor = (10 × 0.5 + 5 × 0.5) / 200 = 3.75%.

[0096] The formula for correcting the advertising conversion rate using negative impact factors is: Corrected advertising conversion rate = Original advertising conversion rate × (1 - Negative impact factor).

[0097] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments.

[0098] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any other combination thereof. When implemented using a software program, it can be implemented entirely or partially in the form of a computer program product. This computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device containing one or more servers, data centers, etc., that can be integrated with the medium. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state disks (SSDs)).

[0099] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A system for evaluating the efficiency of information display in promotional materials, characterized in that, This includes the cloud and several edge devices connected to it; Edge: Used to collect promotional exposure data during batch promotion periods; where batch promotion periods are set up to verify the conversion efficiency of promotional information, and promotional exposure data includes behavioral basic data and device basic data; Cloud-based: Used to categorize users based on advertising exposure data to obtain user advertising reach data; using the advertising reach data as the basis for statistical analysis, users whose conversions are driven by advertising information are selected, and the advertising reach data of the selected users is integrated into an advertising conversion dataset; and, Used to calculate the promotional conversion rate of promotional information based on the aforementioned promotional conversion dataset; Based on the aforementioned promotional reach data, users whose conversions were driven by the promotional information were selected, including: The user's natural conversion tendency is assessed through the aforementioned promotional reach data, and several users are stratified based on the assessment results of the natural conversion tendency. Statistical analysis is performed on the promotional reach data of users in each tier to obtain the reach conversion probability; user screening rules are constructed based on the reach conversion probability and the baseline conversion probability before promotion; wherein, the reach conversion probability includes the conversion probability after reach and the conversion probability after promotion is interrupted; Users who were triggered by the promotional information were selected based on the user filtering rules, and the promotional reach data of the users were integrated into a promotional conversion dataset. Assessing users' natural conversion tendency through the aforementioned advertising reach data includes: User behavioral characteristics are extracted from the advertising outreach data; wherein, behavioral characteristics include static characteristics and dynamic characteristics, static characteristics include spending power tags and interaction frequency, and dynamic characteristics are extracted from the behavioral data before advertising outreach. The behavioral tendency features are input into a pre-trained natural conversion tendency model to obtain the user's natural conversion tendency score; wherein the value range of the natural conversion tendency score is [0,10]. User filtering rules are constructed based on the reach conversion probability and the baseline conversion probability before promotion; wherein, the user filtering rules include natural conversion filtering rules and promotion conversion filtering rules; The natural conversion screening rules are as follows: natural conversion tendency score ≥ 7 points, the difference between the conversion probability after reaching the target audience and the baseline conversion probability before the promotion is ≤ 10%, and the difference between the conversion probability after the promotion is interrupted and the baseline conversion probability before the promotion is ≤ 5%. The conversion screening rules are as follows: the difference between the conversion probability after reaching the target audience and the baseline conversion probability before the promotion is ≥15%, and the difference between the conversion probability after the promotion is interrupted and the baseline conversion probability before the promotion is ≥25%.

2. The information display efficiency conversion evaluation system for promotional information according to claim 1, characterized in that, Collect publicity exposure data during the batch publicity period, including: Set up batch publicity periods for promotional information; these batch publicity periods include publicity placement periods and publicity interruption periods. Collect advertising exposure data of users reached during the advertising period; the advertising exposure data includes device characteristic data and behavioral characteristic data.

3. The information display efficiency conversion evaluation system for promotional information according to claim 1, characterized in that, User categorization of advertising and exposure data includes: Extract basic device data from several of the aforementioned publicity and exposure data; wherein, basic device data includes account characteristic data and device characteristic data; Based on the device's basic data, several types of promotional exposure data are categorized, and the promotional exposure data of the same category are sorted in chronological order according to timestamps to obtain the user's promotional reach data.

4. The information display efficiency conversion evaluation system for promotional information according to claim 3, characterized in that, Based on the aforementioned device data, several types of promotional and exposure data are categorized, including: The account feature data is used to classify several of the aforementioned promotional exposure data to obtain several initial categories; The unclassified publicity and exposure data is reclassified using the device feature data, and several initial categories are filled in based on the reclassification results to complete the classification of the publicity and exposure data.

5. The information display efficiency conversion evaluation system for promotional information according to claim 1, characterized in that, Statistical analysis was conducted on the promotional outreach data for users in each tier, including: Extract promotional outreach data from each tier of users; The conversion probability after reaching the advertised user is calculated based on the advertised reach data. Wherein, the conversion probability after reaching the user = the number of users who converted during the advertised period after reaching the user / the total number of users reached in this tier. Based on the aforementioned promotional reach data, the conversion probability after promotional interruption is calculated; wherein, the conversion probability after promotional interruption = the number of users who converted in this segment during the promotional interruption period / the total number of users who were reached in this segment.

6. The information display efficiency conversion evaluation system for promotional information according to claim 1, characterized in that, The conversion rate of promotional information is calculated based on the aforementioned promotional conversion dataset, including: Based on the aforementioned promotional conversion dataset, the number of users converted by the promotion and the number of users reached by the promotion are statistically analyzed. The ratio of the number of users converted by the promotion to the number of users reached by the promotion is used as the promotion conversion rate.

7. The information display efficiency conversion evaluation system for promotional information according to claim 6, characterized in that, Calculate a negative impact factor based on the aforementioned advertising conversion dataset to correct the advertising conversion rate, including: Determine whether there are negative behavioral scenarios in the promotional reach data in the promotional conversion dataset; if yes, mark the users associated with the promotional reach data as negative scenario users; otherwise, do not mark them. A negative impact factor is calculated based on the number of users in the negative scenarios and the number of users converted from the promotion; the negative impact factor is then used to adjust the promotion conversion rate.

Images