CN113269595A - Advertisement delivery method and system based on real-time prediction ROI - Google Patents
Advertisement delivery method and system based on real-time prediction ROI Download PDFInfo
- Publication number
- CN113269595A CN113269595A CN202110698938.5A CN202110698938A CN113269595A CN 113269595 A CN113269595 A CN 113269595A CN 202110698938 A CN202110698938 A CN 202110698938A CN 113269595 A CN113269595 A CN 113269595A
- Authority
- CN
- China
- Prior art keywords
- advertisement
- data
- platform
- real
- display
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0247—Calculate past, present or future revenues
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0242—Determining effectiveness of advertisements
- G06Q30/0244—Optimization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0272—Period of advertisement exposure
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Strategic Management (AREA)
- Finance (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- Marketing (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an advertisement putting decision method and system based on real-time prediction ROI, belonging to the technical field of advertisement putting in the field of mobile Internet. Aiming at the problems that the data of an advertisement showing mode is not obtained timely, the obtained data can not support a putting decision and the return on investment of the advertisement showing mode is not clear in the prior art, the invention provides a method, which is used for rapidly grasping the user quality obtained by the current advertisement putting through a method of reverse income and income prediction, and realizing the calculation of the advertisement income and ROI of historical user levels by utilizing the advertisement dotting event data in product application, the purchase quantity consumption data of an advertisement putting platform, the income data of the advertisement showing platform and the display data of a polymerization platform; meanwhile, the daily real-time income and return on investment rate of newly-added users are predicted by using daily real-time purchase quantity data and in-application advertisement dotting event data, and intelligent advertisement putting is realized.
Description
Technical Field
The invention relates to the technical field of advertisement putting in the field of mobile internet, in particular to an advertisement putting decision method and an advertisement putting decision system based on real-time prediction ROI.
Background
IAA (In-App Advertisement) and the Advertisement showing mode refer to a showing mode In which a user obtains an Advertisement by showing the Advertisement on an Advertisement showing platform and then gains a profit by using the Advertisement accessed to the Advertisement showing platform In a product. The advertising change is different from the conventional in-purchase mode in that, after attracting the user to an App such as a game, the IAA obtains revenue by presenting the user with an advertisement, rather than directly paying the user to purchase a product or service.
For the mobile phone apps of the IAA advertisement change category, in the process of advertisement putting, putting personnel need to buy the amount of the operated apps on different platforms every day. New users who acquire the amount of purchases will enter the product usage service, in the process, advertisements will be seen, and the product will earn revenue by users looking to click on these advertisements.
From the ROI (return on investment) point of view, the consumption cost of the daily purchase amount can be obtained from the purchase amount platform in real time, but the advertisement revenues of the daily live users need T +1 day. Because data of cost and income cannot be synchronously acquired, new users who acquire the amount of purchases on the same day cannot directly calculate and know what advertising revenues the new users bring, and whether the new users can return the original or not. More specifically, users from different channels buy different advertisement showing effects, and the user quality can not directly know what difference exists. This may result in that the delivery personnel lacks data decision basis when screening plans and monitoring data fluctuation, which affects the advertisement delivery effect.
Currently, to calculate advertisement performance data of new and old users in different channels, two general methods can be used:
1. all the income of the day is evenly distributed to each daily user on the assumption that the income capacity of the daily user is the same, so that the advertising income of different channels and the new and old users can be calculated. However, this approach assumes that the preconditions are too strong to be satisfied, thus resulting in large deviations of the calculated advertising revenue from the actual revenue.
2. Assuming that the number of times a user has been exposed to advertisements in a product represents his or her revenue capacity, revenue is then apportioned to each daily user. The premise of this approach is that it is actually required that all advertisement presentations have the same revenue value, which is obviously not accurate enough, and therefore, the calculated advertisement revenue has a large deviation.
Disclosure of Invention
1. Technical problem to be solved
Aiming at the problems that the data of an advertisement showing mode is not obtained timely, the obtained data can not support a putting decision and the return on investment of the advertisement showing mode is not clear in the prior art, the invention provides an advertisement putting method and an advertisement putting system based on real-time prediction ROI, which can carry out reverse income and real-time prediction, and realize the calculation of historical advertisement income and ROI by utilizing the data of advertisement dotting events in product application, the purchase consumption data of an advertisement putting platform, the income data of the advertisement showing platform and the display data of the advertisement gathering platform; meanwhile, real-time income and return on investment of newly added users on the same day are calculated by utilizing real-time purchase quantity data and in-application advertisement dotting event data, and intelligent advertisement putting is realized.
2. Technical scheme
The purpose of the invention is realized by the following technical scheme.
An advertisement putting method based on real-time prediction ROI acquires advertisement dotting event data, advertisement putting platform data, advertisement showing platform data and advertisement aggregation platform data, and calculates the advertisement dotting event data and the advertisement showing platform data to obtain thousands of display benefits of historical advertisement code positions;
and calculating real-time advertisement display income according to thousands of display income of the historical advertisement code positions and real-time advertisement dotting event data, calculating the real-time investment return rate of newly added users on the day according to the purchase quantity consumption data of the advertisement putting platform, and adjusting the advertisement putting strategy in real time. Advertisement putting personnel can obtain advertisement income refined to channels, namely different platforms according to the obtained calculation result, newly-added user ROI under different plans can be obtained by combining the advertisement putting platform data and the polymerization platform data, advertisement position putting evaluation is carried out according to the newly-added user ROI data of each platform, and accurate automatic advertisement putting is realized, namely the number of times of platform advertisement putting is increased or reduced through the ROI data.
The method calculates the advertisement dotting event data and the advertisement showing platform data by using a revenue reverse-pushing method to obtain historical information of thousands of display profits (eCPM) of advertisement code positions, constructs a revenue prediction method to realize the calculation of T + 0-level revenue, combines the purchase quantity consumption data provided by an advertisement putting platform to give a prediction result of ROI (region of interest) to guide advertisement putting personnel to carry out real-time data decision, realizes the advertisement putting with higher efficiency, has good accuracy, reduces the advertisement putting cost when striving for a high input-output ratio, and ensures the advertisement putting effect.
Furthermore, the display times of the advertisement code positions are determined according to the advertisement dotting event data, the income of the advertisement code positions is determined according to the advertisement showing platform data, and the display times and the income of the advertisement code positions are correlated to obtain thousands of display profits of the advertisement code positions.
Furthermore, the aggregation platform comprises data with advertisement code positions and data without advertisement code positions, and the number of times of advertisement display of the aggregation platform is distributed by using an integer optimization method.
And when the income is calculated reversely, dividing the advertisement dotting event data into aggregation platform data and non-aggregation platform data, and respectively processing the aggregation platform data and the non-aggregation platform data. And counting the display times of all advertisement code positions on the day according to the data of the non-aggregation platform, associating the display times with the revenue of the corresponding advertisement code positions counted by the advertisement showing platform, and calculating the thousands of display benefits of the advertisement code positions of the non-aggregation platform.
For the data of the aggregation platform, because the aggregation platform comprises data of advertisement code positions and data without the advertisement code positions, part of the code positions in the advertisement dotting event cannot be matched with the code positions counted by the advertisement showing platform, only the code positions of the aggregation platform are given, the corresponding code position income information cannot be found, and the loss of income counting can be caused by direct matching. Here, the statistics of revenue of code-slot-free ad-dotting events are solved through an integer optimization algorithm using aggregate platform-related ad slot presentation statistics. The distributed advertisement display data of the aggregation platform can be matched with corresponding advertisement income of the advertisement showing platform, thousands of display profits of a single code position of the advertisement dotting event of the aggregation platform are obtained, and for the display data lacking the code position, the display data are averaged to a single display according to the distribution result, so that the calculation of the thousands of display profits of all the advertisement display events is realized.
Furthermore, the integer optimization method sets constraint conditions and performs optimization calculation on an objective function, wherein the objective function comprises a loss term and a regular term, and the loss term calculation enables the ratio of the advertisement display times of each code position after the distribution of the advertisement dotting events is completed to be closest to the ratio of the advertisement display times given by the aggregation platform; and the regular term calculation enables the absolute deviation between the advertisement space display times and the display times given by the aggregation platform to be as small as possible when the target function loss term meets the requirement.
Further, the integer optimization method has the constraint conditions that:
the advertisement display times distributed to each known code bit are more than or equal to zero;
the sum of the display times of the distributed advertisement code positions is equal to the advertisement display times of the aggregation platform that the advertisement positions cannot be matched with the income;
the ratio of the advertisement display times of the single code bit after the distribution and the advertisement display times given by the aggregation platform is less than 1+ epsilon, and epsilon is greater than 0.
Further, for the real-time advertisement with the advertisement code position on the aggregation platform, displaying the income for thousands of times, and calculating by using an average value; and calculating the thousands of display gains of the real-time advertisements of the aggregation platform without advertisement code positions or the thousands of display gains of the real-time advertisements of the non-aggregation platform by using an exponential smoothing method.
Further, the calculation formula of the exponential smoothing method is as follows:
wherein the content of the first and second substances,
representing that ad slot on the day shows revenue thousands of times,
representing thousands of impressions of the ad slot two days ago,
indicating that the ad slot shows revenue thousands of times a day ago.
And matching historical advertisement code position information obtained by using a revenue back-stepping algorithm with real-time advertisement event dotting data displayed on the same day to obtain revenue predicted values of all purchase quantity levels.
And if the advertisement display event is a non-aggregation platform, extracting thousands of display profits obtained by a reverse-pushing algorithm of the corresponding advertisement code position, and calculating thousands of display profits of the advertisement position of the non-aggregation platform on the same day.
If the advertisement display event is an aggregation platform and the actually displayed advertisement code bit exists, the prediction of thousands of display yields can be carried out by using the part of the day data pulled from the aggregation platform. And if the advertisement display event is an aggregation platform but no actually displayed advertisement code bit exists, processing by referring to a non-aggregation platform. The partial advertisement code slots can be obtained through simple statistics in the revenue back-stepping algorithm.
Further, the real-time advertisement display income of the newly increased users on the day is divided by the purchase quantity consumption data of the advertisement putting platform, so that the real-time investment return rate of the newly increased users on the day is obtained, and the purchase quantity consumption data of the advertisement putting platform is classified according to advertisement putting channels or advertisement putting plans.
Furthermore, the real-time advertisement display income of the newly added user on the current day is calculated by predicting thousands of display incomes through all advertisement display event data of the user and the advertisement code bit corresponding to the advertisement.
After the prediction income on the day is calculated, the real-time ROI of the newly added user on the day can be calculated by means of the advertisement dotting event, the attribution of the user and the purchase amount consumption cost data provided by the advertisement putting platform.
And aggregating to obtain the prediction income of the corresponding level according to the code position information in the multi-level advertisement dotting events such as the user-plan-channel on the day, and obtaining the purchase amount cost according to the consumption data provided by the advertisement putting platform. And dividing the prediction income by the purchase quantity cost to obtain the prediction return on investment of the advertisement delivery on the same day. And the advertisement delivery personnel creates, screens and adjusts the advertisement delivery plan in real time based on the calculated real-time investment return rate of the newly-added users on the same day, so that the ROI maximization is realized.
The system comprises a revenue reverse-pushing unit, a revenue prediction unit and an ROI (region of interest) calculation unit, wherein the revenue reverse-pushing unit calculates the display income of a historical advertisement space according to data acquired by the system, the revenue prediction unit calculates the real-time advertisement display income, and the ROI calculation unit calculates the real-time investment return rate of a newly-added user on the day and adjusts an advertisement putting strategy in real time.
The advertisement putting method calculates the display income of different showing advertisement position types, and corrects the calculated display income by using the advertisement dotting time data to obtain the prediction income of the advertisement position. And then correlating with the purchase quantity data of an advertisement putting platform, calculating the actual income and the return on investment of new and old users in different channels on the purchase quantity side, realizing intelligent advertisement putting by the advertisement putting platform, increasing or continuously putting advertisements for advertisement positions with high return on investment, reducing advertisement putting for advertisement positions with low return on investment, adjusting an advertisement putting strategy in real time, advancing an advertisement putting decision from T +1 to T +0, and improving timeliness.
3. Advantageous effects
Compared with the prior art, the invention has the advantages that:
the advertisement putting method simplifies decision indexes, does not need to specifically analyze filling and clicking conditions of each type of advertisement display, rapidly grasps the user quality obtained by the current advertisement putting through a reverse income pushing and income forecasting method, has high data calculation accuracy, and has the advantages that the difference between income data calculated by the advertisement putting method and actual income data is about 5 percent through limited data analysis, and the calculation accuracy is high. The system automatically adjusts the advertisement putting according to the user ROI calculated by the method, evaluates the advertisement putting according to ROI data, further increases or decreases the putting quantity, automatically corrects the advertisement putting mode and realizes the intelligent advertisement putting.
The method directly uses income and cost to make advertisement putting decision, converts complex event statistical indexes into core variables capable of being measured by financial data, and can directly reflect the good and bad effect of advertisement putting; the advertisement delivery method calculates the return on investment of the advertisement space by a revenue reverse-pushing and revenue prediction method, further adjusts the advertisement delivery strategy in real time, advances the decision from T +1 to T +0, improves the decision timeliness, enables the advertisement delivery to face a changeable flow environment, can quickly and effectively find and solve the problem as far as possible, and improves the efficiency of the whole product operation decision.
Drawings
FIG. 1 is a schematic flow diagram of a revenue regression method of the present invention;
FIG. 2 is a schematic flow chart of a revenue prediction method of the present invention;
FIG. 3 is a schematic diagram of the system of the present invention.
Detailed Description
The invention is described in detail below with reference to the drawings and specific examples.
Examples
Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
First, terms appearing in the description of the embodiments will be explained.
eCPM: the thousand-time display income refers to the advertisement income which can be obtained by displaying every thousand times, and the display unit can be a webpage, an advertisement unit or even a single advertisement. By default, eCPM refers to revenue from thousands of web page presentations, i.e., revenue per thousand ads. eCPM is only used as a parameter to reflect the profitability of a website and does not represent actual revenue.
Advertisement dotting event data: also called as the event data of embedded point, a string of codes are embedded at a certain point, in App, if a user triggers an advertisement, a piece of event data of dotting is returned, and the user knows that the click and the view of the advertisement are generated.
An advertisement putting platform: the platform for advertising the products needing to be obtained can be used for dividing the granularity of channels, plan groups, plans and the like and obtaining the purchase quantity consumption data of different granularities in real time.
Advertisement change platform: a rendition platform using the IAA advertisement rendition mode may be classified as an aggregation platform/non-aggregation platform. The non-aggregation platform can provide advertising revenue on different real advertising positions, and the aggregation platform can provide display data of different aggregation advertising positions but cannot be related to the purchase quantity channels, plans and other levels of the advertising platform.
Polymerization platform/non-polymerization platform: the aggregation platform is a third-party platform for providing advertising service and profit opportunities for IOS and Android mobile terminal developers, and by concentrating the entrances of a plurality of advertising channels into one SDK and combining with the data optimization strategy of the developers, the developers can simultaneously integrate a plurality of advertising resources and define a matching strategy, so that the developers are helped to rapidly promote advertising revenue. In a simple way, in the aggregation platform, a large advertisement space can be randomly split into a plurality of small advertisement spaces, and the aggregated advertisement space is connected with a common advertisement space for advertisement display according to a platform algorithm every time the advertisement is displayed. When the advertisement is put, the proportion between the large advertisement space and the small advertisement space and the proportion between the number of the advertisements and the advertisement platform need to be controlled.
The advertisement delivery system based on the real-time prediction ROI described in this embodiment, as shown in fig. 3, includes a revenue reverse-pushing unit 100, a revenue prediction unit 200, and a ROI calculation unit 300, where the system acquisition data is calculated by the revenue reverse-pushing unit 100, the revenue prediction unit 200, and the ROI calculation unit 300 in sequence, the revenue reverse-pushing unit 100 is configured to calculate eCPM information of a historical advertisement code position, the revenue prediction unit 200 calculates eCPM information of an advertisement code position on the same day, and the ROI calculation unit 300 calculates actual revenue and return on investment rate of different channels and different users, and guides an advertisement delivery strategy.
Aiming at the problems that the advertisement delivery effect on each platform cannot be known accurately in real time, more delivery cost is needed for better obtaining the delivery effect, the input-output ratio is low, the user delivery time is not clear, and the like, the embodiment discloses a more scientific and efficient advertisement delivery method based on the system, the real-time ROI of each newly added user on each platform can be obtained through revenue reverse pushing, revenue forecasting and ROI calculation, the advertisement delivery strategy of each platform is automatically adjusted in real time, the delivery cost is reduced when the high ROI of advertisement delivery is obtained, and the method is specifically explained below.
Step 1: the revenue reverse-pushing unit 100 calculates the advertisement dotting event data and the advertisement showing platform data by using a revenue reverse-pushing method, and obtains the advertisement code position thousands of display revenue in the historical data.
The system collects and obtains the advertisement dotting event data, the advertisement showing platform data and the advertisement aggregation platform display data in the application, and stores the data in the database so as to be convenient for the subsequent calling of the data.
As shown in fig. 1, a revenue backspace method is constructed by advertisement dotting event data and advertisement performance change platform data, so that backspace of specific advertisement code positions eCPM and user-level revenue calculation are realized, and historical advertisement code position eCPM information is obtained. According to different properties of the advertisement positions, the advertisement dotting event data is divided into aggregation platform data and non-aggregation platform data, and calculation is performed respectively.
For non-aggregation platform data, namely common and conventional advertisement slots, counting the display times of all advertisement code slots on the day, associating the display times with the corresponding advertisement code revenue counted by the advertisement showing platform, and calculating to obtain the eCPM of the non-aggregation platform advertisement code slots.
For the data of the aggregation platform, as the data of the aggregation platform comprises data of advertisement code positions and data without the advertisement code positions, part of the code positions in the data of the advertisement dotting events cannot be matched with the code positions counted by the advertisement showing platform, the advertisement showing platform only gives the code positions of the aggregation platform, corresponding code position income information cannot be found, and the direct matching can cause the loss of advertisement income counting.
And at the moment, the times of displaying the advertisement positions of the aggregation platform in the advertisement position display statistical information of the aggregation platform are used, the distributed times of displaying the advertisements are calculated through an integer optimization algorithm, the eCPM of the advertisement code positions of the aggregation platform is further calculated according to the calculated times of displaying the distributed advertisement code positions, and the predicted eCPM of the code position-free advertisement dotting events of the aggregation platform is obtained.
When the number of times of displaying the advertisement positions of the aggregation platform is calculated by using an integer optimization method, setting the following constraint conditions according to business requirements and basic logic:
the advertisement display times distributed to each known code bit are more than or equal to zero and are integers;
the sum of the display times of the distributed advertisement code positions is equal to the advertisement display times of the aggregation platform that the advertisement positions cannot be matched with the income;
the ratio of the number of advertisement display times of a single code bit after completion of allocation to the number of advertisement display times given by the aggregation platform needs to be less than 1+ epsilon, epsilon >0, and epsilon is generally required to be as small as possible as a relaxation condition of a feasible solution of the target problem, where epsilon =0.5 in this embodiment.
The objective function comprises a loss term and a regular term, and the optimization goal of the loss term is as follows: and the ratio of the advertisement display times of each code position after the distribution of the advertisement dotting events is closest to the ratio of the advertisement display times given by the aggregation platform. The regularization term optimization objective is: and under the condition that the loss item is optimal, the absolute deviation between the advertisement space display times in the feasible solution and the display times given by the aggregation platform is as small as possible. To ensure that the objective function is differentiable, a regular form of L2 is used.
And when the optimization target is calculated, the advertisement display times ratio of each code position after the distribution of the advertisement dotting events is completed is closest to the advertisement display times ratio given by the aggregation platform, and the advertisement display times are kept as equal as possible. Under the condition of ensuring that the display times of the advertisement code positions are reasonable, the deviation of two statistical apertures is minimized as much as possible, and the probability of existence of a feasible solution is improved. And solving the integer optimization problem by using a gradient descent algorithm, and distributing the display times of the real advertisement code bit which cannot be found.
And matching the distributed advertisement display data of the aggregation platform with corresponding advertisement revenue of the advertisement showing platform to obtain eCPM of a single code position of the advertisement dotting event of the aggregation platform, and distributing the display data of the lack of the code position of the aggregation platform to single display according to a distribution result to realize the calculation of eCPM of all the advertisement showing events.
After the eCPM information of the historical advertisement code positions is obtained, the income of the advertisement displayed at a single time is calculated according to the advertisement dotting event data of each advertisement code position, the advertisement showing income at the user level is obtained through aggregation at the user level, the advertisement dotting data is further used for calculation to obtain the income data according to different advertisement putting channels or advertisement putting plan groups and the like, and the historical income data is calculated and analyzed at the plan level.
Step 2: the revenue prediction unit 200 calculates real-time advertisement display revenue according to the display revenue of the historical advertisement code slots calculated by the revenue reverse pushing unit 100 and the real-time advertisement dotting event data.
As shown in fig. 2, revenue prediction is performed using real-time advertisement dotting event data and historical advertisement code position eCPM information, and the real-time advertisement display revenue of the day is calculated. And matching the historical eCPM information of the advertisement code position obtained by using a revenue reverse-pushing method with the real-time dotting event data displayed by the advertisement on the day to obtain the advertisement display revenue predicted value of each purchase quantity level.
If the advertisement is displayed on the non-aggregation platform, extracting eCPM information of the advertisement code position obtained by a first-day backward-pushing method of the corresponding advertisement code position on T-2 days (namely two days ago) and T-1 days (namely one day ago), and calculating the eCPM of the non-aggregation platform advertisement position on the current day by using an exponential smoothing method. The specific formula is as follows:
wherein the content of the first and second substances,
a non-syndicated platform ad spot eCPM representing the day,
representing a non-syndicated platform ad spot eCPM two days ago,
representing a day's non-syndicated platform ad slot eCPM.
If the advertisement presentation event is an aggregation platform and there is a really presented advertisement code space, the prediction of the advertisement code space eCPM can be entered using the part of the day data pulled from the aggregation platform
. In order to ensure higher data coverage rate, data of the corresponding advertisement code bit in the current day and the T-2 day are extracted, the average eCPM of nearly three days is taken as a calculation parameter for each actually displayed advertisement code bit, and the specific formula is as follows:
wherein the content of the first and second substances,
representing the aggregation platform real advertisement code bits eCPM on the day,
representing the current day two days ago the syndication platform real ad code site eCPM,
representing the current day of the day, the syndication platform real ad code site eCPM.
If the advertisement display event is an aggregation platform, but no actually displayed advertisement code position exists, prediction is carried out by referring to a non-aggregation platform computing mode, and eCPM of the advertisement code position can be obtained through simple statistics in a revenue reverse-pushing method.
According to the calculation result, the user who obtains the advertisement by putting the advertisement on the day can be obtained, and the corresponding predicted eCPM is displayed on the advertisement every time for subsequent aggregation calculation. Compatible interfaces are reserved in the income reverse-pushing unit 100 and the income prediction unit 200, so that different types of subsequently-appearing advertisement showing platforms can be accessed to the part for prediction calculation according to the characteristics of the platforms.
And step 3: the ROI calculating unit 300 calculates a real-time investment return rate of the newly added user on the current day according to the purchase amount consumption data of the advertisement delivery platform, and adjusts the advertisement delivery strategy in real time.
The ROI calculating unit 300 realizes prediction of ROI and guides advertisement delivery personnel to make decisions by combining attribution information and third-party platform data. After the revenue prediction unit 200 calculates the expected display revenue of the current day, the real-time ROI of the newly added user of the current day can be calculated by the advertisement dotting event, the attribution of the user and the purchase amount consumption data provided by the advertisement putting platform, so as to further realize the control of the advertisement putting.
And according to all advertisement display event data of the user, correlating the advertisement display event data with the advertisement code bit prediction eCPM corresponding to the advertisement display, and calculating the prediction income accumulated by the newly added user to the current moment. And determining the source channel of the newly added user on the same day through the attribution information, and calculating the acquisition cost of the newly added user in different channels and different plans, namely the purchase consumption data according to the purchase consumption data provided by the advertisement putting platform. The predicted revenue is divided by the purchase consumption to obtain the real-time predicted ROI. And the ROI calculating unit is used for automatically adjusting the delivery strategy in real time according to the calculated ROI, guiding delivery personnel to adjust the purchase quantity plan in real time according to the index calculated by the ROI calculating unit, scientifically and accurately performing advertisement delivery arrangement and adjusting the advertisement delivery plan. The advertisement putting personnel can obtain the advertisement putting effect in real time, the putting cost is reduced, and the whole advertisement putting efficiency is improved.
In the embodiment, based on the calculation of the real-time prediction advertisement ROI, the income and the cost are directly used for making an advertisement putting decision, and advertisement putting personnel can create, screen and adjust an putting plan in real time according to the condition of the prediction ROI.
The invention and its embodiments have been described above schematically, without limitation, and the invention can be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The representation in the drawings is only one of the embodiments of the invention, the actual construction is not limited thereto, and any reference signs in the claims shall not limit the claims concerned. Therefore, if a person skilled in the art receives the teachings of the present invention, without inventive design, a similar structure and an embodiment to the above technical solution should be covered by the protection scope of the present patent. Furthermore, the word "comprising" does not exclude other elements or steps, and the word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. Several of the elements recited in the product claims may also be implemented by one element in software or hardware. The terms first, second, etc. are used to denote names, but not any particular order.
Claims (10)
1. An advertisement putting method based on real-time prediction ROI is characterized in that advertisement dotting event data, advertisement putting platform data, advertisement showing platform data and advertisement aggregation platform data are obtained, and the advertisement dotting event data, the advertisement putting platform data and the advertisement showing platform data are calculated to obtain thousands of display benefits of historical advertisement code positions;
and calculating real-time advertisement display income according to thousands of display income of the historical advertisement code positions and real-time advertisement dotting event data, calculating the real-time investment return rate of newly added users on the day according to the purchase quantity consumption data of the advertisement putting platform, and adjusting the advertisement putting strategy in real time.
2. The method of claim 1, wherein the number of times of displaying the advertisement code slot is determined according to the advertisement dotting event data, the revenue of the advertisement code slot is determined according to the advertisement presentation platform data, and the number of times of displaying the advertisement code slot is associated with the revenue to obtain the thousand-time display revenue of the advertisement code slot.
3. The method of claim 2, wherein the aggregation platform comprises data with advertisement code slots and data without advertisement code slots, and the number of times of advertisement display on the aggregation platform is distributed by using an integer optimization method.
4. The method of claim 3, wherein the integer optimization method sets constraints and performs optimization calculation on an objective function, the objective function comprises a loss term and a regular term, and the loss term calculation enables the ratio of the advertisement display times of each code location after completion of distribution of the advertisement dotting events to be closest to the ratio of the advertisement display times of the aggregation platform; and the regular term calculation enables the absolute deviation between the advertisement space display times and the display times given by the aggregation platform to be as small as possible when the target function loss term meets the requirement.
5. The method of claim 4, wherein the integer optimization method constraints are:
the advertisement display times distributed to each known code bit are more than or equal to zero;
the sum of the display times of the distributed advertisement code positions is equal to the advertisement display times of the aggregation platform that the advertisement positions cannot be matched with the income;
the ratio of the advertisement display times of the single code bit after the distribution and the advertisement display times given by the aggregation platform is less than 1+ epsilon, and epsilon is greater than 0.
6. The method of claim 1, wherein for thousands of revenues of real-time advertising with code slots on a syndication platform, average calculation is used; and calculating the thousands of display gains of the real-time advertisements of the aggregation platform without advertisement code positions or the thousands of display gains of the real-time advertisements of the non-aggregation platform by using an exponential smoothing method.
7. The method of claim 6, wherein the exponential smoothing method is calculated by the following formula:
wherein the content of the first and second substances,
representing that ad slot on the day shows revenue thousands of times,
representing thousands of impressions of the ad slot two days ago,
indicating that the ad slot shows revenue thousands of times a day ago.
8. The method of claim 1, wherein the real-time advertisement display revenue of the newly added users on the same day is divided by the consumption data of the purchased volume of the advertisement delivery platform to obtain the real-time return rate of investment of the newly added users on the same day, and the consumption data of the purchased volume of the advertisement delivery platform is classified according to advertisement delivery channels or advertisement delivery plans.
9. The method of claim 8, wherein the real-time advertisement display revenue of the newly added user on the same day is calculated by predicting thousands of display revenue through all advertisement display event data of the user and the advertisement code bit corresponding to the advertisement.
10. An advertisement delivery system based on real-time prediction ROI, characterized in that the advertisement delivery method based on real-time prediction ROI is used, the system comprises a revenue reverse pushing unit, a revenue prediction unit and an ROI calculation unit, the revenue reverse pushing unit calculates the display income of a historical advertisement slot according to the data acquired by the system, the revenue prediction unit calculates the real-time advertisement display income, the ROI calculation unit calculates the real-time investment return rate of the newly added user on the day, and the advertisement delivery strategy is adjusted in real time.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110698938.5A CN113269595B (en) | 2021-06-23 | 2021-06-23 | Advertisement delivery method and system based on real-time prediction ROI |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110698938.5A CN113269595B (en) | 2021-06-23 | 2021-06-23 | Advertisement delivery method and system based on real-time prediction ROI |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113269595A true CN113269595A (en) | 2021-08-17 |
| CN113269595B CN113269595B (en) | 2022-01-11 |
Family
ID=77235778
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110698938.5A Active CN113269595B (en) | 2021-06-23 | 2021-06-23 | Advertisement delivery method and system based on real-time prediction ROI |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113269595B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114266604A (en) * | 2022-01-04 | 2022-04-01 | 北京小熊博望科技有限公司 | Advertisement putting effect evaluation method based on multiple dimensions and storage medium |
| CN115456685A (en) * | 2022-09-23 | 2022-12-09 | 广州唯彩会网络科技有限公司 | Real-time purchase quantity optimization system, method, terminal and computer readable storage medium |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140089106A1 (en) * | 2012-09-27 | 2014-03-27 | Yahoo! Inc. | Method and system for formulating bids for internet advertising using forecast data |
| CN105809458A (en) * | 2014-12-29 | 2016-07-27 | 苏宁云商集团股份有限公司 | Advertisement accurate delivery method and system in e-commerce site |
| CN107067274A (en) * | 2016-12-27 | 2017-08-18 | 北京掌阔移动传媒科技有限公司 | One DSP real time bid ad system based on blended learning model |
| CN110619548A (en) * | 2019-09-20 | 2019-12-27 | 腾讯科技（深圳）有限公司 | Method, device, equipment and storage medium for determining media content delivery strategy |
| CN111080354A (en) * | 2019-12-10 | 2020-04-28 | 杭州古点网络科技有限公司 | Intelligent mobile advertisement system for dynamically selecting advertisers |
| CN111178981A (en) * | 2020-01-02 | 2020-05-19 | 众安在线财产保险股份有限公司 | Advertisement putting method and device, computer equipment and storage medium |
| CN111510746A (en) * | 2019-01-30 | 2020-08-07 | 腾讯科技（深圳）有限公司 | Media resource delivery method and device, storage medium and electronic device |
| CN111737646A (en) * | 2020-05-15 | 2020-10-02 | 广州小迈网络科技有限公司 | Advertisement promotion effect evaluation data processing method, system and storage medium |
| CN112150219A (en) * | 2020-10-22 | 2020-12-29 | 广州万唯邑众信息科技有限公司 | Online advertisement delivery system and use method thereof |
-
2021
- 2021-06-23 CN CN202110698938.5A patent/CN113269595B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140089106A1 (en) * | 2012-09-27 | 2014-03-27 | Yahoo! Inc. | Method and system for formulating bids for internet advertising using forecast data |
| CN105809458A (en) * | 2014-12-29 | 2016-07-27 | 苏宁云商集团股份有限公司 | Advertisement accurate delivery method and system in e-commerce site |
| CN107067274A (en) * | 2016-12-27 | 2017-08-18 | 北京掌阔移动传媒科技有限公司 | One DSP real time bid ad system based on blended learning model |
| CN111510746A (en) * | 2019-01-30 | 2020-08-07 | 腾讯科技（深圳）有限公司 | Media resource delivery method and device, storage medium and electronic device |
| CN110619548A (en) * | 2019-09-20 | 2019-12-27 | 腾讯科技（深圳）有限公司 | Method, device, equipment and storage medium for determining media content delivery strategy |
| CN111080354A (en) * | 2019-12-10 | 2020-04-28 | 杭州古点网络科技有限公司 | Intelligent mobile advertisement system for dynamically selecting advertisers |
| CN111178981A (en) * | 2020-01-02 | 2020-05-19 | 众安在线财产保险股份有限公司 | Advertisement putting method and device, computer equipment and storage medium |
| CN111737646A (en) * | 2020-05-15 | 2020-10-02 | 广州小迈网络科技有限公司 | Advertisement promotion effect evaluation data processing method, system and storage medium |
| CN112150219A (en) * | 2020-10-22 | 2020-12-29 | 广州万唯邑众信息科技有限公司 | Online advertisement delivery system and use method thereof |
Non-Patent Citations (2)
| Title |
|---|
| 任子翰: "广告投放评估系统的设计与实现", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 * |
| 刘华玲: "网络广告点击率预估的特征学习及技术研究进展", 《浙江大学学报(理学版)》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114266604A (en) * | 2022-01-04 | 2022-04-01 | 北京小熊博望科技有限公司 | Advertisement putting effect evaluation method based on multiple dimensions and storage medium |
| CN115456685A (en) * | 2022-09-23 | 2022-12-09 | 广州唯彩会网络科技有限公司 | Real-time purchase quantity optimization system, method, terminal and computer readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113269595B (en) | 2022-01-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106960359B (en) | Full-automatic bidding optimization method and system based on streaming calculation | |
| CN110991875B (en) | Platform user quality evaluation system | |
| CN113269595B (en) | Advertisement delivery method and system based on real-time prediction ROI | |
| Han et al. | Segmentation of telecom customers based on customer value by decision tree model | |
| CN106156878B (en) | Advertisement click rate correction method and device | |
| US20150134462A1 (en) | Dynamic bid price selection | |
| US20120323674A1 (en) | Creation and usage of synthetic user identifiers within an advertisement placement facility | |
| US10445789B2 (en) | Segment-based floors for use in online ad auctioning techniques | |
| CN101483531A (en) | Method and system for prediction type account balance alarm | |
| US20150134463A1 (en) | Performance metric based bid placement | |
| Dullaghan et al. | Integration of machine learning techniques to evaluate dynamic customer segmentation analysis for mobile customers | |
| US20150112795A1 (en) | Efficient budget pacing | |
| CN108154379B (en) | Media information publishing method and device | |
| CN111080354A (en) | Intelligent mobile advertisement system for dynamically selecting advertisers | |
| JP2016516239A (en) | Method and system for using consumer aliases and identifiers | |
| CN106296247A (en) | The online sort method of network information resource and device | |
| CN106408325A (en) | User consumption behavior prediction analysis method based on user payment information and system | |
| CN111737646A (en) | Advertisement promotion effect evaluation data processing method, system and storage medium | |
| CN112465549A (en) | System and method for identifying channel cheating | |
| US20140257938A1 (en) | System and method for analyzing hospitality industry data and providing analytical performance management tools | |
| Mori | It’s not price; It’s quality. Satisfaction and price fairness perception | |
| US20170186047A1 (en) | Optimization of audience groups in online advertising bidding | |
| US20220374923A1 (en) | Computational methods and processor systems for predictive marketing analysis | |
| CN113139826A (en) | Method and device for determining distribution authority of advertisement space and computer equipment | |
| Weber | Methods to forecast television viewing patterns for target audiences |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |