Automatic bidding implementation method for motivating advertiser to really report ROI (Region of Interest) constraint

Abstract

The invention discloses an automatic bidding implementation method for motivating an advertiser to really report ROI constraints. According to the method, an estimated value and ROI constraints of an advertiser are collected before an automatic bidding cycle, an advertisement platform designs a cumulative delivery mechanism based on historical data and experience distribution, and automatic bidding is coordinated through internal and external strategies in each auction; wherein the internal strategy dynamically distributes conversion and a cost target, and the external strategy adjusts the number of advertisement positions through a PID controller to ensure the achievement of an accumulative putting target. According to the method, strategy optimization is carried out on the automatic bidding side of the advertiser, the advertiser can be motivated to truly report self constraints, and the method is suitable for an advertisement auction market where the auction frequency is high and the advertiser has ROI constraints; compared with a traditional automatic bidding scene through auction model modification, the efficiency of the advertisement system can be improved, and the overall social welfare of the advertiser, the advertisement platform and the user can be improved.

Description

Technical Field

[0001] This invention relates to the field of internet computing advertising technology, specifically to an automatic bidding method that incentivizes advertisers to truthfully report ROI constraints. Background Technology

[0002] In the field of internet computing advertising, online ad auctions are the mainstream resource allocation model. In the traditional model, advertisers need to manually bid on a massive number of ad display opportunities in real time. This is not only cumbersome but also places extremely high demands on advertisers' bidding strategy capabilities. Especially for small and medium-sized enterprises lacking professional marketing teams, manual bidding constitutes a huge operational burden.

[0003] To lower the operational barrier and improve advertising efficiency, automated bidding products have emerged and become the industry mainstream. In this model, advertisers no longer need to meticulously bid for each auction; instead, they set high-level marketing goals and budget constraints. The advertising platform's automated bidding algorithm then breaks down these macro-level goals and generates real-time bids for each advertising auction. Products such as oCPM, oCPC, and target ROAS offered by mainstream platforms are typical examples of this model.

[0004] However, the rise of the automated bidding paradigm has also brought new challenges to mechanism design. In traditional advertising auctions, mechanism design mainly focuses on how to incentivize advertisers to truthfully report their click value in a single auction. In the automated bidding environment, the advertiser's role shifts from a "single bidder" to a "reporter of goals and constraints," and their private information expands from a simple "valuation" to complex preferences including "ROI constraints." If the automated bidding system lacks effective incentive guarantees, rational advertisers will tend to strategically report their ROI constraints in an attempt to manipulate the system to secure more favorable campaign results. This strategic reporting behavior leads to two serious consequences: first, the bidding decisions made by the advertising platform based on distorted target information will inevitably be suboptimal, resulting in a loss of overall advertising system efficiency; second, advertisers themselves will also need to incur additional costs to test and calculate the optimal reporting strategy, increasing decision-making costs.

[0005] Therefore, a significant drawback exists in existing technologies: traditional automated bidding systems primarily focus on optimizing bids under given constraints, neglecting potential strategic behaviors by advertisers during the reporting constraint phase and lacking an inherent mechanism to incentivize advertisers to provide truthful reports. This prevents the system from achieving global efficiency optimization based on authentic information, hindering further improvements in the overall social welfare of advertisers, platforms, and users. Summary of the Invention

[0006] In order to achieve optimal global efficiency based on real information, this invention provides an automatic bidding method that incentivizes advertisers to truthfully report ROI constraints.

[0007] The technical solution adopted in this invention is as follows: an automatic bidding method for incentivizing advertisers to truthfully report ROI constraints, comprising the following steps: Step S1. Before the start of the automatic bidding period, the advertising platform collects information from advertisers. Valuation of ad slots for this period Constraints on target return on investment (ROI) ; Step S2. Based on the advertiser's transaction history, competitors' experienced bidding distribution, user experience requirements, and platform revenue goals, the advertising platform designs an automatic bidding cycle cumulative placement mechanism for advertiser i. ; in, This indicates the expected cumulative conversion target within the period. This indicates the target unit conversion cost within the period, and the cumulative delivery mechanism. It is ROI approx. The function satisfies the incentive compatibility condition: Condition 1, during the bidding period, the advertiser's expected ROI is equal to the constraint value of their reported results. Condition two: The advertiser's expected utility function depends on the constraint value of its reported value. Monotonically non-increasing; Step S3. During the automatic bidding period, for each advertising auction request, execute the following automatic bidding sub-steps; S3.1. Evaluate the bid distribution of other advertisers based on historical data. ; S3.2. Invoke internal strategy Based on the accumulated advertising dataset Current cumulative deployment target Calculate the immediate conversion target for this auction. Real-time cost targets and the corresponding constraint slack ; S3.3. Construct and solve the bidding optimization problem to determine the optimal bid for this auction. Its mathematical model is,

[0008]

[0009]

[0010] in, and These are the allocation function and payment function of the advertising auction mechanism, respectively; S3.4. Submit a bid to the advertising auction system ; S3.5. Collect the results of this auction and update the advertising dataset. ; S3.6. Invoke external strategy Based on the updated and cumulative delivery target Dynamically adjust environmental parameters for subsequent auctions; Step S4. After the current automatic bidding period ends, the campaign performance is fed back to the advertiser, and the cumulative campaign mechanism and automatic bidding strategy are iterated and optimized in the next period based on the data from this period.

[0011] Preferred internal strategy The execution logic includes: based on the cumulative conversions up to the current auction. and accumulated costs The remaining conversion and cost targets for the current period will be dynamically allocated to each future auction; the allocation will employ one of the following two strategies: ① Convert the remaining target amount and remaining target cost Distribute equally to the remaining In the next auction; ② Based on the average allocation, a prediction error correction factor related to the auction progress is introduced. The allocation targets are dynamically adjusted.

[0012] Preferably, the constraint slack amount The calculation method is as follows: dynamic adjustment is made based on the cumulative deviation between the target conversion / cost and the actual conversion / cost in each historical auction round. Specifically, in the k+1 round auction,

[0013]

[0014] in, and For the first advertisers before the auction The target conversion rate and target payment, and make and For the first After the auction, the advertiser Actual conversions and payments; slack volume in the first round of auctions. Initialize to (0,0).

[0015] Preferred external strategy This is achieved using a PID controller. The specific execution logic includes: the input to the PID controller is based on the cumulative weighted error. It integrates the achievement of cumulative conversions and cumulative costs; the output of the PID controller dynamically adjusts the number of ad slots available for allocation in the next auction. By controlling the supply of ad slots, the cumulative campaign goals for the period can be achieved.

[0016] Preferred, The calculation formula is

[0017] in, It is the expected total number of auctions at the start of the cycle. and These are the weighting coefficients, and .

[0018] Preferably, the proportional term of the PID controller Integral terms and differential terms The parameter is set to .

[0019] The second technical solution adopted by the present invention is as follows: a computing device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the automatic bidding method for incentivizing advertisers to report true ROI constraints.

[0020] The third technical solution adopted in this invention is as follows: a computer-readable storage medium storing a computer program thereon, wherein when the program is executed by a processor, it implements the steps of the automatic bidding method for incentivizing advertisers to report true ROI constraints.

[0021] The present invention has the following beneficial effects: 1. Effective incentives for advertisers to report truthfully: Through a cumulative delivery mechanism that strictly meets incentive compatibility conditions, that is, the advertiser's expected utility does not increase monotonically with the reported ROI constraint value, the design ensures that for rational advertisers, reporting their true ROI constraint is their dominant strategy. This solves the problem of low system efficiency caused by advertisers strategically exaggerating constraints due to lack of incentives in the traditional automatic bidding environment. 2. Improve the overall delivery efficiency of the advertising system: Since advertisers are willing and inclined to report true constraint information, the advertising platform can optimize bidding based on real and reliable preference information. This enables the platform's automatic bidding strategy to more accurately match the advertiser's actual needs with market display opportunities. On a macro level, this greatly reduces resource mismatch caused by information distortion and significantly improves the delivery efficiency and stability of the entire advertising system. 3. Reduce the decision-making and operational costs for advertisers: It eliminates the extensive and complex calculations and trial-and-error process that advertisers need to go through in order to find the optimal reporting strategy. Advertisers only need to report their true marketing goals to automatically achieve the best campaign results. This simplifies the advertiser's operational process, reduces their decision-making costs and mental burden, and is especially beneficial to small and medium-sized enterprises that lack professional marketing teams. 4. Ensuring the achievement of periodic goals: The two-layer framework, which combines the inner real-time bidding strategy with the outer macro-control strategy, can dynamically and reasonably decompose long-term, cumulative ROI constraints into bidding instructions for a single auction through the internal strategy. At the same time, the PID controller in the external strategy can make macro-fine adjustments to the market environment, making the system more robust and precise in achieving periodic constraints. It can effectively cope with market fluctuations and stably guarantee the long-term returns for advertisers. 5. Achieving a win-win situation for advertisers, platforms, and users: For advertisers, they obtain stable and efficient advertising services based on their true intentions, maximizing marketing effectiveness; for advertising platforms, the overall optimal allocation of traffic is achieved based on real information, improving platform revenue and ecosystem health; for users, improved system efficiency means more relevant and better-experienced advertising content. Attached Figure Description

[0022] Figure 1 This is a schematic diagram illustrating the interaction relationships among participating entities in an automatic bidding environment according to an embodiment of the present invention.

[0023] Figure 2 This is a schematic diagram of a single advertising auction process within an automatic bidding cycle in an embodiment of the present invention.

[0024] Figure 3 This is a diagram of a two-layer automatic bidding framework implemented in this embodiment of the invention to incentivize advertisers to express true ROI constraints. Detailed Implementation

[0025] The present invention will be further described below with reference to the embodiments and accompanying drawings.

[0026] The core of this invention lies in proposing an incentive-compatible two-layer automatic bidding framework, aiming to solve the problem in existing automatic bidding technologies where advertisers may strategically overstate their ROI constraints, leading to low overall system efficiency. The overall working environment and process of the system are first described below with reference to the accompanying drawings, followed by a detailed explanation of the specific implementation details of this invention.

[0027] (a) Overall Interaction Relationships in the Automated Bidding Environment like Figure 1 As shown, the automatic bidding environment involved in this invention mainly includes three stages of interaction, forming a complete closed-loop system: Figure 1 -① Goal and Constraint Reporting Phase: Before the start of the automatic bidding cycle, advertisers express their private information to the advertising platform, including the valuation of the ad slots. and target return on investment (ROI) constraints This is the starting point of the entire process and the key point of application for the excitation-compatible design of this invention.

[0028] Figure 1 -② Automatic Bidding and Execution Phase: The advertising platform designs personalized automatic bidding strategies based on information reported by advertisers, historical transaction data, user experience requirements, and platform revenue targets. When a user query triggers an advertising demand, the platform uses this strategy to place a real-time bid on behalf of the advertiser and participates in the advertising auction.

[0029] Figure 1 -③ Performance Feedback and Iterative Cycle Phase: After the automatic bidding cycle ends, the advertising platform provides the advertiser with feedback on the cumulative advertising performance of that cycle. Advertisers can analyze the performance based on this feedback and resubmit or adjust their valuation and constraint information before the start of the next automatic bidding cycle, thus forming a closed loop of continuous learning and optimization.

[0030] (II) Detailed Process of a Single Advertising Auction This invention is in Figure 1 -② The specific, micro-level process of each advertising auction executed in phase 2 is as follows: Figure 2 As shown, the main steps include: 1. User query: Users initiate search or browsing behavior, generating a demand for advertising display.

[0031] 2. Recall and User Modeling: The advertising platform quickly filters out a set of legitimate and relevant candidate advertisers from a large number of advertisers based on user profiles and contextual information.

[0032] 3. Auction: Recalled advertisers participate in the auction. Advertisers using automatic bidding will have their bids... It is automatically generated by the algorithm of this invention.

[0033] 4. Allocation and Payment: The advertising platform allocates ad slots based on a pre-set advertising auction mechanism, according to the bids of all advertisers (allocation function). ) and payment settlement (payment functions) ).

[0034] 5. Ad Display: The advertiser who wins the auction will have their ad displayed in the designated ad space, completing the campaign.

[0035] (III) Implementation of a two-tiered automatic bidding framework that incentivizes advertisers to truthfully report ROI constraints Based on the aforementioned overall interaction and single process, the present invention proposes a two-layer automatic bidding framework (its overall structure is as follows). Figure 3 As shown, the core objective of incentivizing advertisers to provide authentic reports is achieved through the collaboration between the outer mechanism design and the inner real-time execution.

[0036] 1. Design and initialization of the outer cumulative delivery mechanism Before the start of the automatic bidding period (corresponding to) Figure 1 -①), the system performs the following key initialization steps: ① Information Collection: The advertising platform receives the ad slot valuation from advertiser i. Constraints with target ROI .

[0037] ② Mechanism Design: The advertising platform is based on the advertiser's transaction history and experience distribution of competitors' bidding. Based on estimates, user experience requirements, and the platform's own revenue goals, a periodic cumulative spending mechanism was designed for this advertiser. , This represents the platform's commitment to achieving the advertiser's expected cumulative conversion rate during this automatic bidding period. This represents the target unit conversion cost corresponding to the above conversion volume.

[0038] This mechanism satisfies the following conditions: feasibility requirement, and expected ROI equals reported value. Monotonicity condition: the advertiser's expected utility varies with their reported value. Monotonic and unchanging. This means that falsely reporting an overly strict (higher) ROI constraint will not bring advertisers higher expected returns, thus fundamentally eliminating their incentive to report strategically.

[0039] 2. Core execution flow of the inner real-time bidding algorithm During the automatic bidding period, for each triggered ad auction request (corresponding to...) Figure 2 In the "bidding" stage, this invention utilizes the following core algorithm (whose process corresponds to...) Figure 3The algorithm makes real-time bidding decisions through an inner loop. Under the macro-control of the external strategy (ECS), the algorithm repeatedly executes the internal strategy (ICS) to make micro-bidding decisions. The specific algorithm implementation is as follows.

[0040] Algorithm: An automatic bidding algorithm that implements a cumulative bidding mechanism; Input: Cumulative delivery mechanism ROI report for advertisers ; 1. Initialization: ; 2. While there are ad requests and the period has not ended, do; 3. Environmental Assessment: Evaluate the bidding distribution of other advertisers based on experience distribution, and set it as... ; 4. Internal Strategy (ICS) Decomposition of Objectives: Invoke internal strategy To determine the effectiveness of the target advertising campaign in this auction. and the corresponding relaxation amount ; 5. Solve for the optimal bid: Solve the following optimization problem to determine the bid for this auction. ,

[0041]

[0042]

[0043] in, and These are the allocation function and payment function of the advertising auction mechanism, respectively; 6. Execute a bid: Submit a bid to the auction. ; 7. Data Update: Collect the auction results and update the dataset. ; 8. External Strategy (ECS) Control: Invoking Adjust the subsequent auction environment parameters; 9. end while; 10. Return

[0044] Output: Ad campaign results dataset .

[0045] (iv) Specific implementation examples of internal and external strategies The internal policy (ICS) and external policy (ECS) in the above algorithm are the core of achieving fine-grained control, and their specific implementation is as follows.

[0046] 1. Detailed implementation of Internal Strategy (ICS) ICS is responsible for dynamically and rationally breaking down the periodic cumulative targets into each specific auction. and For the front The cumulative conversions and unit conversion costs achieved in this auction will inform the advertising platform's strategy for future auctions. The plan aims to converge to the cumulative investment target after the bidding period ends. .

[0047] ①Target decomposition: Calculate the first Immediate target of the next auction Two optional strategies are provided: Internal Strategy 1: Average Distribution Method

[0048]

[0049] in, This represents the estimated number of remaining auctions from the start of this auction to the end of the cycle. This strategy is simple and robust, aiming to evenly distribute remaining targets and budgets across the remaining auctions for smooth control.

[0050] Internal strategy two: Allocation method with schedule adjustment.

[0051]

[0052] in, It is the expected total number of auctions at the start of the cycle. This is the prediction error correction factor. Later in the cycle, this factor will be adjusted more aggressively or conservatively to adjust the individual target based on the progress of target achievement.

[0053] ② Calculation of relaxation amount: relaxation amount It is dynamically set based on the cumulative deviation between the target and actual values ​​in historical auction rounds, aiming to provide necessary flexibility for the optimization problem and avoid the bidding strategy from becoming locally rigid due to random fluctuations in a single auction. In the (k+1)th auction, the setting is as follows:

[0054]

[0055] in, and For the first advertisers before the auction The target conversion rate and target payment, and make and For the first After the auction, the advertiser Actual conversions and payments. Slack volume in the first round of auctions. Initialize to .

[0056] 2. Detailed implementation of External Policy (ECS) As a macro controller, the ECS's role is to ensure that the cumulative target for the entire cycle is successfully achieved. This embodiment uses a PID controller to achieve this purpose.

[0057] 1. Control Objective: To ensure that the advertiser's cumulative campaign performance reaches its target at the end of the bidding period. Able to approach the target set by the outer mechanism ; 2. Error signal: at the first Before the auction, the cumulative weighted error input to the PID controller. Defined as

[0058] This error signal integrates the progress achieved in both conversion rate and cost dimensions, with weighting coefficients... Adjustments can be made based on the platform's business strategy; 3. Control Output: The PID controller outputs based on the error. The calculation is performed, and its output is mapped through a hash function to dynamically adjust the number of ad slots in the next auction.

[0059] in It represents the maximum number of ad impressions that users can accept. By controlling the supply of ad slots, the degree of market competition can be adjusted at the macro level, thereby indirectly helping advertisers achieve their periodic goals. 4. Controller parameters: Proportional item Integral terms and differential terms The parameters can be set in this embodiment. The specific values ​​of these parameters can be further optimized through offline simulation experiments or online A / B testing to achieve the best control effect.

[0060] (v) End of cycle and system iteration When the automatic bidding period ends (corresponding to) Figure 1-③), the advertising platform will provide the advertiser with feedback on the final campaign results for this period. Simultaneously, the platform will utilize the new data generated during this period to iteratively optimize the design model of the outer-layer cumulative campaign mechanism and the control parameters of the inner-layer strategy, enabling the entire system to continuously learn and improve, thus adapting to the ever-changing market environment.

[0061] (vi) Summary Through the detailed description above, combined with the accompanying drawings, algorithm flow, and specific strategy embodiments, this invention clearly demonstrates a complete and implementable automated bidding solution, from macro-level system interaction to micro-level bidding decisions. This solution, through incentive-compatible mechanism design and a two-layer collaborative control strategy, not only achieves efficient real-time bidding but also fundamentally incentivizes advertisers to reveal their true ROI constraints, thereby addressing the core pain points of existing technologies and improving the overall social welfare of advertisers, platforms, and users.

[0062] Obviously, the above embodiments of the present invention are merely illustrative examples to illustrate the invention and are not intended to limit the implementation of the invention. Other obvious variations or modifications derived from the essential spirit of the invention still fall within the protection scope of the invention.

Claims

1. An automatic bidding implementation method for encouraging advertisers to report ROI constraints truthfully, characterized in that, The steps are as follows: Step S1. Before the start of the automated bidding cycle, the ad platform collects ad Ad slot valuation for the cycle With target return on investment (ROI) constraints ; Step S2. The advertising platform designs a cumulative delivery mechanism for advertiser i in an automatic bidding cycle based on the advertiser's transaction history, competitor's experience bidding distribution, user experience requirements, and platform revenue goals ; wherein, represents an expected cumulative conversion target within a period, represents a target unit conversion cost within a period, a cumulative bidding mechanism is a function of ROI and satisfies incentive compatibility conditions: condition one, within a bidding period, the expected ROI of the advertising principal for the bidding effect is equal to the reported constraint value ; condition two, the expected utility function of the advertising principal is monotonically non-increasing with the reported constraint value . Step S3. In the automatic bidding cycle, for each advertising auction request, the following automatic bidding sub-steps are performed; S3.

1. Assess other advertisers' bidding distribution from historical data ; S3.

2. invoke internal policy , according to the accumulated advertising data set , the current accumulated delivery target , the instantaneous conversion target of this auction is calculated , the instantaneous cost target , and the corresponding constraint relaxation amount ; S3.

3. Construct and solve the bid optimization problem to determine the optimal bid for this auction with the mathematical model being, wherein, and are the allocation function and the payment function of the advertisement auction mechanism, respectively; S3.

4. Submitting bids to an ad auction system ; S3.

5. Collect the results of this auction, update the ad serving dataset ; S3.

6. Call external policies , based on the updated and accumulated delivery targets , dynamically adjust the environment parameters of subsequent auctions; Step S4. After the current automatic bidding cycle ends, the delivery effect is fed back to the advertiser, and the cumulative delivery mechanism and the automatic bidding strategy are iteratively optimized in the next cycle based on the data of this cycle.

2. The method of claim 1, wherein the method further comprises: Internal strategy The execution logic includes: dynamically allocating the conversion target and cost target for the remaining period to each future auction according to the cumulative conversion and cumulative cost up to the current auction; the allocation is done in one of the following two strategies: ① convert the remaining target volume and the remaining target cost to the remaining auction; ii. Introducing a prediction error correction factor related to the auction progress on the basis of an equal distribution Dynamic adjustment of the target of distribution.

3. The method of claim 1, wherein the method further comprises: Constraint relaxation amount The calculation method is: based on the cumulative deviation between the target conversion / cost and the actual conversion / cost in each round of historical auction, specifically, in the k+1 round of auction, in, and For the first advertisers before the auction The target conversion rate and target payment, and make and For the first After the auction, the advertiser Actual conversion rate and payment; In the first round of auctions, the slack initialized to (0, 0).

4. The method of claim 1, wherein the method further comprises: External strategy The implementation is through a PID controller, and the specific execution logic includes: the input of the PID controller is the cumulative weighted error based on the cumulative conversion and the cumulative cost; the output of the PID controller is the number of ad slots available for distribution in the next auction , and the cumulative delivery target of the period is achieved by controlling the supply of ad slots.

5. The method of claim 4, wherein the method further comprises: The calculation formula is wherein, is the total number of auctions expected at the beginning of the period, and are weight coefficients, and .

6. The method of claim 4, wherein the method further comprises: The parameters of the proportional term , the integral term and the derivative term of the PID controller are set to .

7. A computing device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein, The processor executes the computer program to implement the steps of the automatic bidding implementation method for stimulating the real reporting of the ROI constraint of the advertiser as claimed in any one of claims 1 to 4.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the steps of the automatic bidding implementation method for stimulating the real reporting of the ROI constraint of the advertiser as claimed in any one of claims 1 to 4.

Images