Artificial intelligence techniques for bid optimization used for generating dynamic online content

Patent metadata

Field	Value
Publication number	US-12131350-B2
Application number	US-202117403702-A
Country	US
Kind code	B2
Filing date	Aug 16, 2021
Priority date	Nov 18, 2019
Publication date	Oct 29, 2024
Grant date	Oct 29, 2024

Abstract

Techniques are disclosed for real-time bidding (e.g., for dynamic online content placement) using an optimized final bid. The final bid is determined based on a predicted clearing price and an initial bid. The initial bid represents a value to a prospective content provider, and may be computed based on campaign information. The predicted clearing price is a predicted amount paid, and may be predicted using a model trained using historical winning bids data. The clearing price may be predicted using a quantile regression model, where the quantile can be selected to control bid aggressiveness. In some cases, the quantile is determined based on pacing in an overall campaign. Once the initial bid and the predicted clearing price are calculated, the final bid is computed based on the initial bid and the predicted clearing price.

First claim

The invention claimed is:

1. A method comprising:
receiving, by a bid generation computer system and from a content placement system, a bid request for a dynamic content placement opportunity;
identifying a pace of a campaign, the pace representing an amount intended to be spent per a time unit compared to an amount actually spent in a subset of the time unit;
determining, by the bid generation computer system, a quantile value, corresponding to a selected percentile of historical clearing price data, based on the identified pace of the campaign;
predicting, by the bid generation computer system, a clearing price for the bid request by applying a pretrained model to information associated with the received bid request, wherein the quantile value is a hyperparameter of the pretrained model used to control bidding aggressiveness and to reduce a likelihood of overbidding or underbidding, and wherein the pretrained model was trained on historical winning bids data identifying historical clearing prices corresponding to historical winning bids;
computing, by the bid generation computer system, a final bid based upon the predicted clearing price; and
communicating the final bid from the bid generation computer system to the content placement system, wherein communicating the final bid to the content placement system causes a primary content provider to provide a user device with access to dynamic online content, wherein the receiving, determining, selecting predicting, computing, and causing the final bid to be communicated are performed in less than about 50 milliseconds.

2. The method of claim 1 , further comprising:
receiving information for a winning bid for the dynamic content placement opportunity; and
updating the pretrained model based on the received information for the winning bid.

3. The method of claim 1 , wherein the pretrained model is a quantile regression model.

4. The method of claim 1 , wherein the dynamic content placement opportunity is triggered in response to a request from a user to the primary content provider for publisher content, and wherein the information associated with the bid request identifies the user, the publisher content or a type of dynamic content.

5. The method of claim 1 , wherein computing the final bid comprises:
selecting a lower price of an initial bid and the predicted clearing price as the final bid.

6. The method of claim 1 , wherein computing the final bid comprises:
computing a new price based upon a function of an initial bid and the predicted clearing price; and
indicating the new price as the final bid.

7. The method of claim 1 , further comprising:
selecting, by the bid generation computer system using a model selection vector, the pretrained model from a plurality of pretrained models.

8. The method of claim 1 , wherein the final bid is for an auction and the clearing price is further based on a type of the auction.

9. A computer system comprising:
a bid generation system comprising a network interface and processing hardware, wherein the network interface is communicatively coupled to a content placement system, wherein the processing hardware is configured for:
receiving, from the content placement system, a bid request for a dynamic content placement opportunity;
identifying a pace of a campaign, the pace representing an amount intended to be spent per a time unit compared to an amount actually spent in a subset of the time unit;
using the identified pace to determine a quantile value corresponding to a selected percentile of historical clearing price data;
predicting a clearing price for the bid request by applying a pretrained model to information associated with the bid request, wherein the quantile value is a hyperparameter of the pretrained model used to control bidding aggressiveness and to reduce a likelihood of overbidding or underbidding, and wherein the pretrained model was trained on historical winning bids data identifying historical clearing prices corresponding to historical winning bids;
computing a final bid based upon the predicted clearing price; and
communicating the final bid to the content placement system, thereby causing the content placement system to notify a content provider of the final bid, wherein the content provider thereby provides a user device with access to dynamic online content, wherein the receiving, determining, selecting predicting, computing, and causing the final bid to be communicated are performed in less than about 50 milliseconds.

10. The computer system of claim 9 , further comprising:
receiving information for a winning bid for the dynamic content placement opportunity; and
updating the pretrained model based on the received information for the winning bid.

11. The computer system of claim 9 , wherein the pretrained model is a quantile regression model.

12. The computer system of claim 9 , wherein the dynamic content placement opportunity is triggered in response to a request from a user for publisher content, and wherein the information associated with the bid request identifies the user, the publisher content or a type of dynamic content.

13. The computer system of claim 9 , wherein computing the final bid comprises:
selecting a lower price of an initial bid and the predicted clearing price as the final bid.

14. The computer system of claim 9 , wherein computing the final bid comprises:
computing a new price based upon an initial bid and the predicted clearing price; and
indicating the new price as the final bid.

15. The computer system of claim 14 , wherein the new price is different from the initial bid and the predicted clearing price.

16. A non-transitory computer-readable medium having instructions stored thereon, the instructions executable by a processing device to perform operations comprising:
receiving, from a content placement system, a bid request for a dynamic content placement opportunity;
identifying a pace of a campaign, the pace representing an amount intended to be spent per a time unit compared to an amount actually spent in a subset of the time unit;
using the identified pace of the campaign to determine a quantile value corresponding to a selected percentile of historical clearing price data;
using a pretrained quantile regression model, predicting a clearing price for the bid request using information associated with the bid request, wherein the quantile value is a hyperparameter of the pretrained model used to control bidding aggressiveness and to reduce a likelihood of overbidding or underbidding, and wherein the pretrained model was trained on historical winning bids data identifying historical clearing prices corresponding to historical winning bids;
computing, based upon the predicted clearing price, a final bid; and
communicating the final bid to the content placement system as a response to the bid request, wherein the receiving, determining, selecting predicting, computing, and causing the final bid to be communicated are performed in less than about 50 milliseconds.

Assignees

• Adobe Inc

Inventors

• Sarkhel Somdeb
• Mitra Saayan
• Xie Jiatong
• Kothari Alok

Classifications

• G06Q30/0277 Primary

  Online advertisement · CPC title

• G06N20/00

  Machine learning · CPC title

• G06N5/01

  Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound · CPC title

• G06N5/04

  Inference or reasoning models · CPC title

• G06Q30/0275 Primary

  Auctions · CPC title