Evaluating attribution models

Patent metadata

Field	Value
Publication number	US-9875484-B1
Application number	US-201414186720-A
Country	US
Kind code	B1
Filing date	Feb 21, 2014
Priority date	Feb 21, 2014
Publication date	Jan 23, 2018
Grant date	Jan 23, 2018

Abstract

Methods, systems, and apparatus including computer programs encoded on computer-readable storage media are provided for evaluating attribution models and comparing estimates produced by the attribution models with causal measurements from controlled experiments. An attribution model is identified for use in determining an estimate of an effectiveness of a campaign. An experiment is identified including experiment data that reflects implementation of a change in an experiment environment including identifying a measure of effectiveness of the change within a predetermined confidence level. The estimate is evaluated as compared to the identified measure of effectiveness. A determination is made that the attribution model is an effective measure of the change for the campaign, based at least in part on the evaluating.

First claim

What is claimed is:

1. A method comprising:
conducting a campaign experiment that reflects implementation of a campaign change in an experiment environment to generate experimental results that reflect a causal measurement of value of a campaign, wherein the campaign change includes a current change to one or more campaign parameters for providing content items over a computer network for presentation by user computing devices;
determining, by one or more computing servers, a measure of effectiveness of the campaign change to within a predetermined confidence level, based at least in part on analyzing the experimental results;
determining, by the one or more computing servers, an estimate of effectiveness of the campaign change, wherein determining the estimate of effectiveness comprises referencing campaign data related to the campaign change, referencing interaction data that indicates user interactions with one or more of the content items, correlating the campaign data with the interaction data over a predetermined time period, and applying rules of an identified attribution model that assigns credit to user interactions with one or more of the content items that lead to user conversions;
comparing, by the one or more computing servers, the determined estimate of effectiveness of the campaign change to the determined measure of effectiveness of the campaign change;
determining, based on the comparison, that the identified attribution model provides an estimate of the measure of effectiveness that is within a predetermined range of the measure of effectiveness of the campaign change;
for one or more subsequent campaign changes that include a change to the one or more campaign parameters that matches the current change to the one or more campaign parameters, determining a subsequent estimate of effectiveness of the subsequent campaign change, based at least in part on applying the identified attribution model in lieu of conducting another campaign experiment to determine the subsequent estimate of effectiveness of the subsequent campaign change; and
using subsequent determined estimates of effectiveness of the subsequent campaign changes obtained from application of the identified attribution model as a proxy for measures of effectiveness in lieu of conducting another campaign experiment to determine the measure of effectiveness.

2. The method of claim 1 , wherein the campaign change is a change of spend associated with the campaign.

3. The method of claim 1 , wherein the campaign change is a change in media type or mix of media types in the campaign.

4. The method of claim 3 , wherein the campaign is associated with an advertisement and the media types are one or more of display, search, or video.

5. The method of claim 1 , further comprising providing a user interface to a campaign sponsor, the user interface including a control that enables selection of a particular attribution model from among a plurality of attribution models for the campaign including providing the particular attribution model as a suggestion to the content sponsor associated with the campaign, wherein the user interface includes a visual indication of a confidence band corresponding to a range of values around a point measure of effectiveness of the campaign change, and for each of the plurality of attribution models, a visual representation of the estimate of effectiveness of the campaign change relative to the confidence band.

6. The method of claim 1 , further comprising providing a user interface to a campaign sponsor, wherein the user interface visually presents a suggestion for a model change from a previously selected attribution model used for the campaign to a suggested attribution model including providing the suggested attribution model as the suggestion when an estimate of effectiveness of a campaign change based on an application of the previously selected attribution model is an unacceptable estimate of a measure of effectiveness of the campaign change.

7. The method of claim 1 , further comprising identifying a plurality of attribution models, wherein comparing the determined estimate of effectiveness of the campaign change to the determined measure of effectiveness of the campaign change includes comparing a determined estimate of effectiveness of the campaign change for each of the identified plurality of attribution models to the determined measure of effectiveness of the campaign change, and wherein determining that the identified attribution model provides an estimate of the measure of effectiveness includes determining the identified attribution model from among the plurality of attribution models as an attribution model that indicates an estimate of the measure of effectiveness that is closest to the determined measure of effectiveness of the campaign change.

8. The method of claim 1 , further comprising using the identified attribution model to determine a bid adjustment for the campaign.

9. The method of claim 1 , further comprising suggesting one or more different bids to be associated with the campaign based on the campaign change and the estimate of effectiveness of the campaign change indicated by the identified attribution model.

10. The method of claim 1 , further comprising using the identified attribution model to generate reports on an estimated effectiveness of the campaign.

11. The method of claim 1 , wherein the campaign experiment is a geographic campaign experiment that uses a control group in one geographic region that does not present content from the campaign as compared to other geographic areas where content from the campaign is presented.

12. The method of claim 1 , further comprising evaluating an incremental performance impact of the campaign change using the identified attribution model.

13. The method of claim 12 , further comprising presenting an evaluation of the identified attribution model to a content sponsor associated with the campaign.

14. The method of claim 1 , wherein the identified attribution model is selected from the group comprising first click, last click, and data-driven.

15. The method of claim 1 , wherein the predetermined confidence level is associated with a band, and the estimate of effectiveness of the campaign change determined based on the attribution model is within the band.

16. A computer program product embodied in a non-transitive computer-readable medium including instructions, that when executed, cause one or more processors to perform operations comprising:
conducting a campaign experiment that reflects implementation of a campaign change in an experiment environment to generate experimental results that reflect a causal measurement of value of a campaign, wherein the campaign change includes a current change to one or more campaign parameters for providing content items over a computer network for presentation by user computing devices;
determining, by one or more computing servers, a measure of effectiveness of the campaign change to within a predetermined confidence level, based at least in part on analyzing the experimental results;
determining, by the one or more computing servers, an estimate of effectiveness of the campaign change, wherein determining the estimate of effectiveness comprises referencing campaign data related to the campaign change, referencing interaction data that indicates user interactions with one or more of the content items, correlating the campaign data with the interaction data over a predetermined time period, and applying rules of an identified attribution model that assigns credit to user interactions with one or more

Assignees

• Google Inc

Inventors

• Schnabl Stefan F
• Vaver Jon

Classifications

• G06Q30/0242 Primary

  Determining effectiveness of advertisements · CPC title