Evaluating attribution models based on simulated activity streams

Patent metadata

Field	Value
Publication number	US-9922340-B1
Application number	US-201414250016-A
Country	US
Kind code	B1
Filing date	Apr 10, 2014
Priority date	Apr 10, 2014
Publication date	Mar 20, 2018
Grant date	Mar 20, 2018

Abstract

Methods, systems, and apparatus including computer programs encoded on computer-readable storage media are provided for evaluating an attribution model, based on simulated activity streams. Parameters are specified that describe how users behave in the absence and presence of advertising, and parameters that regulate advertising in a simulation. A first set of simulated activity streams is generated with advertising turned on, observational metrics associated with the first set of streams are determined, and an attribution model is applied to determine a first fraction of incremental conversions associated with one or more advertising channels. Further sets of simulated activity streams are generated, each with a single advertising channel turned off, observational metrics associated with the further sets of streams are determined, and a second fraction of incremental conversions is determined for each advertising channel. The first and second fractions of incremental conversions are compared to evaluate the attribution model.

First claim

What is claimed is:

1. A method comprising:
specifying parameters that describe how users behave in an absence of advertising, parameters that describe how users react to advertising, and parameters that regulate injection of advertising in a simulation, wherein specifying parameters that describe how users behave in an absence of advertising includes at least one of specifying an activity level that influences a length of a simulated activity stream, an activity preference from among a plurality of possible activities, and a conversion rate for a given activity, wherein specifying parameters that describe how users react to advertising includes specifying a probability that a user's behavior will change with advertising exposure, wherein the simulation is conducted using the specified parameters to generate a plurality of simulated activity streams, each simulated activity stream representing a sequence of simulated online actions performed by a virtual user during a simulated web session;
generating, by one or more processors, a first set of simulated activity streams with advertising turned on, based on the specified parameters;
determining observational metrics associated with the first set of simulated activity streams;
applying an attribution model to the generated first set of simulated activity streams and determining a first fraction of incremental conversions associated with one or more advertising channels;
generating, by one or more processors, one or more further sets of simulated activity streams, each with a single advertising channel turned off, based on the specified parameters;
determining observational metrics associated with the one or more further sets of simulated activity streams;
determining a second fraction of incremental conversions for each advertising channel by evaluating the further set of simulated activity streams with the single advertising channel turned off; and
comparing the first and second fractions of incremental conversions to determine an applicability of the attribution model for estimating effectiveness of a campaign, wherein the comparing comprises generating a comparison report that includes information related to the first and second fractions and providing the comparison report for presentation at a client device.

2. The method of claim 1 , wherein the simulation is conducted to generate a different simulated activity stream for each virtual user of a heterogeneous set of virtual users, wherein the activity level that influences the length of the simulated activity stream represents a volume of user activity for a given virtual user over a predetermined time interval, and wherein the activity preference specifies a frequency of a particular activity for the given virtual user.

3. The method of claim 2 , wherein specifying parameters that describe how users behave in an absence of advertising includes generating a transition matrix that includes columns and rows that represent activities and corresponding entries in the transition matrix that define a probability of progressing from a given activity to a corresponding next activity.

4. The method of claim 3 , wherein specifying parameters that describe how users react to advertising includes a modification to the transition matrix based on exposure to a given advertisement.

5. The method of claim 4 , wherein specifying parameters that regulate injection of advertising in a simulation includes specifying one or more of a probability that an advertisement is served in a respective channel of the campaign, a budget cap per channel, a minimum threshold for impressibility for each channel, or a cost of advertising per channel.

6. The method of claim 5 , wherein generating the first set of simulated activity streams includes for a first stream, identifying a starting point on the transition matrix that corresponds with an initial activity, traversing the transition matrix in accordance with defined probabilities, impressing advertising from the channels in accordance with a probability that an advertisement will be served in association with a respective activity, adjusting the transition matrix in accordance with the modification, and continuing to generate the first stream until an end of session activity is determined.

7. The method of claim 6 , wherein determining observational metrics associated with the first set of simulated activity streams includes, for each type of advertising channel, determining one or more of impressions, clicks, spend, reach, frequency, or share.

8. The method of claim 1 , wherein applying the attribution model includes determining an estimate of a number of conversions associated with each advertising channel.

9. The method of claim 1 , wherein determining the first fraction of incremental conversions associated with one or more advertising channels includes calculating an estimated fraction of incremental conversions for each of the channels.

10. The method of claim 3 , wherein generating the one or more further sets of simulated activity streams includes, for a first stream, identifying a starting point on the transition matrix that corresponds with an initial activity, traversing the transition matrix in accordance with defined probabilities, and continuing to generate the first stream until an end of session activity is determined.

11. The method of claim 10 , wherein determining observational metrics associated with the one or more further sets of simulated activity streams includes, for each type of advertising channel, determining one or more of impressions, clicks, or share.

12. The method of claim 1 , wherein determining the second fraction of incremental conversions for each advertising channel comprises:
determining a number of incremental conversions for the single advertising channel;
repeating the generating and determining for each advertising channel;
calculating a total number of incremental conversions for each advertising channel; and
determining the second fraction of incremental conversions for each advertising channel based on a respective number of incremental conversions for a given channel divided by the total number of incremental conversions.

13. The method of claim 1 , further comprising repeating the method for a second attribution model and comparing results for each attribution model to determine a most suitable attribution model for estimating effectiveness of the campaign.

14. The method of claim 1 , further comprising varying the parameters and repeating the method to evaluate the attribution model in view of the varied parameters.

15. The method of claim 1 , further comprising using total spend as a budget, determining an optimal mix of spend across all advertising channels that generates a most incremental conversions using the budget, generating observational metrics in association with the optimal mix, finding a fraction of incremental spend to assign to each channel, and determining applicability of the attribution model including comparing the fraction of incremental spend with the first fraction of incremental conversions.

16. The method of claim 15 , further comprising repeating the method for a second attribution model and comparing results for each attribution model to determine a most suitable attribution model for estimating effectiveness of the campaign.

17. The method of claim 15 , further comprising varying the parameters and repeating the method to evaluate the attribution model in view of the varied parameters.

18. The method of claim 1 , wherein one or more of the further sets of simula

Assignees

• Google Inc
• Google Llc

Inventors

• Vaver Jon
• Schnabl Stefan F

Classifications

• G06Q30/0242 Primary

  Determining effectiveness of advertisements · CPC title