SYSTEM AND METHOD FOR ATTRIBUTING MULTI-CHANNEL CONVERSION EVENTS AND SUBSEQUENT ACTIVITY TO MULTI-CHANNEL MEDIA SOURCES

• US 20120054019A1
• Filed: 08/30/2011
• Published: 03/01/2012
• Est. Priority Date: 08/30/2010
• Status: Active Grant

• Current Assignee: Adap.tv, Inc. (Apollo Global Management, Inc.)
• Original Assignee: Lucid Commerce, Inc.
• Inventors: Kitts, Brendan, Burdick, Brian, Au, Dyng, Wei, Liang, Powter, Amanda
• Granted Patent: US 8,768,770 B2
• Time in Patent Office: Days
• US Class Current: 705/14.41
• CPC Class Codes: G06Q 30/0241 Advertisements
• CPC Class Codes: G06Q 30/0242 Determining effectiveness of advertisements
• CPC Class Codes: G06Q 30/0243 Comparative campaigns
• CPC Class Codes: G06Q 30/0246 Traffic
• CPC Class Codes: G06Q 30/0254 based on statistics
• CPC Class Codes: G06Q 30/0269 based on user profile or attribute
• CPC Class Codes: G06Q 30/0272 Period of advertisement exposure

Abstract

This paper presents a practical method for measuring the impact of multiple marketing events on sales, including marketing events that are not traditionally trackable. The technique infers which of several competing media events are likely to have caused a given conversion. The method is tested using hold-out sets, and also a live media experiment for determining whether the method can accurately predict television-generated web conversions.

Claims

1. A computer-implemented method of attributing a plurality of conversion events to a plurality of media events, the method comprising:

determining a first demographic and psychographic profile for each of a plurality of media events;

determining a second demographic and psychographic profile for each of a plurality of customers corresponding to the plurality of conversion events;

comparing the first demographic and psychographic profiles with the second demographic and psychographic profiles to determine a most likely media channel and media event that generated each of the conversion events.

2. The method of claim 1, further comprising:

obtaining known conversion event-media event pairs;

using the known conversion event-media event pairs to build a model to predict a probability that a specific conversion event results from one of the plurality of media events;

executing the model for unattributed conversion events in conjunction with comparing the first demographic and psychographic profiles and the second demographic and psychographic profiles to determine a probability of attribution to each media event for each unattributed conversion event;

wherein the model takes into account customers associated with the unattributed conversion events who have been exposed to multiple media events.

3. The method of claim 1, wherein a conversion event is one of the following:

phone-in order, web order, or retail purchase transaction.

4. The method of claim 1, wherein media channels corresponding to the plurality of media events include one or more of the following:

television, radio, internet, billboard, print, and direct mail advertisement, and further wherein a media event is an advertisement using one of the media channels.

5. The method of claim 1, further comprising:

attributing a first percentage of credit for subsequent conversion events performed by a customer following the customer'"'"'s first conversion event to the media event that generated the customer'"'"'s first conversion event;

attributing a second percentage of credit to more recent media events.

6. The method of claim 1, wherein the comparison further assigns a probability of having generated each of the conversion events to each of the plurality of media events.

7. The method of claim 1, wherein the plurality of media events is online, and the plurality of conversion events is offline;

or the plurality of media events is offline, and the plurality of conversion events is online.

8. The method of claim 1, wherein the media events occur through multiple channels, and further wherein at least one of the media events occurs through a channel that does not perform conversion tracking and further wherein at least one of the media events does not uniquely track a customer from media event exposure to conversion event.

9. The method of claim 1, further comprising:

generating a report of the conversion events and attribution to one or more media events.

10. The method of claim 9, further comprising:

removing web conversion events from the report that are attributed to television media events; and

showing the web conversion events attributed to television media events as part of television media performance metrics.

11. The method of claim 1, wherein determining a most likely media channel and media event further uses geo-spatial statistics for the plurality of conversion events and the plurality of media events.

12. The method of claim 1, wherein each conversion event occurs through a conversion event channel, and further where the conversion event channel is one of retail or television, and the media event channel is the other one;

or the conversion event channel is one of web or television, and a media event channel is the other one;

or the conversion event channel is one of retail or web, and the media event channel is the other one.

13. The method of claim 2, wherein the model includes one or more of the following:

logistic regression, decision trees, neural networks, other supervised machine learning models.

14. The method of claim 2, wherein the model takes into account a geographic area receiving the media event, a population estimated to have seen the media event, a time of the media event, and the demographics of the media event.

15. The method of claim 2, wherein a threshold is pre-selected such that if a highest predicted probability of a media event being responsible for a conversion event is lower than the threshold, the conversion event is attributed to one or more unknown events.

16. The method of claim 15, further comprising statistically identifying and characterizing the one or more unknown events.

17. The method of claim 15, further comprising:

generating a first report of the conversion events using the known media events and the unknown events; and

generating a second report of only the conversion events attributed to known media events.

18. The method of claim 15, wherein the threshold is pre-selected by an aggregate statistics method.

19. The method of claim 2, wherein the model is assigned an identification code, and attributions made by the model for a set of conversion events and media events are tracked using the identification code, and further wherein multiple models are run to produce attribution decisions, and each model is assigned a different identification code, and further wherein software code that generates each of the models is referenced using the identification code, analyzed conversion events, and analyzed media events to allow different models to be modularly exchanged in production, and further wherein a specific model is deployed by using the identification code of that specific model.

20. The method of claim 19, further comprising:

monitoring accuracy of the multiple models using the known conversion event-media event pairs;

executing automated checks of the model on incoming data; and

triggering an alert if the accuracy is below a predefined value or if the incoming data is erroneous.

21. A computer-implemented method of attributing consumer responses to advertising events, the method comprising:

inferring a model and parameters, wherein the model and parameters accurately predict data from a training set, wherein the training set comprises a first plurality of consumer responses and a first plurality of advertising events, and further wherein each of the consumer responses is believed to be caused by one or more of the advertising events;

using the model to predict which one or more of a second plurality of advertising events caused each one of a second plurality of consumer responses.

22. The method of claim 21, wherein the training set is obtained by selecting advertising event—

consumer response instances, and further wherein the training set is used to train the model to recognize and correctly attribute future advertising event—

consumer response pairs.

23. The method of claim 21, wherein the training set is obtained by inspecting historical data pertaining to the first plurality of consumer responses and the first plurality of advertising events to determine cases consistent with causation of the first plurality of consumer responses with a high probability.

24. The method of claim 21, further comprising using a probability threshold to determine if the second plurality of consumer responses was caused by one or more unknown events.

25. The method of claim 24, further comprising statistically identifying and characterizing the one or more unknown events.

26. A system for attributing a plurality of conversion events to a plurality of media events, the system comprising:

at least one database configured to store a training set comprised of known conversion event-media event pairs;

a processor, communicatively coupled to the at least one database, and configured to;

train a supervised model using the training set;

use the model to predict which one or more of a plurality of media events caused each one of a plurality of responses.

27. The system of claim 26, wherein the processor is further configured to use a probability threshold to determine if the plurality of responses was caused by one or more unknown events.

28. The system of claim 26, wherein the processor is further configured to statistically identify and characterize the one or more unknown events.

29. The system of claim 26, wherein the processor is further configured to:

determine a first demographic and psychographic profile for each of a plurality of media events;

determine a second demographic and psychographic profile for each of a plurality of customers corresponding to the plurality of conversion events;

compare the first demographic and psychographic profiles with the second demographic and psychographic profiles to determine in conjunction with the supervised model a most likely media channel and media event that generated each of the conversion events.