Methods and systems for measuring conversion probabilities of paths for an attribution model

#An online user today is exposed to a plethora of media exposures, such as banner ads, email ads, display ads, organic and paid search results, amongst others. These media exposures can be configured to direct a user to a particular website. When the online user performs a converting act, such as making an online purchase, advertisers would like to know which of the various media exposures the user was exposed to were responsible for the user's converting act. Historically, the media exposure the user was last exposed to would get all of the credit for the conversion, while all other media exposures that the user was exposed to would get none. This attribution model is referred to as last click attribution.

#Methods, apparatuses, and systems for creating an attribution model that assigns attribution credit to not only the last media exposure the user was exposed to prior to a converting act, but to other media exposures that were partly responsible for the occurrence of the converting act are described herein. In particular, the attribution model described herein relies on visit related data of visits to a website, including but not limited to conversion probabilities of paths taken by the visitors visiting the website. As such, the present disclosure relates to a data-driven attribution model for assigning attribution credit to various media exposures associated with paths and a conversion probability determination engine that is configured to determine the likelihood of a path converting.

#According to one aspect, a method for creating a data-driven attribution model, includes, identifying by a processor, for a given time period, a plurality of visits to a particular website. The processor then identifies, for each visitor identifier associated with the identified plurality of visits, a path associated with the visitor identifier. The path including at least one event that has a corresponding index position indicating a position of the event relative to positions of other events included in the path. The processor then determines, for each path type associated with the identified paths, a path-type conversion probability based on a number of visits corresponding to the path type that resulted in a conversion. The processor then calculates, for each of a plurality of the path types associated with the identified paths, a counterfactual gain for each event based on a conversion probability of the given path type and a conversion probability of a path type that does not include the event for which the counterfactual gain is calculated. The processor determines, for each event of each of the plurality of path types, an attribution credit based on the calculated counterfactual gain of the event. The processor then stores, for each of a plurality of the path types associated with the identified paths, the determined attribution credit for each event included in the path type.

#The accompanying drawings are not intended to be drawn to scale. Like reference numbers and designations in the various drawings indicate like elements. For purposes of clarity, not every component may be labeled in every drawing. In the drawings:

#FIG. 1 is a block diagram depicting one implementation of an environment for identifying competitors using content items including content extensions, according to an illustrative implementation;

#FIG. 2A-2C show conceptual illustrations of a plurality of identified paths;

#Following below are more detailed descriptions of various concepts related to, and implementations of, methods, apparatuses, and systems for creating an attribution model that assigns attribution credit to not only the last media exposure the user was exposed to prior to a converting act, but to other media exposures that were partly responsible for the occurrence of the converting act. In particular, the attribution model described herein relies on visit related data of visits to a website, including but not limited to conversion probabilities of paths taken by the visitors visiting the website. As such, the present disclosure relates to a data-driven attribution model for assigning attribution credit to various media exposures associated with paths and a conversion probability determination engine that is configured to determine the likelihood of a path converting. The various concepts introduced above and discussed in greater detail below may be implemented in any of numerous ways, as the described concepts are not limited to any particular manner of implementation. Examples of specific implementations and applications are provided primarily for illustrative purposes.

#As described above, an online user today is exposed to a plethora of media exposures or marketing touchpoints, such as banner ads, email ads, display ads, organic and paid search results, social media posts or notifications, amongst others. A user exposed to such media exposures may likely take an action related to the media exposure, for example, click on the media exposure. Generally, upon taking an action on the media exposure to which the user is exposed, the user can be directed to the website linked to the media exposure, resulting in a user visit. In the event that the user performs a converting act at the website, for example, making an online purchase or registering an account, advertisers would like to know whether any of the media exposures to which the user was exposed deserve to get attribution credit for the user's converting act, and if so, the amount of attribution credit.

#Historically, in a last click attribution model, the media exposure to which the user was last exposed would get all of the credit for the conversion, while all other media exposures that the user was exposed to would get none. The last click attribution model, however, is not equitable as it fails to give credit to media exposures that deserve attribution credit. It is understood that last click attribution is less than ideal and hence various alternative attribution models have been developed. These attribution models include simple rules based approaches such as dividing the credit equally among all the media exposures to which the user was exposed prior to the converting act. However, none of these existing attribution models rely on historical data associated with visits to a particular website to determine the amount of attribution credit deserving media exposures should receive.