Value function-based estimation of multi-channel attributions
View Patent ↗Techniques for analyzing marketing channels are described. Users are exposed to the marketing channels. User responses (e.g., purchases and no-purchases) to the exposures are tracked. Upon a request from a marketer to analyze an attribution of a marketing channel, the user responses are analyzed. The attribution represents the credit that the marketing channel should get for influencing the users exposed thereto into exhibiting a particular user response (e.g., a purchase). The analysis involves multiple steps. In a first step, a non-parametric estimation is used to generate a value function at a user-level. In a second step, a coalitional game approach is used to estimate the attribution based on the value function. A response is provided to the marketer with data about the attribution.
1. A method for intelligently distributing interactive content for accessing online platforms via electronic communication channels, the method comprising:
receiving, by a computer system and from a marketer device, a request for an attribution, to a target user response, of a particular electronic communication channel from a set of electronic communication channels, wherein the target user response comprises a navigation of a user device to an online platform via interactive content distributed among a combination of the electronic communication channels in the set;
accessing, by the computer system, exposure feature vectors comprising binary vector data structures having dimensions representing response data, the response data generated from responses of user devices that correspond to exposures of the user devices to prior interactive content distributed via the set of electronic communication channels;
estimating, by the computer system and based on the response data, a likelihood of the target user response given the combination of the electronic communication channels, the combination comprising the particular electronic communication channel;
generating, by the computer system, an output attribution of the particular electronic communication channel to the target user response based on the likelihood of the target user response, wherein generating the output attribution comprises:
segmenting the response data into (i) a first group corresponding to a first set of exposure feature vectors representing a first combination of exposures of the user devices to the prior interactive content distributed via the set of electronic communication channels and (ii) a second group corresponding to a second set of exposure feature vectors representing a second combination of exposures of the user devices to the prior interactive content distributed via the set of electronic communication channels, wherein the first and second combinations of exposures are different from each other,
generating a first coalition game model from the first group and a second coalition game model from the second group,
computing a first attribution to the particular electronic communication channel by applying the first coalition game model to the set of electronic communication channels,
computing a second attribution to the particular electronic communication channel by applying the second coalition game model to the set of electronic communication channels, and
aggregating the first attribution and the second attribution into the output attribution; and
causing the marketer device to modify a distribution, among the set of electronic communication channels, of interactive content for navigating to the online platform, wherein causing the marketer device to modify the distribution of interactive content comprises transmitting, from the computer system to the marketer device, attribution data including the output attribution.
2. The method of claim 1 , further comprising:
analyzing a plurality of attributions corresponding to the set of electronic communication channels;
generating a recommendation for modifying usage of the particular electronic communication channel based on the plurality of attributions;
providing the recommendation to the marketer device;
receiving a resource reallocation for the particular electronic communication channel from the marketer device based on the recommendation; and
providing, based on the resource reallocation, an interactive content item to a computing device of a user via the particular electronic communication channel.
3. The method of claim 1 , wherein causing the marketer device to modify the distribution of interactive content comprises causing the marketer device to perform a resource allocation to the particular electronic communication channel.
4. The method of claim 1 , further comprising analyzing a difference between the output attribution and an additional attribution of an additional electronic communication channel to improve usage of the particular electronic communication channel or the additional electronic communication channel.
5. The method of claim 1 , wherein estimating the likelihood of the target user response comprises:
generating each potential combination of electronic communication channels of the set of electronic communication channels;
estimating a value function based on each potential combination of electronic communication channels, wherein each value of the value function corresponds to a potential combination of electronic communication channels and comprises a likelihood of a user exhibiting the target user response given an exposure of the user to the potential combination of electronic communication channel.
6. The method of claim 1 , wherein the likelihood is estimated based on a number of users associated with user devices exhibiting the target user response and having been exposed to the combination of the electronic communication channels relative to a total number of the users having been exposed to the combination of the electronic communication channels.
7. The method of claim 1 , wherein the likelihood is estimated by at least:
for each electronic communication channel of the combination of the electronic communication channels, estimating an individual likelihood based on a subset of the responses corresponding to exposures of users associated with user devices to the electronic communication channel; and
estimating the likelihood associated with the combination of the electronic communication channels based on individual values estimated for the electronic communication channels of the combination of the electronic communication channels.
8. The method of claim 1 , wherein the likelihood is estimated by averaging individual likelihoods corresponding to user exposures to individual electronic communication channels of the combination of the electronic communication channels.
9. The method of claim 1 , wherein estimating the likelihood of the target user response comprises estimating a first value function based on potential combinations of electronic communication channels, wherein each value of the first value function corresponds to a likelihood of the target user response given an exposure to a potential combination of electronic communication channels, and further comprising:
estimating a second value function of the set of electronic communication channels based on the first value function, the second value function associated with no exposures of users associated with user devices to the set of electronic communication channels, and wherein the output attribution of the particular electronic communication channel is further estimated based on the second value function.
10. The method of claim 1 , wherein the target user response further comprises one or more of a purchase, a click through, or a page view, and wherein the electronic communication channels comprise one or more of: a web page, an advertisement with interactive content displayed in the web page, an email with a clickable link, or a social network post.
11. A system comprising:
a processor;
a network interface device communicatively coupled to a marketer device;
a memory communicatively coupled to the processor and bearing computer-readable instructions that, upon execution by the processor, cause the system to perform operations:
receiving, from the marketer device, a request for an attribution, to a target user response, of a particular electronic communication channel from a set of electronic communication channels, wherein the target user response comprises a navigation of a user device to an online platform via interactive content distributed among a combination of the electronic communication channels;
accessing exposure feature vectors comprising binary vector data structures having dimensions representing response data, the response data generated from responses of user devices that correspond to exposures of the user devices to prior interactive content distributed via the set of electronic communication channels;
estimating, based on the response data, a likelihood of the target user response given the combination of the electronic communication channels, the combination comprising the particular electronic communication channel;
generating an output attribution of the particular electronic communication channel to the target user response based on the likelihood of the target user response, wherein generating the output attribution comprises:
segmenting the response data into (i) a first group corresponding to a first set of exposure feature vectors representing a first combination of exposures of the user devices to the prior interactive content distributed via the set of electronic communication channels and (ii) a second group corresponding to a second set of exposure feature vectors representing a second combination of exposures of the user devices to the prior interactive content distributed via the set of electronic communication channels, wherein the first and second combinations of exposures are different from each other,
generating a first coalition game model from the first group and a second coalition game model from the second group,
computing a first attribution to the particular electronic communication channel by applying the first coalition game model to the set of electronic communication channels,
computing a second attribution to the particular electronic communication channel by applying the second coalition game model to the set of electronic communication channels, and
aggregating the first attribution and the second attribution into the output attribution; and
causing the marketer device to modify a distribution, among the set of electronic communication channels, of interactive content for navigating to the online platform, wherein causing the marketer device to modify the distribution of interactive content comprises transmitting, from the processor to the marketer device, attribution data including the output attribution.
12. The system of claim 11 , wherein the target user response further comprises one or more of a purchase, a click through, or a page view, and wherein the electronic communication channels comprise one or more of: a web page, an advertisement with interactive content displayed in the web page, an email with a clickable link, or a social network post.
13. The system of claim 11 , wherein the target user response further comprises a conversion resulting from the navigation to the online platform, wherein the operations further comprise:
generating the combination of the electronic communication channels from the set of electronic communication channels;
accessing a set of the responses corresponding to exposures of users to the combination of the electronic communication channels; and
estimating a likelihood of conversion from the exposures of the users to the combination of the electronic communication channels based on the set of the responses.
14. The system of claim 11 , wherein the operations further comprise:
for each user having the target user response:
determining a particular combination of electronic communication channels that resulted in the target user response of the user and that comprise an electronic communication channel; and
estimating a surplus of the particular electronic communication channel based on a corresponding value of the particular combination of electronic communication channels; and
estimating the output attribution of the particular electronic communication channel based on surpluses estimated for the particular electronic communication channel for a subset of the users having the target user response and exposed to the particular electronic communication channel.
15. The system of claim 11 , wherein the operations further comprise estimating a likelihood of conversion based on exposures of users to the combination of the electronic communication channels, wherein estimating the likelihood of conversion comprises:
determining whether a set of the responses is sufficient to estimate the likelihood of conversion;
if the set of the responses is sufficient, utilizing the set of the responses to estimate the likelihood of conversion; and
if the set of the responses is insufficient,
accessing subsets of the responses corresponding to exposures of the users to respective electronic communication channels of the combination of the electronic communication channels;
estimating individual likelihoods of conversion corresponding to the respective electronic communication channels based on the subset of the responses; and
utilizing the individual likelihoods of conversion to estimate the likelihood of conversion.
16. The system of claim 11 , wherein the operations further comprise collecting the responses by storing the responses in a data store accessible to the system, wherein each response is stored in association with a corresponding combination of electronic communication channels to which a user associated with the user device was exposed.
17. A non-transitory computer-readable storage medium storing instructions that, when executed on a computing system, configure the computing system to perform operations comprising at least:
receiving, from a marketer device, a request for an attribution, to a target user response, of a particular electronic communication channel from a set of electronic communication channels, wherein the target user response comprises a navigation of a user device to an online platform via interactive content distributed among a combination of the electronic communication channels;
accessing exposure feature vectors comprising binary vector data structures having dimensions representing response data, the response data generated from responses of user devices that correspond to exposures of the user devices to prior interactive content distributed via the set of electronic communication channels;
estimating, based on the response data, a likelihood of the target user response given the combination of the electronic communication channels, the combination comprising the particular electronic communication channel;
generating, by the computer system, an output attribution of the particular electronic communication channel to the target user response based on the likelihood of the target user response, wherein generating the output attribution comprises:
segmenting the response data into (i) a first group corresponding to a first set of exposure feature vectors representing a first combination of exposures of the user devices to the prior interactive content distributed via the set of electronic communication channels and (ii) a second group corresponding to a second set of exposure feature vectors representing a second combination of exposures of the user devices to the prior interactive content distributed via the set of electronic communication channels, wherein the first and second combinations of exposures are different from each other,
generating a first coalition game model from the first group and a second coalition game model from the second group,
computing a first attribution to the particular electronic communication channel by applying the first coalition game model to the set of electronic communication channels,
computing a second attribution to the particular electronic communication channel by applying the second coalition game model to the set of electronic communication channels, and
aggregating the first attribution and the second attribution into the output attribution; and
causing the marketer device to modify a distribution, among the set of electronic communication channels, of interactive content for navigating to the online platform, wherein causing the marketer device to modify the distribution of interactive content comprises transmitting, to the marketer device, attribution data including the output attribution.
18. The computer-readable storage medium of claim 17 , wherein the output attribution of the particular electronic communication channel is further estimated based on a set of user devices having the target user response.
19. The computer-readable storage medium of claim 17 , wherein the instructions, when executed on the computing system, further configure the computing system to perform operations comprising at least:
for each user having the target user response:
determining a particular combination of electronic communication channels that resulted in the target user response of the user and that comprise an electronic communication channel; and
estimating a surplus of the particular electronic communication channel based on a corresponding value of the particular combination of electronic communication channels; and
estimating the output attribution of the particular electronic communication channel based on surpluses estimated for the particular electronic communication channel for a subset of the users having the target user response and exposed to the particular electronic communication channel.
20. The computer-readable storage medium of claim 17 , wherein the electronic communication channels represent players of one or more of the first coalition game model and the second coalition game model, wherein the one or more of the first coalition game model and the second coalition game model includes a characteristic function for attribution computation.