Methods, systems, and devices for counterfactual-based incrementality measurement in digital ad-bidding platform
A digital ad-buying platform uses counterfactual-based incrementality measurement by implementing randomization and/or a correction for auction win bias to avoid the need to identify counterfactual winner types in the control group. This approach can estimate impact at the individual consumer level. Confidence levels can be determined using Gibbs sampling in the context of causal analysis in the presence of non-compliance.
1. A computer-implemented method for determining one or more bids for an ad impression opportunity based on a causal ad impact of one or more previous bids, the method comprising:
receiving, by a computer system, a first bid request for a first ad impression opportunity;
identifying, by the computer system, one or more advertisers to submit a bid in response to the first bid request to place one or more advertisements on user computing systems of a plurality of users;
executing, by the computer system, a randomization scheme, whereby each of the plurality of users is categorized in one of a control group or a test group;
logging, by the computer system, a control impression for the control group, wherein the one or more advertisements is not transmitted to user computing systems of the control group;
submitting, by the computer system, one or more bids for placing the one or more advertisements on user computing systems of the test group;
receiving, by the computer system, results of the one or more submitted bids, wherein the results indicate whether each of the one or more submitted bids was won or lost;
logging, by the computer system, a test-win impression for a first sub-group of the test group, wherein at least one of the one or more advertisements is transmitted to user computing systems of the first sub-group;
logging, by the computer system, a test-lost impression for a second sub-group, wherein the one or more advertisements is not transmitted to user computing systems of the second sub-group;
accessing, by the computer system, data associated with the control group, the first sub-group, and the second sub-group;
identifying, by the computer system, actions associated with the control group, the first sub-group, and the second sub-group by utilizing the accessed data;
determining, by the computer system, the causal ad impact based at least in part on the identified actions associated with the control group, the first sub-group, and the second sub-group;
analyzing, by the computer system using a machine learning model, the determined causal ad impact by determining the relative efficacy of the one or more submitted bids and determining similarities in characteristics between the one or more submitted bids having similar causal ad impact;
providing, by the computer system, based on the analysis of the determined causal ad impact, one or more recommendations to the one or more advertisers related to submission of one or more bids in response to a second bid request for a second ad impression opportunity; and
adjusting, based on the provided one or more recommendations, the one or more bids in response to the second bid request for the second ad impression opportunity,
wherein the computer system comprises a computer processor and an electronic storage medium.
2. The computer-implemented method of claim 1 , wherein each of the plurality of users is categorized as the control group or the test group based on a fraction probability.
3. The computer-implemented method of claim 1 , wherein the method further comprises:
identifying, by the computer system, one or more consumer responses based on the actions associated with the control group, the first sub-group, and the second sub-group,
wherein determining the causal ad impact is further based at least in part on the one or more consumer responses.
4. The computer-implemented method of claim 3 , wherein the one or more consumer responses comprises one or more of: a site visit, a registration, a subscription, an addition of items to a shopping cart, or a purchase.
5. The computer-implemented method of claim 1 , wherein the computer system comprises a demand side platform.
6. The computer-implemented method of claim 1 , wherein the method further comprises:
determining, by the computer system, a confidence factor for the causal ad impact, wherein the confidence factor comprises an interval and/or a value range associated with a probability percentile.
7. The computer-implemented method of claim 6 , wherein the confidence factor is determined based at least in part on Markov-Chain Monte-Carlo sampling.
8. The computer-implemented method of claim 7 , wherein the Markov-Chain Monte-Carlo sampling comprises a Gibbs sampling scheme.
9. A system for determining one or more bids for an ad impression based on a causal ad impact of one or more previous bids, wherein the system comprises:
one or more computer readable storage devices configured to store a plurality of computer executable instructions; and
one or more hardware computer processors in communication with the one or more computer readable storage devices and configured to execute the plurality of computer executable instructions in order to cause the system to:
receive a first bid request for a first ad impression opportunity;
identify one or more advertisers to submit a bid in response to the first bid request to place one or more advertisements on user computing systems of a plurality of users;
execute a randomization scheme, whereby each of the plurality of users is categorized in one of a control group or a test group;
log a control impression for the control group, wherein the one or more advertisements is not transmitted to user computing systems of the control group;
submit one or more bids for placing the one or more advertisements on user computing systems of the test group;
receive results of the one or more submitted bids, wherein the results indicate whether each of the one or more submitted bids was won or lost;
log a test-win impression for a first sub-group of the test group, wherein at least one of the one or more advertisements is transmitted to user computing systems of the first sub-group;
log a test-lost impression for a second sub-group, wherein the one or more advertisements is not transmitted to user computing systems of the second sub-group;
access data associated with the control group, the first sub-group, and the second sub-group;
identify actions associated with the control group the first sub-group, and the second sub-group by utilizing the accessed data;
determine the causal ad impact based at least in part on the identified actions associated with the control group, the first sub-group, and the second sub-group;
analyze, using a machine learning model, the determined causal ad impact by determining the relative efficacy of the one or more submitted bids and determining similarities in characteristics between the one or more submitted bids having similar causal ad impact;
provide, based on the analysis of the determined causal ad impact, one or more recommendations to the one or more advertisers related to submission of one or more bids in response to a second bid request for a second ad impression opportunity; and
adjust, based the provided one or more recommendations, the one or more bids in response to the second bid request for the second ad impression opportunity.
10. The system of claim 9 , wherein a general advertisement not associated with the one or more advertisements is transmitted to user computing systems of the control group.
11. The system of claim 9 , wherein to determine the causal ad impact the logs of the control impression and the test-win impression are taken over a predetermined time period.
12. The system of claim 9 , wherein the randomization scheme is executed by categorizing each of the plurality of users in the control group or the test group using a hash function.
13. The system of claim 9 , wherein the one or more advertisers are identified based on at least one or more of: targeting requirements, types of consumers, targeting contexts, governing campaign criteria, available budget, desired frequency of exposure, or publisher restrictions on types of advertisers.
14. A computer-implemented method for determining one or more bids for an ad impression opportunity based on a causal ad impact of one or more previous bids, the method comprising:
receiving, by a computer system, a first bid request for a first ad impression opportunity;
identifying, by the computer system, one or more advertisers to submit a bid in response to the first bid request to place one or more advertisements on user computing systems of a plurality of users;
submitting, by the computer system, one or more bids for placing the one or more advertisements on user computing systems of the plurality of users;
receiving, by the computer system, results of the one or more submitted bids, wherein the results indicate whether each of the one or more bids was won or lost;
executing, by the computer system, a randomization scheme, whereby each of the plurality of users is categorized in one of a control group or a test group;
logging, by the computer system, a test group impression for the test group, wherein at least one of the one or more advertisements is transmitted to user computing systems of the test group;
logging, by the computer system, a control group impression for the control group, wherein the one or more advertisements is not transmitted to user computing systems of the control group;
accessing, by the computer system, data associated with the test group and the control group;
identifying, by the computer system, actions associated with the test group and the control group by utilizing the accessed data;
determining, by the computer system, the causal ad impact based at least in part on the identified actions associated with the test group and the control group;
analyzing, by the computer system using a machine learning model, the determined causal ad impact by determining the relative efficacy of the one or more submitted bids and determining similarities in characteristics between the one or more submitted bids having similar causal ad impact;
providing, by the computer system, based on the analysis of the determined causal ad impact, one or more recommendations to the one or more advertisers related to submission of one or more bids in response to a second bid request for a second ad impression opportunity; and
adjusting, based on the determined causal ad impact and the provided one or more recommendations, the one or more bids in response to the second bid request for the second ad impression opportunity;
wherein the computer system comprises a computer processor and an electronic storage medium.
15. The computer-implemented method of claim 14 , wherein one or more integrated identifiers are used to determine the causal ad impact, wherein each of the one or more integrated identifiers is associated with a particular user.
16. The computer-implemented method of claim 15 , wherein over a million users are assessed to create the one or more integrated identifiers.
17. The computer-implemented method of claim 15 , further comprising:
determining, by the computer system, a confidence factor for the causal ad impact, wherein the confidence factor is based on a number of total users assessed to create the integrated identifiers.
18. The computer-implemented method of claim 14 , wherein the computer system comprises a demand side platform.
19. The computer-implemented method of claim 14 , wherein the randomization scheme is executed using a hash function.
20. The computer-implemented method of claim 14 , wherein the randomization scheme is executed subsequent to submitting the one or more bids.