System and process to create a lookalike model for a target audience to deliver advertisements
AssigneeCivis Analytics, Inc.
InventorsDavid Shor, Stephen Hoover, Caitlin Malone, Michael Sadowsky, Zachary Krislov, Evan Sadler
Life of this patent
Estimating the term…
Abstract
A system and process to create a lookalike model for a target audience to deliver advertisements are disclosed. According to one embodiment, the method comprises selecting survey data from a survey database that relates to an advertisement. A heterogenous treatment effect (HETE) model is trained on the survey data. Persuadable customers are identified from the survey database for the advertisement based on the HETE model. An optimized customer list is generated using personally identifiable information.
Claims18
A method, comprising:
- selecting survey data from a survey database that relates to an advertisement;
- training a supervised machine learning model to predict heterogeneous treatment effects on the survey data, wherein the model estimates a probability that an individual would be influenced by the advertisement, as determined by unique characteristics of that individual, and rank orders the individuals by treatment effect;
- identifying persuadable customers from the survey database for the advertisement based on the rank ordering;
- generating an optimized customer list using personally identifiable information for the persuadable customers in the survey database;
- and
- directing the advertisement to a networking website, wherein a subset of users of the networking website will receive the advertisement, and wherein a networking server selects the subset of users, the subset of users sharing characteristics with customers on the optimized customer list.
A non-transitory computer readable medium containing computer-readable instructions stored therein for causing a computer processor to perform operations comprising:
- selecting survey data from a survey database that relates to an advertisement;
- training a supervised machine learning model to predict heterogeneous treatment effects on the survey data, wherein the model estimates a probability that an individual would be influenced by the advertisement, as determined by unique characteristics of that individual, and rank orders the individuals by treatment effect;
- identifying persuadable customers from the survey database for the advertisement based on the rank ordering;
- generating an optimized customer list using personally identifiable information for the persuadable customers in the survey database;
- and
- directing the advertisement to a networking website, wherein a subset of users of the networking website will receive the advertisement, and wherein a networking server selects the subset of users, the subset of users sharing characteristics with customers on the optimized customer list.
Dependencies are inferred from the claim text.