# US 2013/0238422 A1 · Automated Multivariate Behavioral Prediction

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2013/0238422 A1
- Document ID: US20130238422A1
- Status: Published application
- Kind: A1
- Application number: 13/416,756
- Filed: 2012-03-09
- Published: 2013-09-12
- Earliest claimed: 2012-03-09
- Assignees: Exponential Interactive, Inc.
- Inventors: Alexander Saldanha
- Web page: https://patentgrep.com/patent/US20130238422A1

## CPC

- G06Q 30/02: Marketing; Price estimation or determination; Fundraising
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR

## Abstract

An advertising system identifies behaviors from user activity and associates the behaviors with a user profile. Advertisers provide the advertising system with information on conversion rates of users associated with user profiles. A behavioral model of user responses is built to identify the relative frequency of behaviors for increasing the response rate of ads. Incoming advertising requests are matched to modeled behaviors to determine an advertiser's interest in bidding on the ad placement.

## Claims

### Claim 1

A computer-implemented method for developing an advertiser-specific behavioral lift model, the method executed by a computer system and comprising:
maintaining in the computer system a plurality of user profiles, each user profile including a persistent identifier of a user and a plurality of user behavior vectors for the user, the user behavior vectors indicating occurrences of each of a set of behaviors by the user, each occurrence having a date;
receiving, by the computer system, conversion information relating to a plurality of advertisers associated with the plurality of user profiles, the conversion information indicating users' responses to an advertisement of the advertiser, the users' responses associated with user profiles; and
calculating by the computer system a lift model for each advertiser based upon the user behavioral vectors wherein the lift model for each advertiser identifies a sequence of advertiser behavior tuples with respect to that advertiser; and
wherein each advertiser behavior tuple of the sequence of advertiser behavior tuples is associated with a lift for an population of user profiles which do not exhibit any behaviors corresponding to advertiser behavior tuples previously indicated in the sequence of advertiser behavior tuples; wherein the associated lift for the advertiser behavior tuples in the sequence generally decreases through the sequence.

### Claim 2 (depends on 1)

The computer-implemented method of claim 1, further comprising:
receiving, from a user device, a received persistent identifier and a referrer link;
identifying, from a behavioral vocabulary, a behavior associated with the referrer link;
updating the plurality of user behavior vectors for the user profile associated with the received persistent identifier to reflect an occurrence of the identified behavior associated with the referrer link.

### Claim 3 (depends on 1)

The computer-implemented method of claim 1, wherein the sequence of advertiser behavior tuples includes an advertiser behavior tuple comprising a null set.

### Claim 4 (depends on 1)

The computer-implemented method of claim 1, wherein the sequence of advertiser behavior tuples includes an advertiser behavior tuple comprising at least two behaviors.

### Claim 5 (depends on 1)

The computer-implemented method of claim 1, wherein the plurality of user behavior vectors includes at least one user behavior vector reflecting the conversion information.

### Claim 6 (depends on 1)

The computer-implemented method of claim 1, wherein each behavior comprises a semantic category and an associated information type.

### Claim 7 (depends on 1)

The computer-implemented method of claim 1, wherein the plurality of user behavior vectors further include a look back period for each user behavior vector reflecting the occurrence of each behavior for individual periods of time.

### Claim 8 (depends on 1)

The computer-implemented method of claim 1, wherein the user profiles of the plurality of user profiles do not include individually identifiable information.

### Claim 9 (depends on 1)

The computer-implemented method of claim 1, wherein the lift model for each advertiser provides a confidence interval such that each behavior tuple in a lift model indicates an estimated minimum lift value.

### Claim 10 (depends on 1)

The computer-implemented method of claim 1, wherein calculating the lift model for an advertiser includes sampling user profiles.

### Claim 11

A computer-implemented method for developing an advertiser-specific behavioral lift model, the method executed by a computer system and comprising:
maintaining in the computer system a plurality of user profiles, each user profile including behavioral information indicating occurrence of individual behaviors among a set of behaviors;
receiving, by the computer system, conversion information for an advertiser, the conversion information indicating conversions for individual user profiles, the conversions reflecting a user profile response to an advertisement of the advertiser; and
calculating, by the computer system, a lift model for the advertiser comprising a sequence of behavior tuples corresponding at least one behavior in the set of behaviors, the calculation comprising iteratively performing the following steps:
determining an population, the population comprising a group of user profiles of the plurality of user profiles which do not correspond to any behavior tuples of the current sequence of behavior tuples;
for a group of behavior tuples, identifying a behavior tuple which describes the user profiles with the highest ratio of conversions; and
adding the identified behavioral tuple to the sequence of behavior tuples.

### Claim 12 (depends on 11)

The computer-implemented method of claim 11, wherein the step of calculating a lift model comprises identifying a second behavior tuple and adding the second behavior tuple to the sequence of behavior tuples.

### Claim 13 (depends on 11)

The computer-implemented method of claim 11, wherein the ratio of conversions for a behavior is calculated with respect to a confidence interval.

### Claim 14 (depends on 11)

The computer-implemented method of claim 11, wherein the user profiles of the plurality of user profiles do not include individually identifiable information.

### Claim 15 (depends on 11)

The computer-implemented method of claim 11, wherein the group of behavior tuples comprises a behavior tuple including more than one behavior.

### Claim 16 (depends on 11)

The computer-implemented method of claim 11, wherein the group of behavior tuples comprises a behavior tuple indicating an empty set of behaviors, the empty set indicating a user profile which does not match any other behavior tuples of the group of behavior tuples.
