# US 2012/0259620 A1 · Message optimization

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

- Number: US 2012/0259620 A1
- Document ID: US20120259620A1
- Status: Published application
- Kind: A1
- Application number: 13/517,032
- Filed: 2010-11-12
- Published: 2012-10-11
- Earliest claimed: 2009-12-23
- Assignees: Upstream Mobile Marketing Limited
- Inventors: Alexios Vratskides; Ori Einstein; Guy Stephane Krief; Assaf Baciu
- Web page: https://patentgrep.com/patent/US20120259620A1

## CPC

- G06Q 30/02: Marketing; Price estimation or determination; Fundraising
- G06F 40/00: Handling natural language data (speech analysis or synthesis, speech recognition G10L)
- G06F 40/10: Text processing (natural language analysis G06F40/20; semantic analysis G06F40/30; processing or translation of natural language G06F40/40)
- G06F 40/103: Formatting, i.e. changing of presentation of documents (automatic justification G06F40/189; automatic line break hyphenation G06F40/191)
- G06F 40/237: Lexical tools
- G06F 40/166: Editing, e.g. inserting or deleting
- G06F 40/253: Grammatical analysis; Style critique
- G06F 40/30: Semantic analysis
- 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
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

The present invention provides a system and method for optimizing a message. Components of a starting message are identified, and at least one rule is applied for modifying at least one message component to create at least one variation of the starting message. Message variants are tested by sending each variant to a sample of people and measuring a response rate for each sent message variant. The measured response rates are used to create an optimal version of the message. In one embodiment, message variants may be created and tested in multiple rounds.

## Claims

### Claim 1

A method performed by a computer for automatically optimizing a message, the method comprising:
treating key components of the message as independent variables;
receiving one or more values for each of the key components;
automatically creating variants of the message based on the received values for the key components;
sending each of a plurality of the created message variants to a respective sample of people;
measuring a response rate for each sent message variant;
comparing the measured response rates to identify one or more best performing values for each of the key components;
automatically creating syntactical combinations of the best performing values of the key components;
sending a plurality of the combinations to a representative sample of people;
measuring a response rate for each of the sent combinations; and
comparing the measured response rates for the sent combinations to identify a message having a highest response rate.

### Claim 2 (depends on 1)

The method of claim 1 wherein:
the step of creating variants of the message comprises applying rules that specify how to manipulate the key components to transform the message, wherein the rules includes at least one of the following: (i) a rule for replacing a value of a key component with one or more alternate values, and (ii) a rule for changing a position of key component within the message.

### Claim 3 (depends on 1)

The method of claim 1 wherein only a subset of the created message variants is sent to people and a D-optimal design is used to determine the subset.

### Claim 4 (depends on 1)

The method of claim 1, wherein the message variants are created and tested in a plurality of phases, wherein the performance of message variants in one phase is used to determine the message variants that will be created and tested in a next phase.

### Claim 5 (depends on 1)

The method of claim 1, wherein the messages variants are created and tested in accordance with a supervised learning method to identify the key component value combinations that maximize response rate.

### Claim 6 (depends on 1)

The method of claim 1, wherein regression analysis is applied to the measured response rate for each of the tested message variants to identify the best performing key component values.

### Claim 7 (depends on 4)

The method of claim 4, wherein the values of the key components are modelled by an ontology and categorized into families of values.

### Claim 8 (depends on 7)

The method of claim 7, further comprising: in response to receiving a positive response rate for a message variant with a particular key component value, testing other values for such particular key component that are within the same family as the value of the particular key component associated with the positive response rate; and
in response to receiving a negative response rate for a message variant with a particular key component value, testing other values for such particular key component that are in a different family as the value of the particular key component associated with the negative response rate.

### Claim 9

(canceled)

### Claim 10 (depends on 1)

The method of claim 1, wherein, in creating message variants, the message is reduced to a vector and each variant is represented by a different value of the vector.

### Claim 11 (depends on 1)

The method of claim 1, wherein rules for manipulating key components are substantially in the form of a context-free grammar.

### Claim 12

A computer program embodied on a non-transitory computer-readable medium and comprising code, that, when executed by a computer system, enables the computer system to perform the following method for optimizing a message:
treating key components of the message as independent variables;
receiving one or more values for each of the key components;
automatically creating variants of the message based on the received values for the key components;
sending each of a plurality of the created message variants to a sample of people;
measuring a response rate for each message variant;
comparing the measured response rates to identify one or more best performing values for each of the key components;
automatically creating syntactical combinations of the best performing values of the key components;
sending each of the combinations to a representative sample of people;
measuring a response rate for each of the combinations; and
comparing the measured response rates for the combinations to identify a message having a highest response rate.

### Claim 13 (depends on 12)

The computer program of claim 12 wherein:
the step of creating variants of the message comprises applying rules that specify how to manipulate the key components to transform the message, wherein the rules includes at least one of the following: (i) a rule for replacing a value of a key component with one or more alternate values, and (ii) a rule for changing a position of key component within the message so as to enable the computer to generate message variants that correspond to all possible combinations.

### Claim 14 (depends on 12)

The computer program of claim 12 wherein only a subset of the created variants is sent to people and a D-optimal design is used to determine the subset.

### Claim 15 (depends on 12)

The computer program of claim 12, wherein the message variants are created and tested in a plurality of phases, wherein the performance of message variants in one phase is used to determine the message variants that will be created and tested in a next phase.

### Claim 16 (depends on 15)

The computer program of claim 15, wherein the values of the key components are modelled by an ontology and categorized into families of values.

### Claim 17 (depends on 16)

The method of claim 16, further comprising: in response to receiving a positive response rate for a message variant with a particular key component value, testing other values for such particular key component value that are within the same family as the value of the particular key component associated with the positive response rate; and
in response to receiving a negative response rate for a message variant with a particular key component value, testing other values for such particular key component that are in a different family as the value of the particular key component associated with the negative response rate.

### Claim 18 (depends on 12)

The computer program of claim 12 wherein the messages variants are created and tested in accordance with a supervised learning method to identify the key component value combinations that maximize response rate.

### Claim 19 (depends on 12)

The computer program of claim 12, wherein regression analysis is applied to the measured response rate for each of the tested message variants to identify the best performing key component values.

### Claim 20 (depends on 12)

The computer program of claim 12, wherein, in creating message variants, the message is reduced to a vector and each variant is represented by a different value of the vector.

### Claim 21 (depends on 12)

The computer program of claim 12, wherein rules for manipulating key components are substantially in the form of a context-free grammar.
