Message optimization
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.
1. A method performed by a communications system including a communications server in communication with one or more communications devices for automatically optimizing a message text, the method comprising:
receiving on the communications server a message text comprising a plurality of words or word phrases that combine together as non-overlapping parts of the message text;
treating the non-overlapping words or word phrases of the message text as multiple independent variables that are reduced to a message vector having each of the multiple independent variables as components of the message vector;
automatically creating on the communications server a plurality of lexical variants of the message text, wherein the lexical variants are created by replacing a word or word phrase for each of the multiple independent variables with one or more alternate words or word phrases based on one or more value-changing rules being applied to the received word or word phrase in the message text, the lexical variants for each of the multiple independent variables being reduced to a lexical vector such that the message vector is made up of variable-sized lexical vectors;
sending each of the plurality of created lexical variants of the message text to the one or more communications devices;
measuring a response rate for each sent lexical variant of the message text;
identifying one or more lexical variants having the best performing measured response rates for each of the lexical vectors;
automatically creating on the communications server syntactical variants of the identified best performing lexical variants by rearranging the lexical vectors within the message vector based on one or more position-changing rules;
sending a plurality of the syntactical variants of the identified best performing lexical variants to the one or more communications devices, wherein only grammatically-correct syntactical variants are sent;
measuring a response rate for each of the sent syntactical variants; and
identifying a message text having the highest measured response rate for the sent syntactical variants.
2. The method of claim 1 , wherein only a subset of the created lexical variants of the message text is sent to the one or more communication devices and a D-optimal design is used to determine the subset.
3. The method of claim 1 , wherein the lexical variants of the message text are created and tested in a plurality of phases, wherein the performance of lexical variants of the message text in one phase is used to determine the lexical variants of the message text that will be created and tested in a next phase.
4. The method of claim 3 , wherein the lexical variants are categorized into families of words or word phrases having equivalent meaning or effect.
5. The method of claim 4 , further comprising:
in response to receiving a positive response rate for a lexical variant of the message text with a particular independent variable word or word phrase, testing other words or word phases for the independent variable that are within the same family as the word or word phrase of the particular independent variable associated with the positive response rate; and
in response to receiving a negative response rate for a lexical variant of the message text with a particular independent variable word or word phrase, testing other words or word phases for the independent variable that are in a different family as the word or word phrase of the particular independent variable associated with the negative response rate.
6. The method of claim 1 , wherein the lexical variants of the message text are created and tested in accordance with a supervised learning method to identify the independent variable word or word phrase combinations that maximize response rate.
7. The method of claim 1 , wherein regression analysis is applied to the measured response rate for each of the tested lexical variants of the message text to identify the best performing independent variable words or word phrases.
8. The method of claim 1 , wherein rules for manipulating the message vector are substantially in the form of a context-free grammar.
9. A non-transitory computer-readable medium comprising computer program code that, when executed by a communications system including a communications server in communication with one or more communications devices, enables the communications system to perform the following method for optimizing a message text:
receiving on a communications server a message text comprising a plurality of words or word phrases that combine together as non-overlapping parts of the message text;
treating the non-overlapping words or word phrases of the message text as multiple independent variables that are reduced to a message vector having each of the multiple independent variables as components of the message vector;
automatically creating on the communications server a plurality of lexical variants of the message text, wherein the lexical variants are created by replacing a word or word phrase for each of the multiple independent variables with one or more alternate words or word phrases based on one or more value-changing rules being applied to the received word or word phrase in the message text, the lexical variants for each of the multiple independent variables being reduced to a lexical vector such that the message vector is made up of variable-sized lexical vectors;
sending each of the plurality of created lexical variants of the message text to the one or more communications devices;
measuring a response rate for each sent lexical variant of the message text;
identifying one or more lexical variants having the best performing measured response rates for each of the lexical vectors;
automatically creating on the communications server syntactical variants of the identified best performing lexical variants by rearranging the lexical vectors within the message vector based on one or more position-changing rules;
sending a plurality of the syntactical variants of the identified best performing lexical variants to the one or more communications devices, wherein only grammatically-correct syntactical variants are sent;
measuring a response rate for each of the sent syntactical variants; and
identifying a message text having the highest measured response rate for the sent syntactical variants.
10. The non-transitory computer-readable medium of claim 9 , wherein only a subset of the created lexical variants of the message text is sent to the one or more communications devices and a D-optimal design is used to determine the subset.
11. The non-transitory computer-readable medium of claim 9 , wherein the lexical variants of the message text are created and tested in a plurality of phases, wherein the performance of lexical variants of the message text in one phase is used to determine the lexical variants of the message text that will be created and tested in a next phase.
12. The non-transitory computer-readable medium of claim 11 , wherein the lexical variants are categorized into families of words or word phrases having equivalent meaning or effect.
13. The non-transitory computer-readable medium of claim 12 , further comprising:
in response to receiving a positive response rate for a lexical variant of the message text with a particular independent variable word or word phrase, testing other words or word phrases for the independent variable that are within the same family as the word or word phrase of the particular independent variable associated with the positive response rate; and
in response to receiving a negative response rate for a lexical variant of the message text with a particular independent variable word or word phrase, testing other words or word phrases for the independent variable that are in a different family as the word or word phrase of the particular independent variable associated with the negative response rate.
14. The non-transitory computer-readable medium of claim 9 , wherein the lexical variants of the message text are created and tested in accordance with a supervised learning method to identify the independent variable word or word phrase combinations that maximize response rate.
15. The non-transitory computer-readable medium of claim 9 , wherein regression analysis is applied to the measured response rate for each of the tested lexical variants of the message text to identify the best performing independent variable words or word phrases.
16. The non-transitory computer-readable medium of claim 9 , wherein rules for manipulating the message vector are substantially in the form of a context-free grammar.