System and method for prompt modification based on caller hang ups in IVRS
In a method for use in Interactive Voice Response (IVR) systems, an IVR is configured such that voice prompts provided to a caller vary systematically and caller hang-up activity is attributed to the particular version of a prompt played to the caller. Voice prompt modifications are chosen based on how long the caller willingly listens to the voice prompt. A relatively short time listening to a prompt before hang-up, indicates dissatisfaction with that prompt. The system compares caller hang-up rates for each of the phrases in a prompt and chooses the optimum solution, which is the variant of the prompt with the longest caller listening time. Optionally, the system compares hang up rates to a threshold and contingent on that comparison chooses an alternative version of the prompt.
1. A method for evaluating an effectiveness of a prompt sequence played by an interactive voice response system, the prompt sequence containing a plurality of separate phrases, the method comprising:
by an interactive voice response computer, playing the prompt sequence to a caller;
by the interactive voice response computer, receiving an indication of a caller event during the playing of the prompt sequence, the indication including a timestamp;
by the interactive voice response computer, calculating a shifted timestamp by compensating the timestamp for at least one of a system processing time and a caller reaction time;
by the interactive voice response computer, mapping the shifted timestamp to a particular phrase of the plurality of separate phrases;
determining that a second caller event is a caller hang-up of a misdialed call, based on a second shifted timestamp falling immediately after initiating the playing of the prompt sequence;
mapping a plurality of shifted timestamps of a plurality of callers to the particular phrase, wherein the second caller event is removed from the mapping as a caller hang-up, and wherein evaluating the effectiveness of the particular phrase is based on the mapping of the plurality of shifted timestamps of the plurality of callers to the particular phrase; and
evaluating the effectiveness of the particular phrase based on the mapping of the shifted timestamps.
2. The method of claim 1 wherein evaluating the effectiveness of the particular phrase is based on a caller event rate determined as a number of caller events having shifted timestamps mapped to the particular phrase.
3. The method of claim 2 , further comprising:
normalizing the caller event rate for a length of the particular phrase.
4. The method of claim 1 , further comprising evaluating an effectiveness of an alternative phrase, and replacing the particular phrase with the alternative phrase if the alternative phrase is more effective.
5. The method of claim 1 , wherein the caller event is a caller hang-up.
6. The method of claim 1 , wherein the caller event is a caller request for a human operator.
7. A non-transitory computer-readable medium having computer readable instructions stored thereon for evaluating an effectiveness of a prompt sequence played by an interactive voice response system, the prompt sequence containing a plurality of separate phrases, wherein execution of the computer readable instructions by a processor causes the processor to perform operations comprising:
playing the prompt sequence to a caller;
receiving an indication of a caller event during the playing of the prompt sequence, the indication including a timestamp;
calculating a shifted timestamp by compensating the timestamp for at least one of a system processing time and a caller reaction time;
mapping the shifted timestamp to a particular phrase of the plurality of separate phrases; and
determining that a second caller event is a caller hang-up of a misdialed call, based on a second shifted timestamp falling immediately after initiating the playing of the prompt sequence;
mapping a plurality of shifted timestamps of a plurality of callers to the particular phrase, wherein the second caller event is removed from the mapping as a caller hang-up, and wherein evaluating the effectiveness of the particular phrase is based on the mapping of the plurality of shifted timestamps of the plurality of callers to the particular phrase; and
evaluating the effectiveness of the particular phrase based on the mapping of the shifted timestamps.
8. The non-transitory computer-readable medium of claim 7 , wherein evaluating the effectiveness of the particular phrase is based on a caller event rate determined as a number of caller events having shifted timestamps mapped to the particular phrase.
9. The non-transitory computer-readable medium of claim 8 , wherein the operations further comprise:
normalizing the caller event rate for a length of the phrase.
10. The non-transitory computer-readable medium of claim 7 , wherein the caller event is a caller hang-up.
11. The non-transitory computer-readable medium of claim 7 , wherein the caller event is a caller request for a human operator.
12. The non-transitory computer-readable medium of claim 7 , further comprising evaluating an effectiveness of an alternative phrase, and replacing the particular phrase with the alternative phrase if the alternative phrase is more effective.