Setting Callflow Preferences
The Preferences tab in each block dialog box allows you to apply specific rules as to how a call is handled.
A block inherits settings from the application default preferences, or from the module to which it belongs (if this is not an application module). However, specific preferences set within a block take precedence over the default preferences.
The following table explains how to use the basic preferences:
|ASR Language||Selects the language for the speech recognizer to use to identify speech.||Select en-gb for British English.|
|Before beginning of speech timeout||Specifies the time, in milliseconds, after which the system times out if no input is received from the caller. The timer starts counting after the prompt has finished playing. This triggers the standard apology timeout prompt, followed by a timeout prompt.||The number of times the timeout prompt plays to the caller is determined by the Maximum no input count parameter. If this limit is reached, Genesys Intelligent Automation transfers the caller somewhere else.||The default value is 5000 milliseconds.|
|Collection barge-in||Specifies whether callers can interrupt an announcement that is being played to them in order to collect information.||For example, you might set up the initial prompt in a callflow to ask the caller to select a department. If you enable this option, the caller can make the selection before the prompt finishes.||This option is often enabled for IVR applications, especially when long announcements are used and it is not necessary for a caller to hear all of the options.|
|Confirmation barge-in||Specifies whether callers can interrupt a confirmation announcement.||A confirmation announcement asks callers to confirm information gathered by the system is correct. For example, you might ask the caller to specify his or her full address, and Intelligent Automation plays back this information to confirm it was recognized correctly.||Enable this option for efficiency and speed of use. However, consider areas within the application in which this must be disabled (such as confirmation of a PIN number).|
|Collection high confidence threshold & Collection low confidence threshold||Specifies the upper and lower threshold, from 0.0 to 1.0, to evaluate the quality of the caller’s input during speech recognition. The speech-recognition engine generates this confidence score as an indicator of how closely the caller's utterance matches the phrases specified in the grammar.|| The speech-recognition engine awards a high confidence score when the application receives non-ambiguous input. In this case, the application usually accepts the response and continues. However, if the caller's utterance is assigned a confidence score between the high and low thresholds set, Intelligent Automation can ask for confirmation ("Is this correct?").
ImportantYou can set a menu option to Always confirm to confirm inputs regardless of the value set in the Collection high confidence threshold.
If you use the Confirm if necessary confirmation mode (default setting), the following rules apply:
Depending on your desired outcome, you can amend the threshold values as follows:
In summary, if you want to:
|Set these values depending on the type of information collected. For example, if the application asks the caller for feedback on the IVR experience, you can set a low value for the Collection high confidence threshold as it's not imperative that this information is verified. However, if the application asks for a credit card number, a high threshold can be used to ensure the information is correct.|
|DTMF complete timeout||Specifies the time, in milliseconds, after which the system times out after the caller has stopped entering information in response to a DTMF question.||You might ask the caller to enter a five-digit account number. If this option is set to 0, the system proceeds to the next stage without delay after five digits have been collected.||You might set a low value to avoid a long pause after the last digit is entered.|
|DTMF inter-digit timeout||Specifies the amount of time, in milliseconds, the system waits between each DTMF input character before interpreting the DTMF string.||You might ask the caller to enter a 16-digit credit card number, but the caller only enters 15 digits by mistake. Intelligent Automation waits the specified number of milliseconds before timing out and prompting the caller to re-enter the number.||This depends on the complexity of the question the caller was asked. For example, if you asked the caller to "press 1 for accounts and 2 for sales", no inter-digit time delay is required. However, with requests for longer, more complex information, you can give the caller extra time to finish entering digits.|
|DTMF termination character|| If specified, this is the DTMF character the caller must press after he or she has finished entering DTMF information. This termination character indicates the input is finished (for example #). You must inform callers to use this character in all prompts that require DTMF input. For example: "Please enter your 16-digit credit card number, followed by the # key."
ImportantIf the caller does not press the terminating character, Intelligent Automation waits for the value you specified for DTMF complete timeout before accepting the input. Therefore, if you are using a termination character, you must increase the DTMF complete timeout value.
|If you want the caller to enter a five-digit account number, but you do not want a time delay after the caller has finished, you can ask the caller to enter the number, followed by the # key. This way, the system is sure the caller has finished entering data.||This depends on the nature of the data requested from the caller. If you are asking the caller to press one digit, a termination character might not be necessary. However, if you ask the caller to provide a string of digits, such as a credit card number, you can use a termination character.|
|Input mode(s)|| Specifies which input modes to enable.
||You can use DTMF to collect information relating to significant numbers, such as credit card numbers and account IDs. You might also use speech recognition at times when DTMF is not feasible (for example, asking the customer for a full postal address).||Turn off voice recognition if you are expecting the caller to be calling from a noisy environment.|
|Maximum help count||Specifies the maximum number of times a caller can ask for help during a single Menu or Question block before they are rerouted.||You can route the caller to an agent if he or she asks for help several times during a call.||The default value is 2.|
|Maximum no input count||Specifies the maximum number of times a retry prompt plays to a caller if no caller input is received. At this point the call transfers somewhere else.||If a caller is confused by a particular question, he or she might not provide any feedback. If this option is set, the call routes to another number, such as an agent.||The default value is 2.|
|Maximum recognition timeout||Specifies the maximum amount of time the system wait for the speech-recognition engine to recognize an utterance after it has detected speech. If this time is exceeded, you can initiate a retry prompt.||For example, if the caller is talking in a noisy environment, he or she might stop talking but the speech-recognition engine might think the caller is still speaking. This setting allows you to allocate a set amount of time for the engine to analyze the input.||This value depends on the type of question you have asked the caller. If you anticipate a complicated answer, give the speech-recognition engine a longer time to interpret the response.|
|Maximum recording timeout||Specifies the maximum length of time to allow a caller to make a recording. This option is only relevant to Recording blocks.||}Intelligent Automation can detect when the caller is speaking (without speech recognition) and continues onto the next block when the caller stops speaking. However, if the background noise is too loud, Intelligent Automation might confuse this with speech. Therefore, this parameter sets the absolute maximum time the recording can last.||This value depends on the type of question you have asked the caller. If you have asked an open-ended question, you must leave enough time for the caller to give a complete answer. If you anticipate a short answer, set a quicker timeout. Be aware that if you set a higher value, the caller might stop speaking after 10 seconds but Intelligent Automation hears noise on the line and thinks the caller is still talking.|
|Maximum retry count||Specifies the maximum number of retries you want to allow a caller. A retry might be necessary if a digit entered was not recognized during a DTMF response, or if a low confidence is calculated by the speech-recognition engine during a spoken response.||You might ask the caller to enter an account number using the keypad. In this case, you might want to allow at least one retry in case digits are entered incorrectly. However, if you ask the caller to speak an account number, you might allow more than one retry in case speech recognition is more complicated.||This value depends on the difficulty of what you are asking the caller. The more difficult the question, the more retries you might allow. However, too many retry attempts might frustrate the caller. You can also ensure your retry prompt acknowledges the difficulty the caller might be experiencing.|
|Recognition complete timeout||Specifies the length of silence, in milliseconds, after which the system times out after the caller provides an answer that matches something in the grammar.||If the caller provides a recognized response and then stops talking, this value specifies how long to wait before proceeding. However, if the response is not recognized in the grammar, the Recognition incomplete timeout parameter determines how long the system waits before timing out.|| The longer the expected answer is, the larger this value should be. This timeout setting can be set to the same value (or slightly shorter than) the Recognition incomplete timeout value. For example:
This could be followed up by a confirmation prompt, depending on the confidence score attained by the caller’s answer.
|Recognition incomplete timeout||Specifies the length of silence, in milliseconds, after which the system times out if the caller stops talking but has not yet mentioned any required words or phrases.||For example, if a caller pauses in the middle of providing an answer, Intelligent Automation cannot match the utterance against the grammar. If the caller does not continue, Intelligent Automation waits the specified number of milliseconds before timing out. This value gives the caller time to provide the final piece of the answer.||This value should scale according with the expected length of the utterance. This timeout is normally followed by a retry prompt.|
|Recording complete timeout||Specifies the amount of time to wait, in milliseconds, after a caller provides a recorded response to when Intelligent Automation accepts the recording and continues to the next block. This option is only relevant to Recording blocks.||If you asked the caller for feedback during a questionnaire, you might set an amount of time to wait before proceeding to the next block after he or she has finished answering. This wait gives the caller time to think of whether they want to add anything else.||This value depends on the nature of the question. If it is an open-ended question, you can set this value higher in case the caller decides to add more information. However, if it is a straight-forward answer with a one- or two-word response, you can set a lower value.|
|Recovery mode enabled||Specifies whether or not recovery mode is enabled. Recovery mode uses Recovery prompts to return the caller to the start of the current question, allowing them to attempt the question again and avoid the failure path.||An example prompt might be: "I’m having problems understanding your response. Do you want to try again or speak to an agent?" A response of "Try again" returns the caller to the main menu.||This option is disabled by default.|
|Sensitivity||Sets the speech-recognition engine's sensitivity to noise during input recognition. The value entered must be between 0.0 (least sensitive to noise) and 1.0 (highly sensitive to quiet input). Thus, if you set the property to a low value, the recognizer is less sensitive to noise, but the user must speak more loudly in order to be recognized.||You can set the sensitivity level above 0.5 if you expect the majority of your callers to be in a quiet environment, such as their own home. However, set this value below 0.5 if you expect callers to be in a noisy environment, such as a busy workplace.||The default value is 0.5. If you adjust this value, do so using small increments or decrements, one day or week at a time, to ensure the adjustment does not have a detrimental effect on your customers.|
|Transfer timeout||Specifies the time, in seconds, to wait before a call exits with a result of no answer. If several phone numbers are provisioned, this is the maximum time to wait for each one before timing out.||If you ask the caller to select a department, and no one in the department answers and no voicemail is available, the Intelligent Automation waits for the timeout before routing the caller to the receptionist to allow them to leave a message.||Choose a value that allows people a reasonable amount of time to answer the call – but not too long that the caller loses patience.|
- Contact Genesys for information on advanced preferences not mentioned in the table above.
- If a call exceeds maximum values specified in these parameters (apart from Maximum recording timeout), the block exits with a result of recognition failure. This usually results in the call being routed to an agent, but this behavior can be overridden.
- Getting your timeout values right is key to making your application as pleasant to use as possible. Shorter timeouts mean a snappier response to the caller, but may also mean that the caller gets interrupted by the system before he or she finishes speaking.
- Intelligent Automation enables you to quickly make small, incremental changes to these parameters and deploy these changes to production. You can monitor the resulting statistics over time and, if no improvement is obvious (or the changes are detrimental), roll back the changes to the previous value/version. All this can be done without having to make software changes, do extensive release/testing cycles, or take the application offline.