Unidentified sounds during predictive dialing to IVR
Discover the causes of why sounds can be generated in predictive dialing to IVR systems.
Table of Contents
Symptom / Need
The user reports unusual behavior during the dialing process in the operation:
- When a call is established through the predictive dialer, the customer hears approximately 20 seconds of audio with unusual sounds, described as breathing, background noise, among others.
- This situation occurs before the call is transferred to the IVR for normal delivery.
- The user reports that the audio was not configured by the operation.
Context / Scenarios
This anomaly occurs under specific technical conditions:
- Affected module: Telephony Service / Predictive Dialer.
- Campaign Configuration : VAGENT type skill with transfer strategy to an IVR diagram.
- Exclusivity of the error: The strange sound does not occur when the call is transferred directly to a human agent skill.
- Scope: The behavior can be replicated in different campaigns within the same operation.
Answer / Solution
Information to be collected
- Confirm the affected transaction with the client.
- Check the percentage of impact.
- Routing point name and number.
- Campaign name, ID of the affected skill .
- IDs, date and time of the recordings that show the impact.
Initial validations
- Try another campaign configured in the same operation that is affected, to validate if the behavior is the same.
- It performs an audit of call flows and campaign configuration.
- Active Component Identification: Determines if the campaign has the AMD (Answering Machine Detection) module active. The perceived audio is a technical response from this algorithm and not an audio file from the campaign.
- Analysis of Activation Mechanics (AMD): This validates whether the call initially has silence from the receiver. When the AMD (Answering Machine Detection) engine detects a silent channel, it automatically triggers specific frequency tones designed to elicit a voice response or activate recognition at the receiver's end. These technical tones, as they travel over the telephone line, can be interpreted by the human ear as ambient noise, breathing, or background sounds, and their primary function is to confirm the presence of a human before handing the call off to the IVR.
- Delivery Flow Validation (IVR vs. Agent): Confirm the call's destination. The system primarily uses this audible validation tone for calls to the IVR to prevent the diagram from being run into empty space or an answering machine. For deliveries to agents, the validation process is different and typically omits this audible step.
- Parameterization Verification: Checks in the interface whether the sound is configurable. The client should be informed that these tones are intrinsic to the server's detection algorithm and cannot be modified or parameterized through the standard configuration interface.
- Technical Conclusion and Closing: If the validations confirm that the perceived audio corresponds to the frequency emitted by the detection engine, it is a case of "Standard System Behavior".
Possible Causes
- Silence Detection : Activation of the AMD algorithm in the presence of a silent channel to force voice recognition.
- Flow to IVR: The system applies more rigorous audio validation before handing the call off to an automated diagram compared to handing it off to a human agent.
Practical Example
In the "micro" operation, when a predictive campaign is initiated, the system establishes contact with the customer. Because the call is routed to an IVR, the system executes the AMD (Analog-Analog-Machine) process; if the customer does not speak immediately, the system plays background noise to try to capture the user's voice. Once a response is detected, it proceeds to transfer the call to the configured diagram.
Considerations
- AMD frequency tones can be misinterpreted by users as breathing noises or laughter due to their "white noise" nature.
- This function is a technical tool to optimize the effectiveness of calls to IVR.
- If the customer requires a different audio experience or customized detection logic, the request exceeds the scope of corrective support and must be channeled as an improvement or custom development request through the Sales area.