Outbound dialing campaign flow — best practice for disposition mapping

Architect & Flows
1

How should disposition codes in the outbound campaign flow be mapped to our WFM forecasting model?

If the predictive dialer categorizes 40% of calls as ‘No Answer’ and 30% as ‘Answering Machine’, those interactions still consume agent time but produce zero revenue. My Erlang models need to account for this wasted capacity to generate accurate staffing forecasts.

2

To properly map dispositions in Genesys Cloud, please follow this structured approach as outlined in the Resource Center:

  • System Dispositions: Automatically assigned by the platform (e.g., ININ-OUTBOUND-ABANDON, ININ-OUTBOUND-NO-ANSWER). These cannot be modified.
  • Wrap-Up Codes: Manually assigned by the agent after a live conversation (e.g., ‘Sale Made’, ‘Callback Requested’).
  • Contact Sort Dispositions: Used by the campaign sequence logic to determine whether to retry the contact.

Ensure your Architect flow’s ‘Set Wrap-Up Code’ action is aligned with your campaign’s retry rules.

3

Good luck getting this to work on the first try.

The previous admin at this org set up a campaign with 14 custom wrap-up codes, but half of them were mapped to the wrong contact sort disposition. So the dialer was retrying contacts that the agents had already marked as ‘Do Not Call’. Nobody noticed for 6 weeks. The compliance team almost had a heart attack.

Document. Your. Disposition. Mappings.

4

From an audio analysis perspective, you should also consider the ‘Answering Machine Detection’ (AMD) accuracy when mapping dispositions.

The GC AMD engine uses a voice activity detection algorithm that analyzes the first 3-4 seconds of audio. If it misclassifies a live human as an answering machine, the call is auto-disconnected and logged as ‘Machine’. That lost contact is a false negative that your disposition reports will never surface.

5

Before you change the AMD settings, you absolutely must communicate the impact to the agents.

If you disable AMD, agents will suddenly start hearing answering machine greetings instead of live humans. Without proper training and change communication, the agents will panic, hang up prematurely, and your abandonment rate will spike. Roll out a brief training memo at least 48 hours before flipping the toggle.

6

If AMD accuracy is your concern, you can augment the native engine with an AWS-based solution.

Route the initial audio stream through Amazon Transcribe in real-time via a Lambda function. Transcribe the first 4 seconds, check for typical voicemail phrases (‘leave a message after the beep’), and use a Data Action to flag the call accordingly.

# CloudFormation - AMD Augmentation
AMDLambda:
  Type: AWS::Lambda::Function
  Properties:
    Handler: amd_check.handler
    Runtime: python3.11
    Environment:
      Variables:
        TRANSCRIBE_REGION: us-east-1
7

What is our current AMD accuracy rate, and how does it impact our cost per acquisition?

If the machine is hanging up on live customers, we are literally throwing money away. I need the operations team to present the AMD false negative rate at next week’s executive review so we can decide whether to disable it entirely.