This topic describes how to configure and use predictive dialing in Cloud Contact Center.
Function overview
Predictive dialing meets large-scale outbound calling needs while improving agent utilization. Cloud Contact Center uses AI to predict call outcomes and fully automates dialing pacing. It maximizes agent talk time while meeting your call abandonment requirements. In a bank’s debt collection use case, agent efficiency increased by 50% and labor costs dropped by 50% compared to other predictive dialing systems.
Access the feature
Log on to the Cloud Contact Center console. In the left navigation pane, choose Instance Management - V2 and click the endpoint of a specific instance to enter its console.
Click the menu button in the upper-right corner of the page and select Voice Services to open the Voice Services console. In the left navigation pane, choose Batch Outbound Calls > Predictive Dialing. The predictive dialing management page appears.
Create task
A task is the operational unit for predictive dialing. Define a task by specifying the customer list, agent skill group, calling dates and times, dialing strategy, and caller ID. The scheduling system starts the outbound task based on these settings and automatically places and assigns calls using real-time agent and customer data.
On the predictive dialing management page, click Create Task. A configuration window appears. Fill in the fields based on your business needs and click OK to create the task.
Name: Enter a custom name.
Pattern Policy
Ratio-based dialing: Set a concurrency multiplier. The system dials at the specified ratio above available agents. We recommend setting the multiplier to 1 divided by your historical answer rate. The system supports multipliers from 1× to 10×.
PID model: An algorithm that dynamically adjusts concurrency to reach a stable state. Provide your target abandonment rate and historical answer rate to cold-start the model. During dialing, the model continuously optimizes based on real-time data.
Max concurrency mode: Use the CreateCampaign and AppendCases APIs together. In this mode, add contacts to the campaign using the SDK. The system uses all available concurrency for dialing, which may cause call abandonment.
Customer list: Enter or import a customer list directly. In max concurrency mode, you cannot import or enter a list on the page. Instead, use the API to append contacts to the campaign.
Task period: The start and end dates for the outbound task.
Calling hours: Set one or more daily time windows for dialing within the task period.
Retry attempts: The number of times to redial the same number if the first attempt fails.
Retry interval: The time between redial attempts for the same number.
Skill group: The agent skill group assigned to handle the outbound queue.
Caller ID: Choose from Use skill group-bound number, Number, or Number set.
Use skill group-bound number: Calls use the number bound to the selected skill group.
Number: Select a specific caller ID. Note: This number must be included in the numbers bound to the skill group.
Number set: Select a predefined set of caller IDs. Note: All numbers in the set must be included in the numbers bound to the skill group.
IVR flow: The interactive voice response flow triggered after the call connects.
Flash SMS: Disabled by default. If flash SMS integration is not configured, this option is unavailable. To enable it, configure flash SMS in Settings.
Execute task
After creating a task, click Publish. The system starts dialing based on the task’s calling hours and calling time windows. When a call connects, it enters the IVR flow and routes to an agent based on IVR logic. The scheduling algorithm automatically controls dialing pace using super-concurrent dialing to maximize agent utilization.
View an overview of all tasks in the task list. You can stop or restart any task.
Task scheduling and prediction algorithm
The scheduling strategy is fully automated. During task creation, provide historical data to assist cold-start. The system automatically optimizes during execution without manual intervention. The algorithm considers multiple factors: online agent count, agent idle status, calling frequency limits per number, answer rate, wait time before talk, talk duration, and post-call processing time. It also accounts for call abandonment (when customers hang up while waiting for an agent). Using automatic trials, adjustments, and learning, the system controls dialing pace and scheduling to maximize agent talk time and boost productivity.
Task progress monitoring
During task execution, refresh the page to view real-time progress. Task metrics include the following:
Statistics
Total numbers: Total numbers in the task.
Stopped numbers: Total numbers marked as stopped.
Calls placed: Total outbound call attempts, including retries.
Completion rate: Completed calls ÷ (Total numbers – Stopped numbers).
Completed calls: Number of successfully dialed numbers.
Number answer rate: Answered numbers ÷ Completed calls.
Call answer rate: Answered calls ÷ Calls placed.
Answered: Number of answered calls.
Call loss rate: the number of calls that were connected but not answered by an agent divided by the total number of calls made.
Call loss volume: the number of calls lost.
Agent utilization: (Talk time + Post-call processing time) ÷ Overlap of agent online time and task execution time.
Monitoring data
Concurrent calls: Current concurrency level and change trends, updated every minute.
Agents on call / idle / online: Real-time agent status and change trends, updated as conditions change.
Billing
Predictive dialing uses a pay-as-you-go model. Each call costs CNY 0.016.
To use predictive dialing, submit a ticket request as instructed on the page. Operations staff will configure the service based on your actual usage.