Breadcrumbs

Predictive Dialer

Overview

In Expertflow CX, predictive dialing is an advanced outbound capability designed for high-volume campaigns. Its primary goal is to maximize agent talk time and minimize idle time by anticipating when agents will become available and dynamically adjusting the call initiation rate.

Unlike traditional platforms, our system does not rely on a simple "predictive dialer mode" toggle switch. Instead, predictive dialing is a highly flexible, flow-based configuration built within the Conversation Studio. By feeding real-time operational data into your flow logic, you can continuously optimize pacing, resource allocation, and compliance to fit your specific business needs.

The Logic: How Predictive Dialing Works

The predictive dialing engine is powered by mathematical algorithms applied to two critical nodes in your campaign flow:

  • SEIZE Node (Resource Allocation): Controls agent reservations. In a predictive setup, it uses a dynamic query to constantly calculate the number of agents that need to be requested based on live metrics (like answering machine rates and active calls). Learn more: As a Campaign Manager | The "Seize Agent" Node

  • INIT Node (Pacing & Initiation): Acts as the gatekeeper for launching calls. Instead of dialing a fixed ratio of contacts per agent (as in Power Dialing), the INIT node uses a formula to "predict" the exact number of outbound calls needed to yield a live human connection just as an agent finishes their previous interaction. Learn more: As a Campaign Manager | Initiating the Contact: Init Node

The Significance of Outbound Metrics

The true intelligence of the predictive dialer lies in its reliance on Outbound Metrics. The predictive engine is not static; it is dynamically fueled by real-time platform data (tracked via Prometheus).

The mathematical formulas you place in the SEIZE and INIT nodes continuously query these live metrics to make split-second decisions. Without these metrics, the formulas would have no data to act upon. For example:

  • Calculating Pacing (Hit Rate): The INIT formula queries ob_call_human_detected_total and ob_contacts_sent_for_dialing_total to understand exactly how many dials it currently takes to get a real human on the line.

  • Protecting Compliance (Abandon Rate): The advanced Compliance Guard formula constantly monitors ob_abandoned_calls_total. If dropped calls inch closer to your defined regulatory limit (e.g., 3%), the metric triggers the formula to apply an automatic "brake" and slow down the dialing speed.

  • Managing Workload: The SEIZE formula looks at ob_currently_seized_agents and ob_agent_seize_requests_pending_total to accurately calculate exactly how many human agents are needed to handle the upcoming call volume.

To see the complete dictionary of these available metrics and how they are classified, refer to Outbound Metrics.

Configuring the Predictive Dialer

To build and manage a predictive dialing campaign, you will use a combination of flow design, algorithmic formulas, and the real-time metrics mentioned above. Refer to the dedicated documents below to complete your setup.

1. Designing the Flow

Setting up the flow logic ensures that the system checks for resources and dials at the correct intervals. Because predictive dialing relies on real-time adjustments, the flow structure differs slightly from static campaigns:

  • Short Pulses: The Start Node is typically set to a shorter pulse interval (e.g., 5–10 seconds) so the system recalculates agent demand frequently.

  • Smart Routing: Calls launched by the INIT node are processed by the Call Decision Node, which uses Call Progress Analysis (CPA) to filter out answering machines and invalid numbers, passing only live humans to the seized agents.

For step-by-step instructions on uploading contacts, adding these nodes to your flow, and managing campaign statuses, refer here: As a Campaign Manager.

2. Applying Predictive Formulas (PromQL)

To turn your standard flow into a predictive dialer, you must insert Prometheus Query Language (PromQL) formulas into the query boxes of your SEIZE and INIT nodes.

  • Demand-Driven SEIZE: Calculates how many agents to seize based on the current list backlog and the probability of reaching a human.

  • Standard Predictive INIT (Inverse Hit Rate): Calculates the live "Hit Rate" of your campaign to dictate how many lines to dial per reserved agent.

  • Compliance Guard INIT: An advanced variation of the standard formula that includes an automatic "brake" based on live abandonment metrics.

  • Learn More: To copy and paste these exact PromQL algorithms into your flow, please refer to [SIEZE / INIT Formulas].