How Speech Analytics Exposed 18% Repeat Call Volume in a Travel Contact Center 

What Is a Contact Center Speech Analytics Program and Why Does It Matter?

A contact center speech analytics program is a structured capability that uses voice analysis technology to process recorded customer calls and surface patterns in what customers are saying, how agents are responding, and what is driving call volume. Unlike manual quality monitoring, which can realistically review only a small fraction of total interactions, speech analytics operates at scale across every recorded call, providing a statistically valid and complete view of contact center behavior.

For organizations managing high call volumes across complex product lines, the absence of speech analytics creates a fundamental visibility gap: leadership may know how many calls are coming in, but not why. 

Client Opportunity

A large membership retail travel company operated a 500-agent contact center handling millions of member calls annually across multiple lines of business including cruise, vacation packages, hotels, cars, and air. The organization had an established contact center operation with routing infrastructure, a quality monitoring program, and a workforce management function already in place.

Despite that operational infrastructure, the organization lacked granular visibility into why members were calling. Agents used between 6 and 10 manual wrap codes to classify calls after each interaction, producing an incomplete and often unreliable picture of call driver distribution. Quality monitoring was entirely manual, with coaches listening to approximately 2 calls per agent per 28-day period, a coverage rate insufficient to detect systemic patterns or measure program effectiveness at scale.

Andrew Reise was engaged to design and implement a comprehensive speech analytics program, beginning with the cruise line of business as the highest-priority segment, and deliver the strategic framework for extending the capability across all lines of business. 

The Challenge

The contact center was operating with limited analytical visibility into its own call volume, relying on manual processes that produced incomplete data and insufficient quality coverage. Leadership needed an intelligence infrastructure capable of diagnosing call drivers, identifying resolution failures, and supporting self-service and coaching strategies. The challenges compounded each other: each gap in visibility made the others harder to address.

Inadequate call driver data 

With only 6 to 10 manual wrap codes covering a multi-line-of-business operation, the organization had no statistically valid picture of why members were calling. The wrap code taxonomy was too broad to distinguish between distinct call types within a single line of business, and agent-assigned codes introduced classification inconsistency that made trend analysis unreliable.

High repeat call volume 

Repeat calls, defined as a member calling back on an unresolved issue from a prior interaction, represented 18% of all contact center interactions, accounting for more than 73,000 calls. This volume represented both a direct cost in handle time on issues that should have been resolved on first contact and an indirect cost in member dissatisfaction, but the root causes of these repeat contacts were entirely unknown.

Excessive agent hold time 

The cruise line of business had the highest average speed of answer in the contact center, driven in part by unusually long average handle times. Analysis revealed that agents were frequently placing members on hold while calling cruise lines directly to obtain information or execute transactions, averaging 14 minutes and 26 seconds of hold time per call. This behavior was widespread but not yet documented or quantified as a systemic issue.

Manual quality monitoring at insufficient scale 

With coaches reviewing only 2 calls per agent per 28-day period, the quality monitoring program provided a highly limited sample of agent performance. This made it impossible to identify consistent coaching opportunities, assess training effectiveness at a program level, or detect emerging behavioral patterns before they affected member experience at scale.