Estimating Customer Patience and Abandonment Rates in Text-Based Contact Centers with Silent Abandonment

Core Concepts

Silent abandonment, where customers leave the system without notifying it, is a common phenomenon in text-based contact centers that creates significant information uncertainty and operational challenges. Accurately estimating customer patience and abandonment rates in the presence of silent abandonment is crucial for improving service quality and operational efficiency.

Abstract

The paper focuses on the problem of silent abandonment in text-based contact centers, where customers leave the system without notifying it. This creates information uncertainty, as the system may not be aware that a customer has abandoned, leading to wasted agent effort and inaccurate performance measurements.
The key insights are:

Silent abandonment is a widespread issue in contact centers, with 27.4% of abandoning customers doing so silently in a no-write-in-queue system, and up to 71.5% in a write-in-queue system.

The authors develop classification models to identify silent abandonment in write-in-queue systems, where the distinction between served and silently abandoned customers is more ambiguous.

They propose an expectation-maximization (EM) algorithm to accurately estimate customer patience in the presence of both censored and missing data due to silent abandonment.

Accounting for silent abandonment significantly improves the fit of queueing models to contact center data, highlighting the importance of accurately estimating this phenomenon.

Silent abandonment has operational implications, wasting 1.7% of agent concurrency time in a no-write-in-queue system and 15.3% in a write-in-queue system.

The authors provide a comprehensive analysis of the silent abandonment problem and develop methodologies to address the information uncertainty it creates, enabling more accurate estimation of customer behavior and operational performance in text-based contact centers.

Stats

"On average, 69.5% of abandoning customers do so silently."
"The Sab percentage in the no-write-in-queue contact center is 5.2% in total and 27.4% of the abandoning customers."
"The uSab percentage in the write-in-queue contact center is 24.4%; using the authors' models, 51.8% of these uSab conversations are indeed silent abandonment, hence, Sab represents 71.5% of the total abandonment."
"In the no-write-in-queue contact center, agents waste 1.5% of messages and 0.8% of words on Sab customers."
"In the write-in-queue contact center, agents waste 3.2% of messages and 3.6% of words on Sab customers."
"The system in the write-in-queue contact center spends 15.3% of its agents' concurrency capacity dealing with Sab conversations."

Quotes

"Silent abandonment, where customers leave the system without notifying it, is a common phenomenon in text-based contact centers that creates significant information uncertainty and operational challenges."
"Accounting for silent abandonment significantly improves the fit of queueing models to contact center data, highlighting the importance of accurately estimating this phenomenon."
"Silent abandonment has operational implications, wasting 1.7% of agent concurrency time in a no-write-in-queue system and 15.3% in a write-in-queue system."

Key Insights Distilled From

Silent Abandonment in Contact Centers

by Antonio Cast... at arxiv.org 04-09-2024

https://arxiv.org/pdf/2304.11754.pdf

Deeper Inquiries

How can contact centers leverage the insights from customer writing during waiting to further improve service quality and operational efficiency

Contact centers can leverage the insights from customer writing during waiting to improve service quality and operational efficiency in several ways. Firstly, analyzing the content of customer messages can provide valuable information about their needs, preferences, and issues, allowing agents to tailor their responses more effectively. By understanding the specific concerns of customers before they even engage with an agent, contact centers can streamline the resolution process and provide more personalized assistance.
Secondly, monitoring customer writing patterns during waiting can help identify common issues or recurring themes, enabling contact centers to proactively address these issues and implement preventive measures. For example, if a particular issue is frequently mentioned in customer messages, contact centers can develop self-service resources or FAQs to address these concerns upfront, reducing the need for agent intervention.
Additionally, analyzing customer writing behavior can help contact centers optimize agent allocation and resource management. By identifying customers who are likely to require more time or assistance based on their initial messages, contact centers can prioritize these interactions or assign them to specialized agents with the necessary expertise, improving overall efficiency and customer satisfaction.
Furthermore, leveraging insights from customer writing can also enhance training programs for agents. By reviewing customer messages and interactions, contact centers can identify areas where agents may need additional support or training, enabling them to address common issues more effectively and provide better service to customers.
Overall, by utilizing the information gathered from customer writing during waiting, contact centers can enhance service delivery, optimize operational processes, and ultimately improve the overall customer experience.

What are the potential trade-offs between allowing customers to write during waiting and the challenges posed by silent abandonment in terms of agent productivity and customer experience

Allowing customers to write during waiting in contact centers can offer several benefits, such as providing customers with a sense of control and engagement, reducing perceived wait times, and enabling customers to articulate their concerns more clearly. However, this approach also presents challenges, particularly in terms of silent abandonment and its impact on agent productivity and customer experience.
One potential trade-off is that while allowing customers to write during waiting can enhance customer engagement and satisfaction, it may also lead to an increase in silent abandonment. Customers who write their concerns but then abandon the conversation without notifying the system can create inefficiencies in agent utilization and resource allocation. Agents may spend time responding to messages from customers who have already left the queue, resulting in wasted effort and reduced productivity.
Moreover, silent abandonment can also affect the accuracy of performance metrics and service level agreements, as contact centers may not have a complete understanding of customer behavior and preferences. This can lead to suboptimal decision-making and resource allocation, impacting both operational efficiency and customer satisfaction.
To address these challenges, contact centers need to implement strategies to identify and mitigate silent abandonment, such as developing automated systems to detect and flag abandoned conversations, optimizing agent workflows to minimize wasted effort on abandoned interactions, and enhancing communication channels to encourage customers to provide feedback or indicate when they are leaving the queue.
In conclusion, while allowing customers to write during waiting can enhance the customer experience, contact centers must carefully balance this approach with strategies to address the challenges posed by silent abandonment and ensure optimal agent productivity and operational efficiency.

How can the methodologies developed in this paper be extended to other service environments, such as healthcare systems or ticketing queues, that also face challenges related to silent abandonment and missing data

The methodologies developed in this paper to address silent abandonment and missing data in contact centers can be extended to other service environments facing similar challenges, such as healthcare systems or ticketing queues. By adapting the EM algorithm and classification models to these contexts, organizations can improve their understanding of customer behavior, optimize resource allocation, and enhance service delivery.
In healthcare systems, where patients may silently abandon queues in emergency departments or outpatient clinics, the EM algorithm can be used to estimate patient patience and identify patterns of behavior that impact service utilization. By analyzing patient interactions and wait times, healthcare providers can improve patient flow, reduce wait times, and enhance the overall quality of care.
Similarly, in ticketing queues for events, transportation, or customer service, the methodologies developed in this paper can help organizations better understand customer behavior, optimize queue management, and improve operational efficiency. By leveraging insights from customer interactions and wait times, organizations can streamline service delivery, reduce customer wait times, and enhance the overall customer experience.
Overall, the methodologies and approaches outlined in this paper provide a framework for addressing silent abandonment and missing data in various service environments, offering valuable insights and strategies for improving service quality, operational efficiency, and customer satisfaction.