Call Center Case Study 2018-01-20T00:38:30+00:00



The global telecom giant leverages DataSkill’s cognitive analytics software, powered by IBM
Watson, to optimize its call center.

At any given time, a telecom call center receives 4,000 to 7,000 calls worldwide. They need to address thousands of questions accurately within a short amount time, or customer satisfaction decreases.

Even the term “call center” can carry a negative connotation, with customers associating it with long wait times and outsourcing to other countries. Speaking of negative experiences, customer satisfaction scores plummet when issues fail to be resolved by the first agent engaged (via chat or phone).

All large-scale call and help centers share a similar goal—decrease agent costs while increasing customer satisfaction. The term “time is money” rings undoubtedly true in this highly competitive environment. For example, based on DataSkill aggregate data, each transferred call in a telecom call center results in 18 minutes of valuable agent time on average. Assuming each minute of agent time costs 50 cents, these transfers can cost an organization as much as $30,000 per week.

The multinational telecom leader is always seeking operational excellence across its call and help centers. This includes incoming customer chat engagement, which can either result in a positive experience (issues resolved quickly in chat) or a negative experience (issues unresolved and transferred to other agents).

The Challenge.

The company set out to augment its call and help center metrics in the following areas:

Call resolution: When an initial call fails to get resolved, it must be escalated to another agent. This not only costs valuable agent time, but customer satisfaction decreases with unresolved issues. Solving problems quickly and efficiently is critical.

Call-handled time: Call and help centers lose money with each extra minute that an agent handles a call. When organizations cut down call-handled time while solving the customer’s problem, that translates to money saved. It also shaves wait times for customers, which results in a more positive experience.

Customer satisfaction: Negative customer surveys and feedback scores will foreshadow customer churn. Therefore, in reference to the two metrics above, call resolution tactics are worthless unless the issue is truly resolved in the mind of the customer; similarly, reducing call-handled time is worthless if the customer feels rushed and unimportant.

The platform is used across a variety of use cases including sales recommendations,
proactive care, retention marketing, contact routing, and customer intent determination.

— Director of Product Development, Global Telecom Leader (Source: LinkedIn)

The Solution.

After much vetting and proof of concept, DataSkill and IBM were engaged to optimize the call and help center with cognitive intelligence and analytics. With a focus on the chatbot, DataSkill deployed its market-leading ACUMI DataFind (text analytics) and ACUMI DataLearn (knowledge analytics) software programs, which harness the power of IBM Watson.

“Using the client’s data, we identifi ed optimal pipeline confi gurations in terms of mapping chat text for routing,” said Dimitri Popolov, Chief Data Scientist at DataSkill. “The data gave new insight into reasons for chat and call transfers, which result in a negative customer experience.”

For example, a common password reset request can be resolved with ACUMI/Watson technology. Through machine learning, this technology trains data to handle that issue, freeing up an agent’s time to address other questions. Also, ACUMI and Watson translate common language into industry terminology, helping representatives identify problems faster.

The software uses a high level of automation to interpret human questions and human intent, resulting in more one-call (or one-chat) resolutions. This not only elevates customer satisfaction, but it radically reduces call and help center costs. It also automatically shares the knowledge of “high-levewl” agents with “low-level” agents (on their screens). This not only gets problems resolved earlier, but it brings all agents up to the knowledge level of top agents.

“The analytics platform uses statistical analysis and machine learning to drive relevancy …,” reports the telecom leader’s Director of Product Development. “The platform is used across a variety of use cases including sales recommendations, proactive care, retention marketing, contact routing, and customer intent determination.” (Source: LinkedIn)

Upon deployment of the new solution, the company reported an immediate and positive impact around call resolution, call-handled time and customer satisfaction across its call and help center.

DataSkill ACUMI benefits for call centers

  • Detects problems, setting service representatives up for success
  • Scales the expertise from experts and shares it across the organization
  • Provides real-time reactions to huge volumes of data
  • Automates “knowledge worker” intelligence with cognitive models
  • Reduces workload through automatic identification and classification of cases

ACUMI Markets

  • Telecommunications
  • Call Center

ACUMI Product + IBM Solution

  • ACUMI DataFind
  • ACUMI DataLearn
  • IBM Watson
  • Watson Explorer
  • SPSS Modeler

Cognitive Science

  • NLP
  • Ontology Generation
  • Concept Extraction
  • Unstructured Data Analytics
  • Predictive Modeling
  • Neural Networks
  • Deep Machine Learning


  • On Premises installation of hardware and run over network
  • Virtual machine must have 300GB of available hard drive space and at least 8MB of ram.
  • Virus scanning must be turned off for the folder containing WEX data
  • Client is responsible for the process that runs daily, which adds to the SQL database any new data from the past day. This must be a fast process, not removing/replacing all data, but incrementally loading the new data.