Analytics Lab

Creating customer support innovation with 4x return, and doing it safely


Help a globally-distributed service organization innovate the delivery of its premium services to meet the ever-increasing expectations of its customers. The client, a large technology conglomerate, offers a portfolio of premium support services for helping its largest and most valuable customers with the design, deployment, and maintenance of mission-critical networks. Although historically the service has been enormously successful and has one of the highest customer satisfaction scores in the industry, it is now feeling pressure from continued increase in customers’ expectations as well as a corresponding increase in the complexity of the underlying service components. However, innovating inside the constraints of a large and profitable organization is challenging. These premium services are delivered to hundreds of multi-national customers by a global delivery team numbering more than a thousand support engineers, operations managers and relationship managers. In order to minimize disruption and ensure delivery consistency, new capabilities and tools are rolled out globally and en masse. As a result, innovation has slowed to the point where even relatively simple capabilities such as service level scorecards require several quarters to move from conception to launch.

Leadership was faced with a classic example of the innovator’s dilemma – how can it innovate rapidly without disrupting current business operations?


Leadership had two important questions: how to identify, develop and implement innovative capabilities inside the organization, and how to understand and manage the risks associated with disruption. To help answer these questions, Chobanian Group launched a Support Services Analytics Lab that brought together user experience research, digital design, and data science. The lab offered a rigorous yet lightweight mechanism to ideate and prototype innovative capabilities that could be deployed within existing service delivery workflows.

Successful innovation is more than just building platforms, tools, and models. Through a rapid series of group and one-on-one user research activities with global service teams, CG worked to understand the needs and work processes of the stakeholders. The results of these sessions enabled a focused ideation and prioritization process with leadership to create an innovation roadmap that balanced quick-win capabilities, like service performance benchmarking, with longer-term strategic initiatives like digital assistants.

Once the capabilities were identified, the data science team got to work. Fundamental machine learning models and analytics were developed as flexible micro-services that could be leveraged by downstream applications. The digital design team then prototyped new tools, built on top of these micro-services, that could be integrated into the service teams’ current workflows. Lastly, the user research and adoption teams worked to rapidly roll-out these tools to small groups of test users and get feedback.

Service Orginization New Capabilities

  • Machine Learning Case Prediction
  • Natural Language Summarization
  • KPI’s Forecasting
  • Customer Network Benchmarking


Within two quarters, the service organization was already developing and rolling out new capabilities, with adoption rates nearly double the historical norms.

Some of the new capabilities include:
  • Machine learning models that can predict 50% of all support case escalations with more than 24 hours advance warning
  • Natural language summarization techniques that can reduce the length of technical case records by over 90%
  • Forecasting models for key case KPI’s such as solution time and case closure that can reduce SLA/SLO violations by 80%
  • Similarity-based benchmarking for complex customer networks to contextualize performance and identify areas for improvement

These new capabilities have already begun transitioning the offer from break-fix support to being predictive and pre-emptive. Real-time alerts allow operations managers to easily monitor hundreds of cases for potential problems. Natural language summarization has dramatically reduced the time engineers need to spend in order to get caught up on the latest development in a case. Predictive models have sped up the escalation process by allowing operations managers to proactively intervene in cases before a customer becomes frustrated.

For more information, contact us at:

Related Work