Published in Financial Services

Improving NPS Scores with Interactive Voice Response (IVR)

Addressing Systemic Issues and Using NLP Analytics with Unsolicited Feedback to Eliminate Failure Demand

A leading healthcare insurance company with one of the most sophisticated voice of the customer programs in Australia is always looking for new sources of insights that go beyond surveys.

The healtchare insurance CX team was tasked with addressing a critical pain point for customers (excessive contact centre wait times), by reducing failure demand (avoidable call volume) rather then increasing available staff.

Being a low-cost insurer, the available project budget was low requiring clever use of data and creativity on behalf of the CX team who looked to leverage existing voice of customer capabilities.

To determine what was driving the biggest volumes of avoidable calls, the team reviewed multiple data sets including call reason codes, agent call notes, NPS survey results, complaints data, agent feedback, and qualitative research. No individual data source delivered sufficient levels of insight into the underlying challenges nor were they providing adequate coverage across all calls they were receiving.

As a result, the companny leveraged NLP (Natural Language Processing) powered text analytics which was applied to 12 months worth of agent call notes. The subsequent analysis generated a custom text analytics category set incorporating 184 unique categories.

Tens of thousands of call notes were then categorised at a phrase level with sentiment also assigned and each comment was married up to detailed operational data to enrich the analysis.

Addressing Pain Points and Taking Action Leads to Positive Financial Impact

By addressing the pain points surfaced through the failure demand analysis, the CX team was able to REDUCE 20% of total calls coming into the call centre.  The financial impact was very significant, saving the company millions of dollars in operating costs and ensuring that the business did not have to continue expanding its contact centre workforce in line with overall business growth.

The text analytics categorisation and sentiment helped to determine why higher volumes of avoidable calls occurred and enabled them to pull additional segmentation data on the customers making those calls.

For example, the top line analysis and a more detailed analysis identified that

  • 15.5% of total calls related to challenges with payments
  • Only 13% of customers had accessed the OMS (Online Member Services) in the 48 hours before calling

The team then ran CX ideation design workshops with the wider team to determine how to address the payment problems leading to avoidable calls. A total of 29 initiatives were created, varying in size.

They sized and prioritised the ideas from the workshop and came up with recommendations to solve key pain points. Some were very large impacting all of  aspects of the company, while others were small process and design changes.

For example, many customers attempted to make payments via Interactive Voice Response (IVR), however the IVR doesn’t include how much they need to pay, so customers were waiting to speak to an agent. As a result, ahm immediately rolled out IVR enhancements to fix this problem.

What Makes the Interactive Voice Response (IVR) Stand out?

The heathcare insuracne CX team have demonstrated an innovative and extremely effective approach to solving the challenge of reducing failure demand within their contact centre.

To determine what was driving the biggest volumes of avoidable calls, they reviewed multiple data sets including call reason codes, agent call notes, NPS survey results, complaints data, agent feedback and qualitative research.

No individual data source delivered sufficient levels of insight into the underlying challenges nor were they providing adequate coverage across all calls recieved.

As a result, the company leveraged NLP (Natural Language Processing) powered text analytics which was applied to 12 months worth of agent call notes. The subsequent analysis generated a custom text analytics category set incorporating 184 unique categories.

Tens of thousands of call notes were then categorised at a phrase level with sentiment also assigned and each comment was married up to detailed operational data to enrich the analysis.

The key benefits of this approach included understanding multiple topics discussed per call and relative sentiment associated with each topic. The approach also allowed for an objective, scalable analysis that provided a truly representative view of why high volumes of avoidable calls were occurring.

With this insight the team was able to prioritise improvement opportunity areas and created 29 CX improvement initiatives which have all been successfully delivered. The resulting 20% reduction in total call volumes and reduction in average call waiting time has delivered a huge business benefit and improved customer interaction and experiences.

Overview of Results:

  • No additional cost to the business for the improvements made
  • 20% reduction in call volume
  • 29 new CX improvement initiatives identified