7 Ways to Mine CX Coal

By: Dheepan Ramanan

December 15, 2015

Tags:
Sentiment Analysis
social listening
Speech Analytics
text analytics
Voice of the Customer

When you were a kid, Santa Claus may have threatened to add you to his “naughty list,” however you were lucky enough to be spared the lump of coal in your stocking on Christmas morning. Unlike those past Christmas morns of yore, receiving customer feedback is not always such a cheerful experience. Every good organization experiences their fair share of CX coal. However, it is especially critical to mine that negative feedback to find actionable insights. As part of our seasons givings, we wanted to impart 7 ways to mine CX coal. Let’s get digging!

1. Monitor trends and spikes in themes

It’s not only important to view overall volumes of feedback, but also to see how they shift over time. Trending data allows your analysts to see movement in theme volumes by days, weeks, years, or even in a small 24-hour period. When you witness spikes in data, you have the opportunity to rapidly react to customer trends or to monitor on-going issues to gauge their importance.

Example: A global chain gauged reactions to a risky campaign, and pulled the plug when the negative voices were clearly outweighing positive ones.

2. Apply sentiment analysis to find areas of customer unhappiness

Sentiment analysis is an incredibly useful tool in analyzing all sources of unstructured data. Sentiment analysis examines your data for positive or negative context, giving you a second dimension of analysis beyond volume or Key Performance Indicator (KPI) metrics. Using sentiment analysis, you can see how different themes affect customer experience, either positively or negatively. Additionally, sentiment can act as a singular metric of performance across all sets of data.

Example: Despite mostly high marks a service provider had low scores in one area. They drilled down, found issues, and made fixes to raise sentiment.

3. Segment to identify key differences

Segmenting data makes it easy to find drivers of customer satisfaction among different groups. What themes do high-level revenue customers like more, compared to lower revenue customers? Which of these themes do high-value customers mention most often? Analyzing different segments of your data using the same models lets you locate points of differentiation.

Example: By segmenting their feedback data, a consumer advocacy group was able to isolate the different needs of members of different ethnicities.

4. Use KPI metrics in conjunction with unstructured themes

You’re probably already using KPI metrics such as review rating, Customer Satisfaction scores (CSAT), or Net Promoter Score (NPS). However, additional insights can be found by mapping these metrics against unstructured feedback to reveal the “why” driving your scores.

Example: A leading managed cloud provider gets more value from two survey questions (NPS and “why?”) than they used to from fifteen.

5. Pair sentiment analysis with CSAT metrics

Looking at KPI metrics in conjunction with themes shows you the high- and low-performing touchpoints along the customer journey; however it does not always show you the best ways to improve those metrics. For example, you may find that long call wait times produce low sentiment—but they are also associated with average CSAT scores. This tells you that although people are unhappy with call waiting, it does not significantly impact their overall customer experience.

On the other hand, you might find that low sentiment regarding website Navigation is also associated with low CSAT. Improving your website, then, is likely to generate improvements to your overall CSAT score.

Example: Customers of a large financial org complained about service descriptions and gave low CSAT scores. They fixed the text & scores went up.

6. Perform multivariate regression to find root cause

It can be difficult to find the root cause when customer volume, sentiment, or satisfaction scores start moving. This difficulty is multiplied when multiple data sources are thrown into the mix, and when several layers of structured attributes may be contributing to the increase or decrease. Using multivariate regression, multiple attributes, themes, and Natural Language Processing (NLP) attributes (word proximity and relationships) can be analyzed at once. Then the significance of these factors can be measured, unlocking the most important issues.

Example: Multivariate regression revealed the root cause of low sentiment for a cafe chain: birds interfering with customers sitting outdoors.

7. Combine all techniques to create a data-driven culture

Data analysis can be incredibly powerful, but you will never realize its value if you don’t take action on the insights you uncover.  Use dashboards, alerts, and case management functionality to push data and insights throughout the organization. These tools allow the information to be delivered to the right person at the right time to take action, opening communication across departments and at all levels of leadership. The volume of customer feedback is never going to slow down—your organization must learn to absorb, react, and learn from data faster and more efficiently than ever.

Example:  Corporation unites 50 sources of customer data to identify issues and improvements, and impacts $110 million in customer transactions.

No one wants to get coal in their CX stocking, but if you have the tools to mine insights, it can be a valuable gift that keeps on giving.

We are counting down the 12 Days of CXMas. Don’t miss out!

The 12 Days of CXMas

11 Jolly Customer Experience Quotes

10 CX Stars Shining Brightly

9 Merry Customer Engagement Tips

8 Ways to Avoid the CX Naughty List

 


Dheepan Ramanan is a data scientist at Clarabridge. Follow him on Twitter @DheepanRamanan.
2016-1486_clarabridge_demo_request_cta