Using Natural Language Processing for Deeper and Broader Customer Support Insights
There are a lot of customer service channels available today. Customers expect multichannel experiences and the ability to engage via their channel of preference. Businesses that fail to offer their customers omnichannel support options are at competitive disadvantages. Some of the ways that customers expect businesses to engage with them when it comes to support include:
• Live Web Chat
• Social Media
• Mobile Apps
• Online Knowledgebase
• Online Communities—Company and Third Party
• Support Portals
Customers also want to use any number of devices, with mobile—smartphones and now smartwatches—transforming when and where engagement takes place.
The implications of not offering customers the support engagement channels they want to use isn’t only confined to enterprises. Technology disruption has shrunk the barriers to entry so that small businesses can tap into multichannel support delivery just as quickly and easily as much larger enterprise counterparts. And businesses don’t even need to have staff in place to add these support channels to their portfolio with the availability of managed services such as Davinci’s Live Receptionist and Live Web Chat services and automated chatbot-like services like Davinci Auto Receptionists.
Chatbot support engagement is the hottest topic and trend today. The accuracy of natural language processing (NLP) and machine learning—namely, the ability to understand human language—has improved dramatically (with a word error rate of less than 5% for many). As these chatbots improve in identifying behavioral intent, the capabilities of these support chatbots will improve further, providing customers with human-like support conversations.
Possibilities of NLP
NLP offers organizations a broad range of support opportunities. Chatbots are just one area. When each of the different customer engagement channels are considered, the breadth and depth of structured and unstructured data that businesses capture is substantial. However, according research firm IDC, upwards of 80 percent of customer data goes unanalyzed. There are certainly numerous use-case scenarios—both within support organizations and across organizations—when it comes to business insights that can be derived from the analysis of customer data:
• Customer sentiment in aggregate and across each support engagement channel
• Top issues ranked in terms of sentiment polarity and/or frequency of occurrence
• Persona development based on needs and values analysis, personality types, language, social, and emotion tone analyses
Individually or in total, these can be used to deliver better service, faster identification and remediation of product or support issues, more effective marketing experiences, and develop products that resonate with customers better.
Customer data is defined in terms of three factors: volume (amount), velocity (rate with which information changes), and variety (range and types of data). As each of these increase, the depth and breadth of insights expand in scope, revealing hidden patterns and relationships. Analytics can be organic whereby the data determines the actual areas of business insight, or forensic whereby a specific topic or issue is examined.
When it comes to the actual analysis of data, there are four different models:
1. Descriptive. Looks at data based on what is happening now and develops business insights that reflect the status quo.
2. Predictive. Looks at data through the lens of future scenarios (what might happen) and provides a predictive model that forecast these potential use-case outcomes.
3. Diagnostic. Looks at past performance to determine what happened and why.
4. Prescriptive. Deeper analysis that overlays different data sources for cause-effect scenarios that are used to formulate specific rules and recommendations.
NLP in Action
Key-phrase (or topic) analysis augmented with sentiment polarity is a common form of NLP. Using Cognition Insights, a technology-enabled managed service, TIRO Communications has worked with clients across multiple industry segments to pinpoint actionable business insights. Let’s look at some of the recommendations we’ve given clients based on key-phrase and sentiment analysis as well as advanced persona development.
Example One. A software client discovered that its sentiment polarity is much higher than that reflected on external review sites in comparison to competitors (viz., rating scales of 1 – 5 or 1 – 10 are subjective whereas NLP is based on customer language). The same client also honed its customer advocacy program, identifying topics where additional reviews are needed and opportunities where reviews provide competitive advantage.
Example Two. Another client in the unified communications sector thought it had a problem with audio quality. The company did not know to what extent, and how it stacked up against key competitors. Cognition Insights helped them understand the frequency and types of audio quality mentions by customers and competitor customers (forensics analysis). It also revealed the nature of support issues, which the client prioritized based on the extent of the problem and ease of resolution.
Example Three. A software client wanted to base its demand-gen campaign calendar for the year on the top topics identified by key-phrase and sentiment analysis. Competitive comparison revealed topics where the client had the greatest likelihood to displace individual competitors.
Example Four. A client offering consultative services to global enterprises wanted to improve the effectiveness of its demand-gen and sales initiatives. Analyzing the natural language of buyers for its top 40 accounts, the client derived and assigned a persona grid to each buyer (a matrix of 192 possibilities) to fine tune language tone in sales outreach emails based on those assignments and to identify content types most likely to resonant with each buyer.
One of the most obvious takeaways from the above examples is the fact that customer support data provides actionable insights across multiple organizational functions. Another takeaway is that the same data can yield different types of insights for different business areas. For example, key-phrase and sentiment analysis can uncover business process support issues, product quality areas for remediation, and competitive strengths, weaknesses, opportunities, and threats (SWOT) for sales battle-card development.
NLP is creating some exciting opportunities for customer support. Organizations large and small are leveraging the data gathered across each engagement channel to create business insights that deliver tangible value that includes identification and remediation of support and product problems, more effective marketing campaigns, better content utilization rates, segmented customer advocacy activities, and communications and content that align with customer personas.