Customer sentiment analysis is a method of processing information, generally in text format and often from social media sources, to determine customer opinions and responses. Analysis of the data allows organizations to assess whether customer reaction to a new product was positive or negative, or whether owners of a product are experiencing major technical difficulties. Analysis of aggregated data over time provides insights into trends, while analysis of individual cases in near real time lets companies address and resolve customer issues quickly.
At the heart of customer sentiment is text analysis, a complex process based on statistical and linguistic analyses. Text analysis is used for many different applications, including fraud detection and analysis of scientific or intelligence data. The broader the range of content, the more difficult it is to get a clear interpretation. In addition, many of the social media streams are filled with slang, abbreviations and sarcasm, all of which are difficult for analytical tools to process. Depending on the application and the software tool, users of customer sentiment solutions have varying degrees of success.
Three general categories of text analytics tools are available:
- The first is the analytics engine itself, which is designed to be incorporated into an application or a platform.
- The second category is a point solution aimed at a particular function such as analyzing the data from social media streams such as Twitter. This type of solution may be offered as a service by a third-party provider, with the results delivered through reports or on the client's dashboard, or may be used directly by the organization.
- Third, platform-based solutions integrate customer feedback from multiple channels, analyze the results and push them out to the appropriate department such as marketing or customer service.
Analytics engine does the work
Lexalytics provides an analytic engine called Salience, which can be incorporated into applications or platforms and used for many types of text analytics, including customer sentiment. Salience is incorporated into products that are used by data services such as Bazaarvoice, which offers social media monitoring, and by companies such as Oracle, which has integrated Salience into Oracle Endeca for text analytics.
Salience first breaks sentences into parts of speech, and looks for phrases that match a dictionary of sentiment-bearing phrases. By associating the phrases to entities in the text with natural language techniques, Salience can score entities for sentiment. Those phrases are then related to entities in the text. Seth Redmore, VP of product management at Lexalytics, says, "We assess sentiment as it relates to each entity, which allows a granular approach to sentiment analysis. We were also the first company to ship a Twitter-specific vocabulary and to handle emoticons."
Text analytics in general and sentiment analysis as a specialty area are fairly pervasive, according to Leslie Owens, research director and principal analyst at Forrester. "Entry barriers are quite low, especially for SaaS-based products, so adoption of social media monitoring and sentiment analysis is widespread in many industries," says Owens. "What is more difficult is to convert these analyses into actionable information."
Platforms with a broad reach
A key trend in text analytics is to improve the delivery of the output from social media monitoring to those who can make use of the information. "After the data is processed by cleansing, clustering and analysis, it needs to be passed along to the lines of business that will act on it," Owens explains. Platforms such as those offered by Attensity and Clarabridge have extended their capabilities to automate routing information to the appropriate recipients and providing mechanisms for responding.
Attensity's social analytics and engagement solutions include Attensity Pipeline, which collects data from more than 150 social media and online sources, including Twitter, Facebook, Google Plus and YouTube. "In addition to social media, we incorporate data from e-mails, surveys, CRM systems and call center notes," says Catherine van Zuylen, VP of products at Attensity, "and can detect positive and negative attitudes toward entities such as products, people and companies."
The near real-time performance of Attensity allows quick response to issues that are being discussed in social media outlets. "Social media is the canary in the coal mine," van Zuylen says. "It provides early warning of issues that can become major problems if they are not detected quickly. We are seeing a big commitment to real-time interaction." For example, an airline that was monitoring comments from its customers noted that some of them were sending Tweets indicating they were going to miss a flight due to long lines. The airline was able to add agents and alleviate the problem.
The component that provides those capabilities is Attensity Respond, which is geared toward customer service, and detects customer dissatisfaction, suggestions and opportunities. It also allows automated responses for up to 90 percent of incoming communications, and can manage escalation and provide visibility of high-priority queues. The reporting feature in Attensity Respond enables comparing sentiment analysis results over time.
Attensity's business process and workflow technology manages the sequence of events for customer interactions, access by agents to different types of information and processes for escalating calls. It also determines the route that sentiment analysis information takes after it is generated. "Attensity can route Facebook messages that indicate an intent to churn or that contain negative sentiment directly to the right contact agent or salesperson," van Zuylen explains. "This is a huge benefit for customer service."