By Seth Grimes, Founding Chair, Sentiment Analysis Symposium
Given: Sentiment — opinion, emotion, and attitude — is a clear customer-experience indicator. Discover business value in feelings in their many forms via sentiment analysis, technology that helps you measure and quantify opinion that helps you respond and engage. This article will discuss the sentiment-customer experience link and analysis approaches that derive value.
Solutions deliver customer insights, linked to product and service features if done right. This detailed, explanatory analysis complements common high-level satisfaction measures such as the Net Promoter Score. Whether applied to enterprise feedback (such as survey responses) or for social monitoring and response, sentiment technologies deliver a key Voice of the Customer asset.
The experts agree. “Emotional” is one of three key experience components, according to customer-experience authority Bruce Temkin. Forrester places emotional engagement at the top of its Customer Experience Pyramid. The challenge then, for brands, agencies, and researchers, is to accurately measure sentiment and transform findings into actionable customer-experience strategy.
Sentiment analysis, via text analytics, has been part of market-leading solutions for several years now. Technical approaches most often start with a lexicon that assigns words — “good,” “fast,” “expensive,” “hot” — to positive and negative categories. Complications quickly become clear. That coffee is hot is (typically) a good thing, while a hot room is not, and the “hot” in “hot chocolate” is merely descriptive. Add in words such as “not” that reverse meaning, modifiers such as “very,” and a world of idiom, metaphor, abbreviations, and emoticons — plus language and cultural complications — and you face a complex analytical challenge.
These problems are solvable. More-advanced solutions are adapted to particular industries, business functions, and information sources. They accommodate differences in vocabularies and expressions, for instance in hospitality, consumer electronics, and healthcare — think of the very different significance of “thin” and “hot” in each of these domains — and in contact-center transcripts, survey responses, and social postings. They’re tailored to support disparate customer service, market research, and corporate compliance needs, and they’re fully translated to each supported language.
Solution providers deliver adaptations — adaptations equate to accuracy, relevance, and completeness — via technical approaches that may include syntactic rules, taxonomies and ontologies, and lexical networks, often allowing for customization by business users and for improvement via technologies such as machine learning.
I don’t have space for a detailed explanation of syntactic rules and lexical networks; but you don’t need to fully grasp the science to make use of the technologies, if you opt for a comprehensive, higher-level solution or service rather than a lower-lower toolkit. Code libraries and programming environments are great for developers but not suitable for most business users, who look for graphical interfaces that insulate them from implementation complexity.
What other criteria matter, in addition to adaptation to your industry and business tasks, solution accuracy, and insulation from technical details? I’ll offer three:
1. Business alignment. Your solution should fit your business rather than force you to rework your business processes and practices.
2. Data availability and integration. Data is key, so your solution should make it easy to bring in data from transactional and operational systems and from online and social sources as well as enterprise feedback. Vendor alliances and vendor-provided data adapters can speed time-to-insight.
3. Results usability. Many solutions, particularly in the social-media monitoring space, surface stats via glitzy dashboards that do little to inform your business decision making. Your solution should provide usable, useful insights that you can act on, with the capability to drill through to particular customer interactions.
The best sentiment technologies are accurate, reliable, and accessible, of central value in any comprehensive customer experience, market research, or social engagement program. I’ve described challenges, solutions, and benefits and covered basic solution requirements. Via the right solution choice, you will be well positioned to translate sentiment — opinion, emotion, and attitude — into business sense.
About the Author: Seth Grimes is an analytics strategist with Washington DC based Alta Plana Corporation. He organizes the Sentiment Analysis Symposium, a twice-yearly conference, and writes frequently for InformationWeek, Social Media Explorer, and other publications. He is a leading industry analyst covering text analytics and sentiment analysis. Follow him on Twitter @SethGrimes.