[This article first appeared in the Clarabridge Bridgepoints newsletter, Q4 2008.]
Understanding customer satisfaction is a key element of Customer Experience Management (CEM), a practice that benefits organizations that work to respond to the spectrum of customer needs, to hear and act on the “voice of the customer” (VOC). Customer voices can be heard in the course of everyday interactions. These interactions may be direct or indirect. We actively engage customers (and prospects) who visit our Web site or store or hotel, phone the contact center with an account or service issue, or respond to a survey. Other encounters are at-a-distance, for instance, when a customer posts to a review site or blog. Some of these interactions generate transactional data, which is typically captured in CRM systems. Others involve qualitative information in diverse,unstructured forms.
Analytics is key to understanding customer satisfaction – text analytics in particular for unstructured, VOC data – but how to start, and how to turn all of this information into a comprehensive CEM program, targeting satisfactions (and, implicitly, profitability) with measurable success?
Start Small in Each of Four Dimensions
Case studies show that you can succeed by starting small, with a modest, targeted initiative.
- You can target particular segments of your business or particular sets of customers.
- You can monitor and react to information from high-value channels – maybe start with surveys or contact-center notes if you’re concerned with quality or satisfaction issues – and defer bringing in other, lower-impact or difficult-to-attain sources.
- You can start by aggregating customer sentiment by theme. Apply bread-and-butter (fundamental) BI reporting and analysis techniques. Identify and address the issues that most needs attention and plan to expand later to more sophisticated business intelligence, data mining, and predictive modeling.
- You might start with selected line managers and plan to build up gradually to delivering executive/managerial-level performance reports.
These start-small examples are drawn from four CEM dimensions: business processes, information sources, analytical techniques, and users. Whichever you start with, do expect benefits to grow as you extend your efforts to more fully and more deeply cover additional CEM dimensions.
Understand Perspectives; Align for Action
There are two sides to the CEM coin: the customer (request) side and the provider (delivery) side. The two sides must be in balance: each inquiry receives a response, each sale involves product or service delivery, each complaint (or product return or warranty claim) is redressed and should also trigger quality improvement processes. Each of the two facets has both quantitative, transactional elements and softer, qualitative elements.
Established CEM approaches have looked solely at transactions. One leading online customer experience vendor, for example, pitches tools for behavior analysis (request side) and for service optimization (delivery side), unfortunately neglecting qualitative enterprise feedback expressed in the customers’ own voices. The vendor’s products look at Web logs and the like but do not support analysis of the most important, direct customer feedback, which is in textual form, or even of interaction data captured in Customer Relationship Management (CRM) systems.
Studying clickstreams and server-side response times – even replaying a Web-site visitor’s session – will tell you what the customer experienced, but those inputs won’t tell you what attracted the customer or motivated his or her choices. So the quantitative-only approach, while valid, provides no sense of the root-cause sentiment. You miss out on the why. Without strong root-cause understanding, you may be able to improve response to particular behaviors but you may miss and fail to correct fundamental, larger issues and you can’t address motivations.
“Triangulation” is a term that is often heard in the CEM world, describing hybrid application of multiple (i.e., three), correlated measurement and analysis methods that will get you to a more complete and accurate answer than will any single method. That is, while it’s practical to start small, a single source or method won’t deliver the full customer-satisfaction picture.
Text analytics should be one of the triangulation’s vertices, perhaps the first, supplying a means of getting directly at the voice of the customer (or market or patient in other contexts), what the customer is saying in the medium and words that he or she chooses. Behavioral/response analysis – looking at the transactional what (even if not at the qualitative why) of customer interactions – can be an excellent complement to analyzing textual feedback.
The use of text analytics to make sense of unstructured sources, in conjunction with traditional BI applied to operational information such as the interaction history captured in CRM systems and records from transactional databases, creates analytical lift, combined results that are superior to findings from any single method. The combination is particularly effective when you’re working with data from “semi-structured” sources such as surveys, comment forms, warranty claims, etc. For example, joint analysis of “fielded” and coded responses to survey questions like “Rate room cleanliness on a scale of 1 to 5,” and of classifiers such as asking the customer to select from a list of product or service topics, can give a huge boost to analyses of survey verbatims, comments, and free-text descriptions.
Psychological/emotive analysis may be chosen as a third triangulation approach, one that relies on the capability of leading-edge text analytics tools to extract sentiment and emotion. The analysis looks for expressions that reflect personality, values, and motivations in the voice of the customer. Triangulation associates psychological profiles with attitudinal expressions and behaviors — How do certain classes of customers, who express themselves in certain ways, act? — to enhance predictive capabilities, contributing to the management end of the practice of customer experience management.
Focus, Extend, Act
How should you start? Indicators such as the Net Promoter Score will guide you to the highest priority topics for initial investigation. Choose the sources you go after first – surveys, customer e-mail, contact-center notes, forum postings, transactional data, other – based on immediate business goals. There’s much to be gained by cross-pollinating information from disparate sources, but don’t underestimate the effort that may be involved. Focus.
Start with modest, achievable goals where you can hit quick-win ROI targets while learning how to adapt text-analysis and other technologies and processes in a comprehensive analytics program. Plan to apply lessons learned in order to build out to the full customer-experience picture encompassing the spectrum of data sources and analysis methods that make sense for your organizations.
There are no one-size-fits-all approaches to understanding customer satisfaction… except when there are. A good step zero, to lay the groundwork, is to study best practices, to learn what has worked for other businesses similar to yours. You may find examples, and vendor solutions, that will advance you on the path to improved customer satisfaction.