“Examine what is said and not who speaks.”
– African proverb
Fill in all the pieces in your CX puzzle with unsolicited, unstructured, omnichannel feedback
No one disputes the need to understand customers if our purpose is to deliver optimal and personalized experiences. Customer insights are a key ingredient of customer experience (CX) management, which is why CX professionals use customer feedback to prioritize investment in better products, services and customer experiences. Thanks to effective customer experience management, companies that are able to monitor and manage that feedback also have a more complete understanding of their customers and can easily measure their satisfaction and build customer loyalty.
Your typical customer insights are stale
Understanding customer needs, perceptions and expectations has always been at the core of experience management. But for many companies the typical ways of collecting customer feedback are based simply on data captured in surveys and panels or, at best, the result of customer interviews and focus groups.And, in a world where every touch point with the brand shapes an experience, these techniques do not provide effective feedback across the entire customer lifecycle. The problems your typical customer insights suffer from are:
It is inaccurate, insincere and non-contextual
This is usually feedback that consists of customer self-report, far in space and time from the natural points of contact with our company and, as such, based on recall and conscious expression. And, as such, it is tainted by errors in that recall and by the respondents’ unconscious desire to be nice and seek approval. It is, therefore, unreliable and inaccurate feedback.
It is contaminated by the internal view of the supplier
Especially in surveys and panels the questions tend to reflect the supplier’s view of customers and their problems. The aim is to identify answers to issues that the supplier already knows he does not know and has previously categorized (his “known unknowns”). This may not match the customers’ worldview and can lead to biased results and loss of unanticipated insights.
It is intrusive and has low response
On many occasions the request for feedback catches the customer at an inopportune time and away from the experience being asked about. This causes it to be perceived as an unwanted interruption and, as a consequence, low response rates or (worse) random answers are given so as not to waste time.
In many ways, these typical techniques capture (poorly) the past. To be able to effectively manage customer feedback our company should be able to frame it in a context of increasingly digital customers. Omnichannel signal collection is the only way to understand the connection between customer needs, perceptions, intentions and behaviors.
That’s why companies are looking at other sources of insights.
Product usage data
Especially in digital products, actual usage data from different users (product analytics) provides the definitive information on preferences and satisfaction. It’s no use that users tell us in a survey that they are delighted with a certain feature if, when it comes down to it, the measured data indicates that no one is using it.
This is feedback data shared by customers without having been asked or requested by the business. Social media (networks, forums, communities) and the omnichannel contact center (voice, email, chat) provide endless opportunities and means for customers to express their issues and opinions. This is a growing trend due to the proliferation of communication technologies and increasingly digital customers. It is to this type of feedback that we dedicate the rest of this post.
“The truth is out there” (The CX-Files)
Why should we care about unsolicited feedback? Unsolicited feedback is not only present throughout all the customer experience and more abundant. It also helps compensate for some of the limitations of traditional solicited feedback.
It is more reliable and honest
Unsolicited feedback is spontaneous, immediate and contextual. The “point of emotion” is the window in which the consumer experiences and attitudes that you need to understand remain vivid and recall is maximized. By being much closer to that point of emotion unsolicited feedback lives “in the moment” – in real time, not in recall time. And it is not tainted by the artificial framing of many solicited feedback techniques, by the desire to please, political correctness or inaccurate recall. Solicited feedback picks up what users say they remember they felt, and that can miss several points. Unsolicited feedback, on the other hand, captures in an unfiltered and possibly more honest way what is felt at the time.
It captures the customer’s point of view
Solicited feedback imposes (in its questions, in the possible answers) a view of the problems from the supplier’s point of view. Unsolicited feedback, on the other hand, is expressed in the customer’s own words and from the customer’s point of view, which allows us to understand how the customer perceives the problems and frustrations he is facing. This free way of expressing oneself, without being subjected to a previous framing, gives rise to the use of more exploratory analysis techniques focused on the discovery of unsuspected issues and aspects.
It is flammable
Solicited feedback is confined within collection channels. But unsolicited feedback is expressed through public, open channels, which opens the possibility of being shared and amplified by all participants in the channel, eventually leading to a reputational crisis. A comment expressed on a social network is not only important for its meaning, but also for its potential for multiplication and expansion.
It is affordable
Feedback expressed in the multichannel contact center or on social media is available in massive quantities and updated in real time at essentially zero incremental cost. Although these characteristics of volume, availability and price also have a downside in terms of representativeness and difficulty of processing.
The use scenarios for unsolicited, multichannel feedback are uncountable. In this section we present two of the most frequent ones. Surely you have experienced them in your company.
The omnichannel contact center (phone, email, chat, social) is one of the main hubs of interaction with our current and potential customers. Both inbound and outbound calls establish a rich and deep dialogue that would be wasteful to neglect to analyze. Interaction analysis allows us to better understand customers, detecting emotions, root causes and drivers of dissatisfaction, discovering churn signals, mentions of competitors and brand associations. All this allows us to improve our offer and better serve our customers. But in the case of the contact center, interaction analysis also empowers the company to optimize the operations of the center itself, for example, by analyzing intent and context to route and assign the call, providing agents with real-time guidance and supervision to bring the interaction to a successful conclusion, detecting satisfaction with the service and levels of quality and compliance, discovering best practices that allow replicating top performers, or preventing fraud.
Review sites and forums
Review sites (e.g. Google My Business, Amazon Customer Reviews, Yelp, TripAdvisor, Trustpilot) and forums (especially industry-focused: finance, health, etc.) are where users publicly comment on their experience using all types of products. It is essential that companies understand and engage with them, as they are the tool that drives feedback and conversation between our customers and us. Reviews and forums not only allow us to capture feedback on our products, but to better understand the product category, detect relevant competitors, identify the hierarchy of attributes that customers use to evaluate the category, discover how they rate competitors on those attributes, and build perception maps. Analyzing reviews and forums allows us to not only respond quickly to potentially harmful comments, but to make strategic decisions that help us differentiate and improve our offering.
The challenges in analyzing unsolicited feedback
But achieving the benefits of analyzing this unsolicited feedback is not for the faint of heart. These scenarios pose the typical challenges of Big Data projects:
- Large amounts of data (Volume): when we think of analyzing calls from a contact center or ratings on a review site we are talking about millions of interactions / comments per month.
- High speed (Velocity): some applications require immediate analysis and response (for example, a reputation crisis on social media or call guidance for contact center agents), which requires a high speed of response.
- Multimedia (Variety): unsolicited feedback does not come in scores from 0 to 10, but in free text, emojis, and even audio and video. Even if we can convert everything to text (for example, by transcribing phone calls), the ambiguity of natural language makes it very difficult to extract its meaning.
These characteristics make the use of automatic methods based on Artificial Intelligence unavoidable. Certainly, while managing and analyzing unstructured customer feedback data has become an increasing challenge for many companies, natural language processing technologies can make this data more manageable and easier to structure.
That’s why a growing number of companies are using these methods and techniques to delve deeper into the customer experience, discover patterns and trends, and extract insights about their customers’ opinions, perceptions, emotions, and motivations.
What you should look for in a solution for analyzing unsolicited feedback
Not all text analytics and natural language processing solutions are valid for these scenarios. Classifying a small number of internal documents according to fixed, unchangeable categories is not the same as trying to extract value from such an abundant and volatile raw material. A suitable solution for analyzing structured feedback should have the following features:
Discover your unknown unknowns
Using predetermined models can limit the viewpoint from which feedback is considered. But unsolicited feedback is totally free in terms of topics, emotional color or type of language. A solution for analyzing free feedback must be powerful enough to go beyond what the company knows it doesn’t know (“known unknowns”) to discover what it doesn’t know it doesn’t know (“known unknowns”). To do this, it must be able to surface the structure of meaning that emerges from the comments, with their levels of importance, without the need to pre-code domain-specific knowledge.
Identify relevant and actionable meaning elements
Understanding comments requires much more than classifying them according to canned categories. In addition to discovering the topics that surface from the data, it is necessary to identify the relationships between them, the polarity expressed in them, the emotions that are made explicit, the drivers behind them, and the users’ intentions. Altogether, a series of elements that express and encode the deep meaning of the interactions. This is the only way to make our insights truly actionable.
Adapting the analysis to our domain in a simple way
Rarely is the result of the first iteration of the analysis the best we can get. As in all data science projects, it is useful to apply an iterative process that refines the analysis by incorporating information from the domain and materializing it in a model adapted to our needs. These models should also be reusable, so that they can be utilized at different times and with different scopes to compare results and detect trends. The solution must have interactive tools that enable an iterative and agile process of refinement and that is desirably intuitive enough to be used by business personnel, not experts in data science.
Collaborate and manage corporate projects
Obtaining customer insights is rarely the endeavor of a single person in the organization and is often deployed in multinational projects that attempt to apply homogeneity in the analysis in order to compare results. These often distributed teams require collaboration features and the management of analysis models so that they can be shared, versioned and adapted. It is also important that the system can be integrated with the tools and processes used in the organization (in this respect, open APIs can be a great advantage). Finally, in international (and necessarily multi-lingual) projects it is necessary that the solution covers the languages involved with the highest possible functionality and quality. To this end, it is recommended that it treats these languages natively (avoiding translating into a common language, analyzing and translating the results back into the original language) and provides facilities for the management of intrinsically multilingual models.
Finally, the solution must accelerate the collection and dissemination of insights and deliver results in a matter of hours, not weeks.
Does your text analytics vendor provide these features? If you want to know how Gavagai can help you don’t hesitate to call us.