Despite all of the expert predictions, digital customer-service has not been the proverbial nail in the coffin for the voice (phone) channel. On the contrary, the voice channel has become an even more complicated, emotionally charged, and integral model of communication for customers.
Hence, a successful speech analytics program has become an essential part of any successful contact center.
When digital communication channels fail, customers angry, frustrated, desperate for help, pick up their phones, and make a call. At that pivotal experiential moment, it’s imperative that the agent answering the phone be able to resolve the customer’s issue quickly, expertly, politely. Every company’s brand reputation depends on those pivotal moments!
The Biggest Challenge in Contact Centers Today
The biggest challenge facing organizations today is escalating consumer expectations, as organizations
born of the digital age with modern, integrated data infrastructure and sophisticated analytic capabilities anticipate customer needs and deliver seamless, personalized interactions across channels.
How are mature organizations with aging infrastructure and siloed data expected to compete? By transforming their customer experience one channel at a time, with technology, best practices, and highly-skilled people. And since one must start somewhere, why not start with the complex, emotionally charged channel?
What is Speech Analytics?
Speech Analytics is the technology leading organizations use to gather insights related to customer
experience, brand reputation, process and technology barriers, competitive differentiation, cost containment opportunities, and much more.
Speech Analytics leverages Natural Language Processing (NLP) and Machine Learning (ML) to convert
unstructured human conversations into minable, searchable, structured data. This Artificial Intelligence-
driven technology has incredible transformational power when the right people manage it.
For all of its insight-generating capabilities, this technology is also a prime candidate for ultra-expensive
self-ware. Like many analytic initiatives that preceded it, the power of Speech Analytics is realized when
people act on the insights uncovered. The organizations that are successful in leveraging Speech Analytics to transform their organizations make an investment not just in technology, but in people and
process as well.
People. The Linchpin for Speech Analytics Success.
From a people perspective, there is a minimum of two unique skill sets that a Speech Analytics practice needs:
1. Data operator
This individual is someone with extreme attention to detail, hands-on experience with data integration and harmonization, and background in experience hypothesis testing and testing methodologies. This individual is the tactician on the team, charged with ensuring that each project completed successfully meets the objectives of the project. (Here I would link to the content created for 101 ‘Hiring a Speech Analyst’).
2. Evangelist, story-teller, and influencer
This individual is an expert at translating data and insights into stories and recommended actions that connect with the audience and propel them into action. S/he is a practiced C-suite presenter and influencer, expert at gaining executive buy-in, sponsorship, and partnership. (Here I would link to the content created for 101 ‘Hiring a Speech Analytics Leader’). These two could also be combined into a single document if desired.
To be a valued organizational asset, a Speech Analytics practice needs both of these individuals to gain
the right level of traction within an organization and successfully execute insight projects that deliver actionable insights that are aligned to organizational objectives.
Selecting the Right Speech Analytics Solution
While the Speech Analytics market is relatively mature and the large providers in the space have a significant degree of uniformity in critical features and functions, some key differentiators are essential
to consider before making a technology selection:
1. Phonetic vs. Linguistic approach to understanding human conversation
Some of the technology solutions understand human communication by converting conversation into
phonemes, the building blocks of language, indexing those phonemes, then using AI to understand conversational content (phonetic approach). Other technology solutions transcribe each word uttered in a conversation, then apply AI to understand conversational content(linguistic approach).
Both methods will achieve roughly the same outcomes, and there are benefits and drawbacks to each. Understanding those and evaluating them in the context of the current and future use-cases for Speech Analytics technology within your organization will ensure your technology selection supports future vision & strategy. (Here I would link to the content created for 501 ‘Questions for future SA Stakeholders’ as a way to uncover future needs & strategies).
2. Open vs. closed architecture
Some of the technology solutions available on the market today were developed with the intent of sharing of data smooth, automated, seamless. Other technology solutions were built with flexibility and scalability to bring hundreds, if not thousands of data points into the solution. Yet others make it easy or hard to do both. When selecting a solution, it’s essential to consider your organizational culture. How siloed is your
data? How territorial are your “data owners” and what is their work backlog?
What is the easiest way for your organization to connect Speech Analytics data to other critical data sources? By bringing data into a Speech Analytics solution or by sharing Speech Analytics data to a database, data lake, or the like? By understanding the architectural approach in which Speech Analytics solutions were designed, an organization can select the Speech Analytics solution, which best matches the way the organization functions.
Designing a Process by which Speech Analytics Delivers Value – Today and in the Future
“A goal without a plan is just a wish” – Antoine de Saint-Exupery.
Whenever I establish a new business unit or team, I dedicate time to designing two distinct set
- A set of processes that would govern how my team would vet, prioritize, and execute work, interact with key stakeholders within the organization, deliver value and quantify the financial value of their activities to the organization. All of the details, tools, and resources have been included in the Speech Analytics MasterClass.
- A single process – strategy really – defining the ways in which my team delivered value would evolve over time – based on market trends, competitive trends, emerging technology, evolving customer expectations, and the like. Because of the industry shifts, it is vital that you engage with industry peers.
Advantages of The Analytics Value Model
One of the strategic frameworks I use are below. There are a few reasons why I love this model:
- The model recognizes that one does not need to use complex analytics methodologies in order to deliver value. Phase 1, descriptive analytics, utilizes the most simple of methodologies– addition, averages, medians. However, when combined with posing the right questions (hypothesis creation) and the right data collection methodology, can be immensely powerful and insightful.
- Each stage of the model establishes practices and methodologies that are necessary for the next stage in the model, forming a foundation and confidence for the next stage.
- The final stage of the model has historically been aspirational. Even today, operational, wide-scale execution of prescriptive analytics is elusive, but it is becoming a closer reality daily. In the meantime, it serves as something to aim for at scale, while applying to and executing with key segments of the business or its customers.
Using the Analytics Value Model for a Successful Speech Analytics Program
Stage 1: Descriptive Analytics
Descriptive analytics aim to answer the question ‘what happened?” at some defined point in the past. From a tactical perspective, answering the question ‘what happened?’ is typically done by building specific queries / categories, utilizing KPIs and metrics generated by Speech Analytics solutions, visualizing those KPIs across time, locations, teams, vendors, etc.
From a business value perspective, descriptive analytics can help businesses identify the impact of an outage, marketing campaign, service delivery failure (FCR), industry or environmental event or a competitor action, as well as to compare and contrast KPIs across teams, regions, locations, vendors, customer segments, lines of business and the like.
Stage 2: Diagnostic Analytics
Diagnostic analytics aims to answer the question, ‘why did it happen?” at some defined point in the past. Answering this type of problem could be accomplished through deep-dive projects in which a specific event or failure is studied deeply (for example, studying customer conversations related to a new process or a new marketing campaign), or through analytic methodologies such as root cause analysis, correlation analysis, and anomaly detection.
While these methodologies are slightly more sophisticated than those used in the prior stage of the value model, these methodologies would be within the skill set of an analyst with some foundational coursework in statistics or machine learning. It is important to note that this is the last stage, which is backward looking.
Stage 3: Predictive Analytics
Predictive analytics aims to answer the question, ‘what is likely to happen in the future?” This stage of
the value model brings with it two unique challenges – data and skillset. From a data perspective, the
more sources of data available for model-building, the richer, more accurate, and more actionable the
Examples of data sources (outside of Speech Analytics) could include CRM data, survey data, demographic data, behavioral data, social media data, digital interaction data, and much more! Some of the challenges involved with bringing in these additional data sources include data silos, harmonization, and cleanliness.
From a people perspective, the analytic methodologies at this stage of the value model are more complex, and the skill sets you need to execute the methodologies (predictive modeling, which is statistical in nature, or machine learning which is data-mining-related) are challenging to find and retain.
These individuals may reside within a Speech Analytics Program, within an organization’s predictive or Big Data team, available for engagements through your technology provider’s professional services team or a consulting organization specializing in data analytics.
Stage 4: Prescriptive Analytics
Prescriptive analytics leverages the outcomes of predictive analytics to answer the question, “what can I
do to change or amplify likely outcomes?” Large-scale operationalization of Predictive Analytics via Prescriptive Analytics remains elusive, given the data environments that exist within most organizations
but can deliver significant value when operationalized on a small-scale.
For example, a predictive model may reveal the top 4 triggers for customer cancellation. That model may help organizations identify “churn” candidates. A prescriptive model may then determine a specific offer that some customer segments prefer. While the prescriptive model may not be able to “run” in real-time to enable the agent to deliver the optimal offer during the initial call (large-scale operationalization of prescriptive analytics), a follow-up call can be made to retain or win back the customer in a targeted fashion, enabled by prescriptive analytics.
Should your Contact Center Implement a Speech Analytics Program?
Speech Analytics is a powerful, AI-driven technology that elite organizations have utilized up until now. But, to meet consumer needs and expectations, all organizations must implement Speech Analytics practices that live inside organizations that invest in people, processes, and technology.
- Do you know how to select the right speech analytics solution for your speech analytics program?- Click to Tweet
- The analytics value model can help you implement a successful speech analytics program- Click to Tweet
- How well do you understand speech analytics?- Click to Tweet
- Understand these things before implementing a speech analytics program- Click to Tweet
Jim Rembach is the Editor in Chief of the Customer Service Weekly and it’s Podcast host. He is President of CX Global Media and the creator of the Call Center Coach Virtual Leaders Academy. As the host of the Fast Leader Show Podcast, he has interviewed hundreds of experts, authors, academics, researchers, and practitioners on various angles, viewpoints, and perspectives for improving the customer experience. He has held positions in retail operations, contact centers, customer support, customer success, sales, and measured the customer experience. He is a certified Emotional Intelligence practitioner, Employee Retention Specialist, and recipient of numerous industry awards.