Employee insights can come from a multitude of sources like unstructured data; and, with churn at record levels (overall turnover rate is estimated at 57.3 % per year, and with Gen Z changing jobs at a rate 134% higher than in pre-pandemic 2019)—and the cultural, operational, and customer value discontinuity this can create—it’s essential for every company to have, and apply, every piece of relevant data.  

Information from employees gives businesses power and can be leveraged to enhance customer experience, resulting in higher retention, more positive customer behavior, and stronger business outcomes.

Workforce Analytics and Voice of Employee

Employee data streams come from two principal frameworks: People Analytics and Voice of Employee (VoE).  People Analytics aka Workforce Analytics, are the data sets HR uses to make recruitment more effective, increase retention and longevity, and improve fit, alignment, and productivity. The pandemic has had a profound effect on people analytics, with challenges coming from differing industries, job/role types, and locations. Today’s most successful companies can and do, utilize internal and external data to enhance workforce strategy through better planning.

Voice of Employee is a bit more complex, and given today’s talent landscape and heightened set of employee responsibilities, perhaps even more crucial. VoE programs collect, analyze, and distill employee feedback to identify areas of performance, challenge, and opportunity. These programs were largely manual until recently, which is both costly and time-inefficient. Also, when traditional people analytics tools were applied to unstructured data, the resulting text analytics were superficial, yielding little real actionability. The best and most contemporary approach for employee-generated text analytics is natural language processing, or NLP.

…Organizations that use workforce analytics have the most engaged workforces, and they thrive in tough conditions. 

— Tim Ringo, Workforce Analytics Isn’t as Scary as It Sounds

Leveraging Natural Language Processing for VoE Analytics

With Natural Language Processing for VoE, organizations can gather an in-depth understanding of factors driving emotionally-based behavior and performance, resulting in clear and impactful programmatic recommendations that drive engagement, loyalty, and commitment. 

• Gather: All data sources (surveys, reviews, messages, emails, chat threads, and other communication) can form a single stream

• Process: NLP analyses can be run utilizing HR language, with customized dashboards, or they can be exported to the organization’s business intelligence tool

• Analyze: Identify areas of focus and experience and emotionally-based sentiment with the power of NLP

• Act:  NLP enables narratives on topics, trends, and patterns to be developed, along with root cause issues and supporting data

Figures 1&2 / Polarity data visualization from insurance company reviews

Using NLP helps businesses identify key topics, categories, themes, intentions from every document in the data stream, and detailed sentiment analysis. And, when compared to open source and traditional people analytics techniques, NLP is more efficient and requires less technical support. NLP is, as well, both highly configurable and completely secure with any infrastructure.  

All employees have stories about their experiences and those of customers. NLP helps organizations hear, understand, share, and leverage those stories to make business decisions that make work life and their customers’ lives better.

• Pearl-Plaza teamed up with Tethr 
• Pearl-Plaza acquired Lexalytics 
• And others

You might be thinking, so what? Should I be excited about this? Let’s look into it. 

What the Acquisition and Partnership Means in a Nutshell

In a nutshell, Lexalytics, and Tethr are data analytics platforms focusing on structured and unstructured customer data, as well as solicited and unsolicited feedback. With such acquisition/partnership, companies like Pearl-Plaza strengthen their capabilities in the “text analytics” space, meaning their ability to analyze unstructured data and extract meaning and actionable insights. But also in a broader way to be able to connect unstructured and structured data sources to generate insights from within one platform.

The Humble Beginnings of Surveys 

Before I jump into the deep end, let’s start at the beginning. Not that long ago, if we wanted to know what a customer thought, how they felt about interacting with your brand (website, store, call center, etc.), or how loyal they are to you, we had to ask them. We sent a survey and asked them what we wanted to know. In fact, almost every company sent surveys, to an extent that customers got rather fed up with it. We ran into the problem of survey fatigue, which plagues many of us. 

But it’s not just survey fatigue that challenges the trusted old survey, it’s also the accuracy of insights we gain from it. We sometimes ask questions the customer may or may not know the answer to—for example, did we resolve your issue today? The customer is likely thinking “hmm, well I hope so, the agent promised me to fix it..” We also ask questions we should know the answer to, like “did you travel with us in the last 30 days?”  And finally, we ask questions that seem irrelevant or unimportant to the customer, but we want to know more about it, like “did you remember seeing any advertisements on your flight today?” So, we kinda capture the “voice of the customer”, at least on things that are important to the company, and from those customers that can be bothered to respond. 

In addition to that, we tend to look at survey results in isolation, and then look at things like financial results, churn reporting, or customer complaints data, in isolation as well. Depending on the data maturity level in your business, you may combine some of your data, but not all of it. You may analyze some of your data, but not all of it—which we know is limiting, as data is best utilized when combined with various sources, rather than analyzed in isolation. 

So that’s why I’m excited about the recent announcements. It’s not that I oppose using surveys—absolutely not. They are a great tool in our toolbox, but they are only one tool, not THE tool. 

Extracting Meaning from Unstructured Data 

There’s one resource that has long been underutilized for mining data—the contact centre! The contact centre is an absolute treasure trove of customer insights and has long been underutilized from a customer insights perspective. It’s an amazing source of customer feedback. We have agents on the phone, email, live chat, and social media messaging. We have bots, call notes, and so much more. So instead of sending a survey, we can now analyze the data we already have, and potentially supplement what’s missing with a survey. 

Conversational analytics is also powerful as we are no longer limited to low numbers of survey responses, or hearing only from those customers that take the time to respond. Analyzing the conversation that just took place between your company and a customer means we have 100% of the conversation to use to generate insights from. It means more volume, but also a deeper understanding of your customers’ experiences, as we “hear” from all customers that interact with us.

With acquisitions and partnerships, companies like Pearl-Plaza strengthened their capabilities in this space, using ML (Machine Learning) and NLP (Natural Language Processing) to extract as much insight as possible from those unstructured data sources to tell us what the conversation was about, how the customer (and agent) felt about the interaction, and even predict what the experience was like (e.g. customer effort). Effort and ease, or CES (Customer Effort Score), is a super valuable metric to use in the interaction environment, as it tells us so much about how an experience went from a customer point of view, and is strongly correlated to customer loyalty. Based on unstructured data (the conversation that just took place between agent and customer) as well as operational data (e.g. call history, wait times, transfers, channel hopping) we can predict the level of effort the customer had to put forth to get their query resolved, all without a survey. 

Analyzing call or chat data helps us understand the conversation that took place, but also what it was all about. It allows you to narrow down on your customer “intent”, or reason for contacting. While we typically rely on agents to choose a “call reason” from a drop down menu, if you work in this space you probably know the accuracy levels of that data. That’s not just because an agent may opt to take a short cut and choose whatever option is right at the top of the drop down menu, it’s also limited to the options we provide, and one option only. Often calls may cover more than one reason, or the contact reason differs from the actual problem that needs to get addressed. Some telephone platforms now offer “intent recognition” and we can also get that information from our VoC platforms if we ingest that data. 

Beyond our contact centre data we can also leverage external sources such as social media or reviews. It’s another source of “free” customer feedback we can leverage to better understand our customers, their needs, and potential improvement areas. And again, we pull it into the same platform to have it in the same place as our other customer feedback data for enriched analytical capabilities. 

The Power and Limitations of Technology 

While those VoC platform announcements are super exciting, it’s not as simple as plugging them into our company tech environment and we have full access to all the shiny toys. You may end up with an (expensive) Ferrari in the garage, unable to drive it. The more data we can ingest into these VoC platforms, the better the quality of our customer and employee insights. However, which data we can share—from a policy, privacy, or tech point of view—determines to what extent we can leverage the tools. If you’re faced with a stack of legacy systems that don’t integrate easily, or can’t even connect the (data) dots between your systems, things become more challenging. 

Another incredibly exciting area is predicting experiences, or rather experience metrics. A word of caution here as well—we all know how unique and unpredictable we are as humans. A lot of testing is required before you have satisfactory accuracy levels for your particular organization (similar to intent work). So again, a great example of how we can leverage survey data to gain insights into customers’ perceptions of experiences. Expectations and perceptions make predicting experiences rather interesting. 

Wrapping Up

So to wrap up, from a conversational analytics point of view, we’re heading to a state where we know why the customer got in touch and what the interaction was about, what the experience was like from a customer point of view, how the customer felt (emotion and sentiment), and the impact the agent may have had. It’s pretty powerful to have that all in one place, but what do we do with this information? 

Firstly, we can enrich it even further with not “just” unstructured data from internal sources, but external sources like social media as well. We can also add key operational or financial data we have on the customer (e.g. call metrics such as handle time, customer tenure, value segments, churn risk, and others). 

Secondly, when we bring it all together we see a picture emerging on two levels, the operational level and the strategic level. 

1. On an operational level we may gain insights to help us train our agents or uncover root causes that we can tackle. Those are typically limited to a specific area, e.g. a call center team, and smaller in nature. 
2. On a strategic level we are able to uncover an end-to-end view of the customer experience, enabling us to look at company-wide experience improvement areas. Whether that’s overall, or broken down e.g. by specific journey stages. Again, effort is a great metric to use here as you can map out friction areas (aka areas for improvement) across journey stages by channel, or intent. You can also view this by e.g. product or specific services, overlay churn risk or value segments, the list is endless. It should give you a clear idea where to focus your improvement efforts and track performance over time. 

Many VoC tools can do parts of what I outlined here, but what we’re seeing now is a strong focus within our industry to mature our capabilities further, particularly in the conversational analytics space. It enables us to use the data we already have and use surveys only when we really need them. And that, in my humble opinion, is fantastic! 

Thanks for “listening”.