About two weeks ago, I was having a conversation with a candidate for a sales role. We talked for almost an hour about evolving technology in improving customer experience measurement. Toward the end, a question came up that I’ve heard several times in the last year or two. He asked, ‘Do we need surveys anymore? Can’t we just replace them with speech and text analytics on interactions?’ It stopped me for a minute as I haven’t always explained the differences clearly in the past. But this time, I answered ‘Great question. To help answer that, can you tell me what I am thinking as we sit here and talk?’
He said ‘No.’
And that’s where unstructured data analytics are today as a tool for measuring customer experiences. The technology can’t understand the unstated; mind-reading is still an aspiration. There’s so much valuable information from conversations that we can capture and analyze to infer good and bad interactions: sentiment, content, speech patterns, etc. But technology can’t measure what’s not said or written – and that’s where crucial information lies. To elicit what customers find important and expect, we need to ask questions in surveys.
Wait a minute … can’t you just use machine learning to predict customer satisfaction or NPS from your customer interaction transcripts? The answer is yes. We’ve been active in this area of research for the past five years, and in a representative engagement, we matched customer surveys to earlier interactions and deployed speech analytics in tandem with machine learning to develop satisfaction models. In what has been a repeating pattern, we see that the group predicted most likely to be dissatisfied (top 10% of customers) were over five times more dissatisfied than the general population.
But although you can predict survey scores broadly, the specific scores on the survey itself remain harder to predict. In the same example, we would have assigned the wrong score to a given interaction three times out of ten, even using a strong model. And for the predictions we got right, we still wouldn’t know what customers thought of the interaction.
It’s not an either/or choice. You need unstructured data analytics and you need surveys to truly understand customer experiences. For the rest of this post, I’ll focus on how to deploy them together successfully.
What should your unstructured data analytics program do?
You absolutely should be deploying speech and text analytics across your voice, chat, and messaging interactions. The power of unstructured data analysis is that it truly can be expanded to cover every interaction. Surveys give only a sampling of your environment. Speech and text analytics coupled with 100% recording ensures you know what happened in all of your interactions. So – what information should you be gathering, where should you be finding it, and what should you be doing with it?
Compliance – Ensure that every interaction aligns to the way that your organization wants to treat its customers. Make sure your interactions are designed to deliver on your brand promise.
Experience scoring – By integrating your survey program results into predictive models against your interactions, you gain the ability to score each interaction and identify when to take action. Being able to diagnose every interaction helps ensure you’ll take action where and when it’s needed to improve the experience. Carefully aligning potential outcome actions to high-certainty interactions helps mitigate false positives.
Root cause analytics – Turning service interactions into hard data brings to light the product and service factors affecting customers, and helps us understand what in-service behaviors create differentiated customer experiences.
Coaching and performance management – Recorded and scored coverage of every customer interaction a front-line employee has is a powerful tool for coaching. With access to all interactions, you’ll know that what you’re coaching is broadly applicable, and not an isolated occurrence.
Where should your survey measurements focus?
Analyzing conversations can yield an immense amount of data for analysis. Your survey programs can focus data collection on the information unstructured analytics can’t give you: insight into the ‘why’ of customers’ behavior and emotions.
What do your customers really think and feel about interactions? While it’s true that unstructured data analytics uncover what is said, social biases often prevent individuals from sharing their true emotions and evaluations during interactions with strangers. At least, not until they get really upset …
What are their expectations? Surveys do a good job capturing what a customer’s expectations for that interaction were. Do we know what she or he was looking for? Or what criteria had to be met in order for it to be considered a positive interaction? Often, expectations aren’t spoken; they’re implicit.
What do they want from you in the future? Although it’s true that product and service enhancement ideas can come from listening to conversations, customer-focused companies know the importance of asking about what they don’t know.
- What are their journeys like? What happens in one particular interaction is often just one touchpoint along the way in a longer customer journey. By asking questions about the end-to-end experience associated with a particular journey, surveys can help us move past the ‘moment in time’ of a particular service experience.
Analyzing data from conversations is mission-critical for organizations to understand what works and what doesn’t work in the interactions they have with their customers. But by focusing only on what has happened and the record of those conversations, organizations run the risk of being too inwardly focused on analyzing the data and prioritizing process improvements. If you don’t explicitly ask your customers what you don’t know and how they feel, you’ll miss out on what they truly think of their relationship with you. Until our AI technologies can read minds, supplementing conversational analytics with surveys is the clearest window into our customers’ hearts and minds.
Image source: Getty Images