Can Speech Analytics Really Increase Customer Satisfaction?

Speech analytics ads are saturated everywhere and the headlines are starting to blend together.  “Uncover actionable insights”, “amplify the voice of the customer”, “gather sentiment with AI”, and so on. But do speech analytics actually deliver on the multitude of promises they’re sold with? The short answer is yes, but the caveat to that yes sounds like something a life coach tells you: you’ll get out of speech analytics whatever you put into them. 

Making Speech Analytics Deliver on its Promises

Let’s focus on a promise we’ve heard repeatedly. You know, the one about uncovering actionable insights. If actionable insights is an apple pie, speech analytics are the oven, not the ingredients, and the problem with this lofty claim about insights is that businesses don’t use good pie ingredients, and even when they do, the oven is complicated to operate. But we’re getting ahead of ourselves. Let’s go over the ingredients, then we’ll bake. 

Where is your data coming from?

Analytics filter and organize data you collect from your customers. So traditionally this data could be surveys, metadata, social media posts, the list goes on. Speech Analytics require audio input. Seems simple enough, but so much speech technology is simply used to make a transcript or check for compliance on a call. 

Call recordings contain an immense amount of customer data that's waiting to be tapped but is often ignored. This is because so many companies simply record calls as a matter of compliance. The only time they listen to any of those recordings is to spot check quality assurance. Meanwhile, lofty promises about speech analytics go unrealized. The first step to obtaining some of these benefits is to process those calls with a fine-toothed comb that produces keywords. 

What keywords are you hunting for?

This is where the life coach tells you you have to put some time in if you want to see any real benefits from the process. Every business has its own language. there's the language the customer uses, and the language your employees use. Say you run a retail operation and a customer calls in saying “I want to return something I bought”, and your phone agent says “I’ll need to process a return merchandise authorization”. 

Now let's say you run a pretty big operation, and phone calls like that happen 100 times a day. Your employee might consistently use the same language because you train them to do so, but your customers have many ways of saying if you want to return a product. Some are eloquent, others, not so much. 

Your speech analytics application needs to be trained through a process called machine learning. Whenever we throw around terms like artificial intelligence, we’re most commonly referring to machine learning. For speech analytics, you'll want to load up your word bank with as accurate a custom vocabulary as you can, because artificial intelligence isn't really intelligent, you are. You know your business, you know what your agents are dealing with day in and day out, and you know what the customers are calling in to speak about. An AI doesn't know what you know, but it can very capably spot what you tell it to look for. 

This opens up another can of worms, namely, how much data do I want to extract from my call recordings? The possibilities are fairly endless, but will cover those next.

Speech Analytics Topics

Topics represent the insights you seek from speech analytics. Consider our fictional call center. customers call up to order new products, to return broken ones, to seek technical support, to complain about a bill, to find pre-sales information, and so on.  Every reason your customer calls about has a list of recurring words and phrases associated with that reason. These recurring keywords identify the topics each call touches on. 

To demonstrate, say our first topic is “Tech Support”. For this you would want to identify phrases like it's not working, I think it's broken, and it won't turn on, when they occur in customer calls. Let’s also define the topic “Billing Issues”. You configure your speech analytics engine to hunt for words and phrases like extra charge, already paid, I don’t owe, etc. Modern speech analytics are very flexible and don't need an exact phrase for successful detection.

Topics humanize the collected data so that we can assimilate the insights in a way that’s natural to us. This is an evolutionary step forward for speech analytics because it goes beyond measuring the number of times a recurring phrase happens and grants users the power to design highly accurate custom analytics models. 

There’s bound to be overlap with calls that touch on multiple topics, just the way a human conversation about one topic branches out to others. Just as you could tell a call center manager to notify you if there’s an uptick in sales or product returns, you can have your analytics platform alert you when certain topics trend up or down. 

Speech Analytics Vs. Customer Surveys

Surveys are a time-tested means of extracting insights from customers, but they have several flaws that make them hard to trust by design. First, a survey must be answered, and no matter what incentives you throw at your customers to take them, there are large numbers of people who simply have no desire to take a survey. Others will speed through survey questions just to retrieve whatever incentive you promised them. Consider also that when you survey a customer, even on who wants to share their opinion, they are mostly limited to your multiple choice answers that may not let you hear what the customer is actually trying to tell you. 

But perhaps above all, surveys fail to show us the customer’s sentiment simply because you are the one asking for an opinion, after a sale, return, or tech support session. The customer is likely in a different frame of mind, distracted by other things going on, less engaged with your company than they were when they called you. 

But what if instead of asking them for their opinion after the call you simply converted the call into the insights you were looking for? Think of it, you have a view of your customer’s sentiment in its purest state, when they’ve bothered to pick up the phone and reach out to you. They’ve looked up your number, navigated your IVR system, pressed multiple numbers to get routed properly, and listened to your on-hold music, all just to talk to one of your live agents. 

This customer is currently focused on you and your products, and will tell you in no uncertain terms what they’re feeling, what they want, what they do and don’t like, and even how they feel about the agent sometimes. 

“Ok, great”, you say, “but how do I use analytics to turn that interaction into the data I was getting from my surveys?” The answer is with topics, The keywords you sweep for should reverse engineer your survey questions. For example, see how some basic customer survey questions are found in calls with the image below. We’ll divide the topics we’re trying to gather data for as Ease of Use, Features, and Contract.

A few minutes thinning about how data is shared from customer to agent over a call will allow your to design keyword-powered topics that reveal what’s driving your call center. 

Analytics Vs. Surveys, Can’t I Do Both?

Without a doubt, more data is usually a good thing, but when it comes to Speech Analytics Vs. Customer Surveys, the answer to can’t I do both is not a simple yes. It’s more an issue of should you do both? Let’s consider the ROI for both methods of collecting customer data.

Analytics Vs. Surveys — Specificity

Surveys let you directly ask customers a specific question, but multiple choice answers limit your ability to fully understand what a customer really wants to say. Speech analytics let you sweep broadly for topics and drill down your results while not forcing opinion to fit within a multiple choice answer. 

Analytics Vs. Surveys — Take 2, 3, 4 and so on

When you reach a customer with a survey, they answer it and the transaction is over. Any opinions the customer had that didn’t get represented in your multiple choice answers are lost. This is handled by having custom fields in your survey that allow customers to explain their answers, but it adds work to the survey assessment process. Contact center speech analytics allow you to constantly refine your assessment of your customers by changing your keyword sweeps. For example, a week’s worth of calls can have multiple topic results extracted from it. You can refine your keyword sweeps to concentrate on customer service, billing issues, compliance checks, upgrade requests, competitor chat, and so on. When you’ve exhausted that week’s calls, you can continue to extract topical data from the next week’s calls. 

To put it simply, you already record your calls, so opting for surveys instead of reaping data from the recording makes no sense regarding time or money invested. Speech analytics yield significantly more actionable and relevant data on an evolving basis, whereas surveys capture a customer snapshot with very limited data. 

Analytics Vs. Surveys — Customer Sentiment

Surveys reach a customer after a transaction is complete. Even if the customer takes the survey right after the transaction, they’re in a different state than they were when the transaction began. Speech analytics capture data when the customer’s emotional state is most valid, with the full sentiment of their language intact. More succinctly, call interactions happen when the customer has called you, and that will be reflected in the language they use. This will be when the customer’s honesty is at its height, and the language they use will come from them, not from your survey.

Surveys simply don’t provide the bang for the buck that speech analytics provide. Do both if you want, but consider which one takes priority and which one is yielding the most ROI.

Yes, Can Speech Analytics Really Increase Customer Satisfaction

As the expression goes, knowledge is power. When you know how a person is feeling, it drives the decisions you make around them. When you accurately measure your customer sentiment, you’ll make the most informed choices about serving them. You’ll know where to concentrate your product development, your marketing, your pricing, and your position against competitors. And with speech analytics, you’ll uncover richer data than you’ve had before. 

Learn more about how speech analytics help improve customer satisfaction by downloading our new MiaRec White Paper.