First-party data is your most important advantage in today’s highly regulated marketing environment, and customer conversations have become the ultimate first-party data source. But extracting value from first-party data hasn’t been easy. In this webinar, you’ll learn how artificial intelligence (AI) can deliver smarter customer insights, resulting in more effective search, social, and display ad campaigns.
In this on-demand webinar, our data analytics experts will show you how to use conversational AI to analyze customer conversations and scale those insights to boost conversion rates, optimize buying experiences and drive more revenue-generating calls.
You’ve probably heard this before: the first step in learning about your customers is to stop making your marketing about you and make it about them. But to know what they want, you have to listen to what they say.
The traditional approach to figuring out what your customers want is to look at your website analytics. There are a lot of specific metrics you can look at to see how people are acting. The problem with that is you only know what they are doing. You put it into a bunch of colorful and fascinating reports and visualizations and call it a day.
You know what brought them to the website, whether or not they had trouble navigating it, the content that they consumed, how long they were there and all of that fun stuff. But you are still missing a big part of the picture — why are your customers doing what they do? The why is something that is difficult to distill from quantitative information and when you start to look at the qualitative, you end up with a whole different set of questions.
To figure out how people feel, you have to listen to them, which can be tough to do looking at click-through and conversion rates. Were they aware that we have new products or new features? Did they understand the value proposition? Do they like what they hear? Why did they come back and engage with the website or with the email or on social media and did it change their opinion about how they feel? Now, did they buy it? But the biggest qualitative is did they like the experience? This is a tough question to answer, especially using the traditional methods of measuring customer experience.
Marketers have created myriad ways to understand how people feel. From the phone surveys of yesteryear to Survey Monkey to usability labs and Net Promoter Scores (NPS), we’ve tried all kinds of crafty ways to get to the bottom of how customers feel about companies. The problem is, when you ask people how they feel about you, they’re probably not going to be completely honest.
The nature of human beings is that they will tell you what they think you want to hear. They’ll tell you whatever it takes to get access to whatever you’re offering. So, yes, my name is firstname.lastname@example.org. Now can I please have the whitepaper? Or, conversely, they’ll tell you what they think of themselves rather than what they’re actually doing. Which is equally useless.
So, how do you get around this problem and find out how people feel when you ask them straightforward questions and they lie to you? You have to listen to what they say. One of the best ways of doing this is listening to what they say in your call center.
In the call center, you experience the entire range of unfiltered customer emotions. They’re asking real questions because they don’t understand your product. They are telling you all about the problems they are having, and they are not going to hold back because they want to get it solved and they don’t want to feel like they are wasting their money and time.
This is where you can start to establish some patterns. When the first 20 people call with a problem, say, navigating your website, you might think “woah, that’s weird.” Then you see 50, 500, 5,000 people are having the exact same issue and you know that you have a problem on your hands.
The best marketers will pay attention to this. They’ll spend time listening to calls at the call center, until they figure out that there is no way to make that scale. That is, until machine learning came around.
Before you non-data science types are lulled off to sleep by the thought of another egghead explanation of machine learning, don’t fret. This will be in plain English.
Generally, you can think of supervised machine learning (ML) as something you use when you know the answers and you want to teach a machine to recognize them. The classic example is teaching a machine to recognize a cat. You show it 100,000 pictures of cats. Then you show it a picture of a skunk and it has to decide if it’s a cat or not. It may have some characteristics of a cat, but if your data is properly labeled, it will say “no, this is not a cat.” Viola, you have trained a supervised AI model.
Unsupervised machine learning is when you don’t know what the question is and you want to find out something that you don’t know. You hand over all of your data and the goal is for the ML to find relationships. Finding these relationships, however, can be correlations rather than causal. This can be useful—or not.
The ML may find that sales on the website go up and retail sales go down when the weather is bad. Well, true, but I knew that already. So, I didn’t learn anything, but thank you, machine.
On the other hand, it may find that death by drowning goes up when sales of ice cream go up. But this is obviously not causative. It just means that it’s summertime, more people are swimming, and since it’s hot, they’re also buying ice cream, but that’s not what’s causing the accidents. This is an opportunity for machines to discover things, but it’s up to humans to figure out whether or not they matter.
Reinforcement machine learning is for when there is no definitive “yes or no” answer, but some behaviors are considered better than others. For example, say you have 100 different email subject lines to test. If you send out 1,000 emails, you can see which ones are working and which ones are not. You get rid of the ones that don’t work, then you find some more that are like the ones that work and you send out another 1,000, and again and again and again and over time, it gets more opens. So, great. The machine through reinforcement says, “Oh, this works better, this works better, this works better,” and it learns what works.
So, now let’s talk about how this relates to the call center and marketing.
Invoca performed a survey where we asked marketers who are using call tracking solutions about their use and view of first-party data. We found that 95% of them think that it’s important for their customers to be able to connect with their business by phone during that buying process, and 93% see calls as having a big impact on their organization’s bottom line. That’s not a huge surprise, but what that says to us is that when you’re thinking about your top sources of first-party data, conversations need to be right up there with data from your website and purchase history.
Because if 50% of your revenue is coming by inbound phone calls, then the data from your contact center or your call center is just as important as the data from your website. So, if you’re blind to the insights that are happening within the contact center and you cannot connect them to your digital journey, then you’re blind to 50% of your revenue.
The same type of data that you expect from your digital journey, you also need for the offline experience. That’s exactly where a call tracking and analytics solution comes in.
There’s a recognition amongst marketers that conversations with customers are crucially important. There’s a recognition that they need to understand, they need to hear directly from customers, and they’re doing whatever they can to get at that data. But that can involve a very manual process of listening or combing through transcripts. It’s not easy, but it’s worth it to their business.
But what that says is that these marketers are actually missing out on that first-party data because they can’t possibly listen to tens of thousands of calls. They also can’t connect that to all the digital identifiers out there and use it to optimize their marketing. So, they take a kind of spot check approach instead, leaving a lot of potential revenue tied up in the call center.