Table of Contents
- What is Voice of Customer Data (VOC)?
- Collecting VoC at All 7 Stages of Customer Experience
- What Information Comprises VOC Data?
- 5 Key AI-Powered Techniques Used To Unlock Value of VOC
- The Impact of AI on the 11 Key Components of VOC
- Big Data + AI for VOC Analysis
- Key Use-Cases for AI-Powered VoC
Voice of the Customer (VOC) is defined as the information you collect from your audience about what they like and don’t like, their behaviors and interests, their opinions, and their demographics. It’s the data that helps you do a better job of delivering the right messages, products and services to your customers.
VOC data powers two primary strategic imperatives at most companies: innovation and retention. With properly collected VOC data, companies create new, customer-centric innovations, answering unmet needs in the marketplace and seizing opportunities that competitors miss. VOC data also helps retain customers by solving problems specific to their needs and interactions with a company.
VoC helps to avoid guessing at the needs of your customers, and instead enables you to use data to understand who they are and what they want. While the best way to get into the heads of your customers is to collect the Voice of the Customer data, this can be a big challenge for companies to do, let alone do well â€“ so we’ll walk through how to utilize artificial intelligence to assist in this endeavor.
But to start, when you’re collecting VOC data, you want it to be comprehensive. It’s easy to collect little bits of information here and there, but it doesn’t tell the whole story. In order to get the full picture, you should be collecting information at every stage of the customer experience.
The customer experience (CX) is the process that someone goes through when they move from stranger to customer, to part of your company’s family. Within each stage of the CX, there are opportunities to collect VOC information. During the buyer’s journey, you can collect information like search intent, which ads have caught their attention, and what additional information they’re interested in. This information will help inform how you communicate about your product or services to external audiences. You can collect information about sentiment, usage, and case studies during the owner’s journey, which will help you in attracting similar customers.
VOC data comes from a variety of sources, including but not limited to:
- Face to face interviews
- Focus groups
- Market research surveys
- Observational data (watching the customer, shadowing the customer)
- Role play
- Customer advisory boards
- Environmental and landscape data collection
- Conferences and trade shows
- UI/UX data collection
One of the challenges of VOC data is that very little of it tends to be quantitative in nature. Much of it is unstructured text data, which can present challenges to companies seeking insights to extract. This is where data science, artificial intelligence (AI) and machine learning (ML) play a vital role in unlocking the value of VOC data.
1. Data Cleaning
Virtually all VOC data requires cleaning of some kind. This typically involves normalization of text data as well as imputation of missing quantitative data using machine learning techniques such as random forest imputation.
2. Exploratory Data Analysis (EDA)
VOC data â€“ and any customer data, for that matter â€“ needs substantial exploration to understand what’s in the data. EDA identifies anomalies, outliers, trends, distributions, statistical characteristics, and the overall shape and nature of the data. How much of it is there? How much of it is usable?
3. Text Mining/Natural Language Processing
VOC data is primarily in the form of unstructured text, such as interview transcripts, chat logs, transcribed audio, etc. Natural language processing techniques, from basic word frequency and ngram extraction to more sophisticated techniques like vectorization and transformers, extract value from text data, effectively transforming it into quantitative data that can be analyzed mathematically. From that analysis, we extract information such as sentiment, emotional valence, semantically related topics, and even inferences.
4. Forecasting and Predictive Analytics
VOC data by itself presents a historical snapshot of where the customer’s mind was at the point of data collection. Using the previously mentioned techniques, we examine the changes in data over time and incorporate other credible third-party data sources to forecast important VOC metrics, such as search intent for specific language or customer satisfaction numbers from survey data. Techniques used in forecasting range from very simple linear regression to ARIMA-based statistical models and deep learning/neural models at the highest levels of sophistication.
5. Regression Analysis
Some VOC metrics have predictive power, but understanding which metrics matter requires techniques such as advanced regression analysis. Technical tools such as linear regression, logistic regression, stepwise regression, Elasticnet regression, gradient boosting, gradient descent, multiple regression subset analysis, and neural networks make regression models more robust and informative for understanding what matters most in a dataset.
1. Advertising and Marketing
With ads, you’re spending money to catch someone’s attention at different stages of the customer experience. When you’re putting ads out to your customer base, you want to make sure that you’re providing the right messaging at the right time â€“ where the customer is in their journey with you. You can accomplish this with A/B testing. A/B testingOpens a new window is creating variations of the same thing and running experiments to see which version resonates. With manual A/B testing, you can manage a handful of iterations. You might test subject lines, images, copy, or audience segments.
A really great example of using AI to do A/B testing is Coca Cola. While most people execute one or two tests simultaneously, CokeOpens a new window figured out that they could use AI to run millions of permutations of ads across the world and knock out the underperforming campaigns until they had a set that would be used. They had to give the AI a training data set of keywords, images, and audience profiles, as well as thresholds of success or failure for each ad campaign. Once the planning was done, the AI took over and Coke monitored what was working and learned from the experiment how to reach their audiences globally.
For advertising, you’ll use regression analysis most.
2. Market Research
It’s tough to pull opinion data out of quantitative information, so why not just ask your customers what they like and don’t like? Market research allows you to dig into specific preferences, thoughts, and feelings that you would otherwise just infer from your static data. Market research can come in the form of 1-1 interviews, panel surveys, or general population surveys. This kind of research is generally the springboard for understanding what your specific audience voice will look like.
For market research, you’ll use regression analysis and exploratory data analysis most.
3. Social Conversations
People do NOT hold back their opinions on social media. You can find people from all walks of life that have thoughts about your company, whether you’ve asked or not. If you want to know exactly how people feel about your company, start a conversation on social media, sit back, and wait. You’ll get the good, the bad, and everything in between. There are a lot of social listening tools on the market already that can pull some of this information, but the limitation of the tools are the platforms that they integrate with. If your customers are on the major social channels but are also found on less conventional platforms you might be missing out on key information about what your customers want.
For social conversations, you’ll use text mining and natural language processing most.
4. Product/Service Reviews
Similar to social conversations, people are not shy about saying how they feel about you in online reviews, especially if they have had a negative experience. This data is a gold mine in terms of learning where to improve and what to keep doing. In addition to the conversation, you can extract the sentiment of reviews to create a model that will help you understand the elements of a five-star review, versus what garners only one star.
For reviews, you’ll use a combination of text mining and regression analysis most.
5. In-person/private messaging
A lot of people are more comfortable interacting digitally than directly with someone in-person or over the phone. That’s where chatbots become very helpful for your business. They can take on the role of interacting with your customers and collecting intake information that can be passed along to a service representative for further investigation and problem-solving. On the flip side, chatbots can help answer repetitive questions that your customers may have. To get this information, you would look at your other VOC components, like social listening, product reviews, and customer service calls, to understand what questions are asked repeatedly and where you can have AI step in.
Email is another way that customers feel more comfortable interacting with you. AI can help with standard responses and setting up nurture campaigns that will move your customers through a pre-defined experience. Both chatbots and email can be set up with predictive text and responses to ensure a consistent customer experience. You can then collect information about the responses and conversations to help inform your products and services.
The big thing to remember is to ALWAYS disclose your data collection policies â€“ so if you’re going to mine the content for information, your customers need to know this ahead of time so that they can decide what they share and what they don’t.
For messaging, you’ll use text mining and natural language processing most.
6. Search Intent Data
Search Intent data is a good place to start when you’re not sure what your audience looks like. Often used at the awareness stage of the CX, search intent data can inform your content marketing and advertising plans. Starting with keyword research, you can use AI to help with permutations of words to get a better understanding. Once you know what keywords you want to be known for, and what keywords your audience is using, you can create predictive calendars to serve up content when they want it. You can combine your data sources and other VOC components to really hone in on the language that your audience is using to talk about you, and what questions they have that you can answer in your content.
For search intent data, you’ll use a combination of predictive analytics/forecasting and text mining.
7. Sales Interactions
You can learn a lot about the behaviors of your customers by what they buy, and how often they purchase. You should have access to your e-commerce data, online cart information, and customer relationship management (CRM) software. Within your CRM you can chart out your sales funnel and see how quickly your customers take to convert. With your e-commerce and cart data you can see if people start the process and then abandon part way through. Understanding these behaviors can shape your customer experience and help expedite the experience for your audience.
For sales interactions, you’ll use a combination of text mining and regression analysis most.
8. Customer Service
People typically don’t call customer support because they are happy with what they have. Your customer service calls, emails, and chats are a treasure trove of information â€“ specifically what your customers don’t like and where you can make improvements. Accompanied by a data collection disclaimer, you can mine this data for keywords and topics that would otherwise get lost in the shuffle. If you’re recording your customer support calls, you can have AI transcribe the audio and then you can run text mining against the content to see what bubbles up. Using AI, you can also parse out the sentiment and tone of your interactions. Knowing the trends of positive or negative sentiment can help you train your service reps for all kinds of situations that may come up and allow them to provide even better support.
For customer service, you’ll use a combination of text mining and regression analysis, along with exploratory data analysis.
9. User/Owner Groups
User and Owner groups tend to be more focused on how people are using your products, which helps you to better understand your user experience. User groups can highlight when your audience creates â€œworkaroundsâ€ or â€œhacksâ€ when features or functionality don’t exist or work as expected. Getting this first-hand information from your customers about how they use your product (often in conjunction with other products) is invaluable and will help inform how you should innovate moving forward. A lot of companies fall into the trap of â€œI know what our customers wantâ€ and miss out on creating exactly what they really need.
For user groups, you’ll use a combination of text mining exploratory data analysis most.
10. Customer Metadata
A lot of companies rely on ONLY this information to determine the profile of their customers. While this information is important, it’s also too broad and tends to lead to assumptions. A great example of this is with My Little Pony. If the marketing executive had solely focused on the young girl ages 4-11 segment, they would have missed out on the â€œBronyâ€ segment, which is adult men who are interested in the lore and fantasy of MLP. This segment also has more disposable income and is willing to spend it with MLP and Hasbro.
Customer metadata is a great place to start your information gathering about your customers but should not be the only thing you collect. Systems like Google Analytics and your social media platforms will have the inferred demographic data around your website visitors and followers, such as age, gender, location, and interests.
For customer metadata, you’ll use exploratory data analysis most.
Many companies have the missed opportunity to ask people quick surveys such as, â€œdid you find what you wanted?â€ â€œdid we answer your questions?â€ or â€œwas the information helpful?â€ These surveys can be quick, can be one question, and can tell you a lot about your customers. Using AI, you can program these surveys to be on your website to be triggered when a customer takes a certain action, or to be sent to a customer’s email address at the conclusion of an interaction. This information adds an additional layer of information to inform your VOC to understand if you’re taking the right actions with your customers, and how they are feeling at the moment.
Customer feedback surveys also give you the data you need to calculate your net promoter score (NPS)Opens a new window . Companies rely on this number to gauge their customer’s loyalty, as well as demonstrate to other potential customers that they are worth doing business with. A company with a higher NPS is more likely to attract new business.
For surveys, you’ll use regression analysis most.
With all of these components of the VOC, how do you analyze all of the data? That’s the last component of the customer experience, Big DataOpens a new window . Big Data is essentially having large amounts of information that require assistance from AI to collect, clean, analyze, and make actionable. Each VOC component alone is a large analysis project. Tackling all of the components together will require a lot of planning and coordination. Once you’ve put together a system to collect data on all of these pieces, you’ll have a comprehensive view of your VOC, broken down into discrete segments, topics, and conversations. Using AI to assist in this venture will help expedite the process and ensure more accurate information. Having one piece of the puzzle is a good start, having all of the pieces to make the full picture is ideal and will give you a strong competitive advantage.
Once you’ve collected and analyzed your VOC data, the next logical step is to make use of it. The two primary use cases for VOC data are innovation and retention.
Based on the analysis you conducted in each part of the customer experience, what trends did you notice? What terms, words, ideas, and phrases kept showing up? Most of all, what did you notice that was unexpected? This is fertile ground for innovation, to identify things that customers desperately need but haven’t clearly articulated it. As a subject matter expert, you can infer what the actual need is based on your VOC data and produce the working solution. The business trope here is Henry Ford’s quip, â€œIf I asked customers what they wanted, they would have said faster horsesâ€. AI and machine learning allow us to identify those needs faster, more accurately, and at lower cost.
The second use case of VOC data is retention. In almost every business, it costs more to acquire a new customer than it does to retain an existing customer. With outputs like sentiment analysis, emotion and tone detection, and other techniques, you can identify what customers are most unhappy about and fix it before you lose them. In turn, you’ll stabilize your revenue and reduce your overall costs for sales and marketing by increasing evangelism. Customers who feel their needs are being met and their expectations exceeded will do some of your marketing for you.
Like we said, VoC helps to avoid guessing at the needs of your customers, and instead enables you to use data to understand who they are and what they really want. We hope this feature helps you figure out the best ways to collect data about the Voice of your Customers, as well as gain insight on how you could leverage artificial intelligence to help drive optimal value from this valuable data!