Voice of the customer (VoC) is the practice of capturing what customers say they need, like and dislike, across surveys, reviews and support conversations, and feeding it into decisions. It puts what customers actually say ahead of internal assumptions. In customer service, the richest VoC source is the conversations your team already handles every day.
In short
- VoC stands for voice of the customer: the feedback customers give, asked for or not.
- Sources include surveys, reviews, social posts, interviews and support conversations.
- Support conversations are the largest VoC source, and the least used.
- A VoC program collects, analyzes, routes and acts, then checks that the fix worked.
- VoC pays off only when insight reaches the teams who can act on it.
What is voice of the customer?
Voice of the customer is everything customers tell you about their needs and experience, collected and analyzed so the business can act on it. The term covers both the feedback itself and the program that turns it into decisions.
VoC programs pull from two kinds of feedback. Solicited feedback is what you ask for: a CSAT survey, an NPS question, a customer interview. Unsolicited feedback is what customers volunteer without prompting, in a support ticket, a review or a social post. Most programs lean heavily on the first and neglect the second, even though unsolicited feedback is larger and more candid.
The output of VoC is not a score but a set of themes: what customers struggle with, what they ask for, and how they feel about it. In support, the fastest way to see those themes is sentiment analysis and topic tagging across every conversation, so the picture reflects everyone who reached out rather than the few who answered a survey.
What does VoC mean in customer service, call centers and BPOs?
In customer service and call centers, VoC means the feedback customers give during and after a contact, used to improve service, products and processes. In BPOs, VoC is also the feedback the outsourcer collects and reports back to its client.
In a support team, VoC usually covers three things: post-contact surveys, the comments customers leave with them, and what customers say in the conversations themselves. A team lead reads it to coach agents, and a product or operations owner reads it to fix what causes the contacts.
In a BPO or outsourced contact center, the term has a second, everyday meaning. Many BPO teams use “VoC” for the post-call or post-chat survey and its score, reported per agent and per client program.
The outsourcer also owes its client a VoC report: what the client’s customers are complaining about, which the client cannot see without it. When that report stops at a survey score, the client misses why customers called, which is why BPOs rarely tell clients what is broken in the product.
Where voice of the customer data comes from
Voice of the customer data comes from surveys, reviews, social media, interviews and support conversations, and each source has a different bias.
| Source | Type | Strength | Weakness |
|---|---|---|---|
| Surveys (CSAT, NPS, CES) | Solicited | Structured and easy to trend | Low response rates, skewed to the happiest and angriest |
| Reviews and social | Unsolicited | Candid and public | Skewed toward extremes, hard to tie to one customer |
| Support conversations | Unsolicited | High volume, honest, in the customer’s own words | Unstructured and often ignored |
| Interviews and user tests | Solicited | Deep context on why | Small sample and time-intensive |
| Product usage data | Behavioral | Shows what customers do, not what they say | Explains little about why |
The underused source: conversation data
Support conversations are the largest and most honest VoC source most companies have, and the least analyzed, because they are unstructured and hard to read at scale.
Surveys capture a slice of customers who choose to respond, usually the happiest or angriest. The far larger signal sits in the conversations customers already have with support every day, where they describe problems in their own words without being prompted.
Analyzing those conversations turns thousands of tickets and calls into recurring themes: the same feedback a survey would surface, but drawn from everyone who reached out rather than the few who filled in a form. Text analytics does this for written tickets, chats and emails, and speech analytics does it for calls once they are transcribed.
How to run a voice of the customer program
A voice of the customer program runs in five steps: set your questions, collect from several sources, analyze themes, route each theme to an owner, and measure whether the fix worked.
Step 1: Decide what you want to learn
Start with two or three questions the business needs answered, such as “why do customers cancel in the first month?” or “which part of onboarding creates the most tickets?”. A program without questions collects data nobody reads.
Step 2: Collect from more than one source
Combine at least one solicited source (a short survey after contact) with one unsolicited source (your support conversations). The survey gives you a trend line; the conversations tell you what sits behind it.
Step 3: Analyze feedback into themes
Tag each piece of feedback with a topic, a sentiment and, where it fits, a contact reason. Done by hand, this only works for a small sample. Done automatically, it covers every conversation, and you can rank themes by how often they occur and how negative they are.
Step 4: Route each theme to an owner
An insight without an owner stays a chart. Product bugs go to the product team with example conversations attached, missing answers go to the knowledge base owner, and agent behavior goes to coaching.
Step 5: Close the loop and measure
Tell customers what changed where you can, and check the same theme a month later. If contacts about that topic dropped and sentiment improved, the fix worked. If not, the theme goes back to step 3.
Voice of the customer examples in support
Voice of the customer examples in support are recurring themes in conversations that point to a specific fix, such as a billing question that keeps coming back or a feature customers keep asking for.
| What customers say | What VoC analysis shows | Who acts | Action |
|---|---|---|---|
| “Where is my refund?” in many tickets after a policy change | One contact reason doubled in a month, with negative sentiment | Operations | Rewrite the refund email and the help article |
| “How do I export my data?” | A frequent how-to question with a stable answer | Knowledge base owner | Publish and link a step-by-step article |
| “I already told the last agent this” | Repeat contacts on one issue type | Support lead | Fix the handover and coach on full resolution |
| “It would be great if it worked with our calendar” | A recurring feature request from larger accounts | Product | Add it to the roadmap review with conversation examples |
Each row starts with words customers used, not a score. That is what makes VoC useful: it tells you what to change, not only that something changed.
Voice of the customer metrics to track
The core voice of the customer metrics are survey scores (CSAT, NPS, CES) plus conversation measures: sentiment, top contact reasons and repeat contacts on the same issue.
| Metric | What it tells you | Limit |
|---|---|---|
| CSAT | Satisfaction with a recent interaction | Only covers customers who answer |
| NPS | Willingness to recommend, a loyalty signal | Says little about why |
| Customer effort score (CES) | How hard it was to get help | One question, one moment |
| Sentiment across conversations | How customers felt, including those who never answer a survey | Needs context to read polite frustration |
| Top contact reasons | Why customers get in touch at all | Only as good as the tagging |
| Repeat contacts | Where the first answer did not solve it | Needs a clear definition of “same issue” |
How QA and conversation intelligence surface VoC
Quality assurance already reads support conversations, so the same analysis that scores agents can surface what customers struggle with and ask for, turning QA into a continuous VoC feed.
Making conversations usable as VoC takes a system that reads them at scale, and this is where QA and conversation intelligence contribute. In 2024 research, Gartner found that among service leaders who rate their QA program as highly valuable, 52% say the value comes mainly from voice of the customer and customer experience insights, and fewer than 20% point to rep performance.
Kaizo is a QA and coaching platform that works with Zendesk and Salesforce. It scores customer sentiment and agent empathy across 100% of messages and shows whether a dip comes from a broken process, product friction or a skill gap, with the conversations behind each one. The conversations a team reviews for quality then become a continuous, unsolicited voice of the customer feed rather than a separate survey program. Our customer experience software guide shows where that layer sits.
Related terms
- Conversation intelligence
- What sentiment analysis is
- Customer sentiment analysis for support teams
- Customer service insights
- Conversation analytics
- Interaction analytics
Hear the voice of the customer in your conversations
Pick one question from step 1 and look for the answer in last month’s support conversations, not in a survey. In a live demo we will show you, on sample data, how Kaizo surfaces the themes, requests and frustrations customers are already telling you.
Let's talk Explore Kaizo Insights
Frequently asked questions
What does VoC stand for?
VoC stands for voice of the customer. It refers both to the feedback customers give about their needs and experience and to the program a company runs to collect, analyze and act on that feedback. In call centers and BPOs, VoC is also used as shorthand for the post-contact survey and its score.
What is the difference between VoC and CSAT?
CSAT is a single survey metric that measures satisfaction after an interaction. Voice of the customer is the broader discipline of capturing and analyzing feedback across many sources, including CSAT, reviews and conversations, to understand what customers need. CSAT is one input to a VoC program, not a replacement for it.
Why are support conversations a good VoC source?
They are high volume, they happen in the customer’s own words, and they are unsolicited, so they are not skewed by who chooses to answer a survey. The challenge is that they are unstructured, which is why analyzing them at scale with topic tagging and sentiment matters more than reading a sample.
How do you analyze conversations for voice of the customer?
You gather conversations across channels, transcribe any calls, then use analysis to tag topics, sentiment and outcomes. This turns unstructured tickets and calls into recurring themes you can trend, the same way you would a survey, but across everyone who contacted you rather than the few who responded.
Who uses voice of the customer insights?
VoC insights feed product teams deciding what to build, support operations spotting rising issues, and leadership tracking experience over time. A program works best when those insights reach the people who can act on them quickly, with example conversations attached so the evidence is hard to dismiss.






