Voice of Customer: What You’re Measuring Isn’t What Customers Mean

A voice of customer framework is a structured process for capturing what customers actually think, say, and feel about your product or service, then translating those signals into decisions. Done properly, it tells you not just what customers say they want, but why they behave the way they do and where the gap between their expectations and their experience is widest.

Most companies have some version of this. NPS surveys, post-purchase emails, the occasional focus group. What most companies don’t have is a system that connects those signals to anything that changes how the business operates.

I’ve run agencies and managed client relationships across thirty industries. In almost every case, the businesses that struggled with marketing were the ones that had lost touch with what their customers actually thought of them. Not in a philosophical sense. In a very practical one. They were optimising ads and refining landing pages while customers were quietly churning for reasons that never showed up in the campaign data.

Why Most VoC Programmes Produce Data Nobody Acts On

The failure mode is almost always the same. A company sets up a survey tool, collects responses, publishes a monthly report, and files it somewhere that nobody reads. The data exists. The insight doesn’t.

I spent time working with a retail client who had been running customer satisfaction surveys for three years. They had thousands of data points. Their NPS score had moved two points in either direction across that entire period. When I asked what had changed in the business as a result of the programme, the answer was essentially nothing. The surveys had become a compliance exercise, something to show the board to demonstrate that the company was “listening to customers.”

The problem wasn’t the tool. It was the absence of any mechanism for turning a response into a decision. Nobody owned the output. Nobody was accountable for acting on it. The data sat in a report and the report sat in a folder.

A functioning VoC framework has three components that most programmes skip: a collection architecture that captures signal across multiple touchpoints, a routing system that sends the right insight to the right person, and a feedback loop that closes the circle by confirming what changed as a result. Without all three, you have a measurement programme, not a learning programme.

If you want broader context on how VoC fits into a full research and intelligence function, the Market Research and Competitive Intel hub covers the surrounding disciplines in detail.

What Signals Are Worth Collecting and Where to Find Them

The instinct is to go straight to surveys. They’re clean, quantifiable, and easy to report. They’re also the weakest signal in most VoC programmes, because they capture what customers are willing to say in a structured format rather than what they actually think and feel.

The richest signals are usually already inside the business. Sales call recordings. Support ticket transcripts. Churn interviews. Online reviews. Social comments. These are places where customers use their own language, unprompted, to describe their problems and their experience. That language is worth more than a satisfaction score, because it tells you how customers frame their situation, and that framing is what your messaging needs to match.

When I was building out the research capability at an agency I ran, we started pulling language directly from client support tickets and mapping it against the copy on their product pages. The gap was striking. Customers were describing their problems in one set of words. The marketing was using an entirely different vocabulary. The product hadn’t changed. The messaging hadn’t changed. But once we closed that language gap, conversion rates moved in ways that no amount of ad spend optimisation had managed.

The signal types worth building into a VoC framework fall into roughly four categories:

• Solicited structured feedback: surveys, NPS, CSAT, post-purchase ratings. Useful for tracking trends over time. Weak on nuance.
• Solicited unstructured feedback: open-text survey responses, customer interviews, focus groups. Higher signal quality. More expensive to process at scale.
• Unsolicited structured feedback: review platforms, app store ratings, third-party comparison sites. Customers saying what they think without being prompted. Highly credible.
• Unsolicited unstructured feedback: social mentions, community forums, support conversations, sales call recordings. The closest thing to unfiltered customer thinking you’ll find.

A mature VoC programme draws from all four. An early-stage programme should prioritise the unsolicited categories first, because that’s where the honest signal lives.

How to Structure a VoC Framework That Actually Routes to Decisions

The architecture question is where most programmes fall apart. Companies collect feedback but have no system for deciding what happens next. A useful framework has five stages, and the last two are the ones most teams never build.

Stage 1: Define the questions you’re trying to answer. Before collecting anything, be specific about what decisions the VoC programme is meant to inform. Are you trying to understand why customers churn? Why conversion rates are lower than expected? What objections are blocking purchase? Vague listening produces vague insight. The collection architecture should follow the question, not precede it.

Stage 2: Map the collection touchpoints. Identify where in the customer experience you have natural opportunities to capture signal. Post-purchase, post-support, post-onboarding, at renewal, at cancellation. Each touchpoint captures a different kind of signal. A cancellation interview tells you something a post-purchase survey never will.

Stage 3: Process and categorise the signal. Raw feedback is not insight. Someone needs to read it, tag it, and identify patterns. This is where qualitative and quantitative signals need to be handled differently. A recurring theme in open-text responses is worth more attention than a single dramatic complaint. Volume and frequency matter.

Stage 4: Route the insight to the right owner. This is the stage most programmes skip. A product insight needs to reach the product team. A messaging insight needs to reach the marketing team. A service insight needs to reach operations. If all insights go into the same report and the same inbox, nothing moves. Build a routing protocol that maps insight categories to named owners with defined response timelines.

Stage 5: Close the loop. Track what changed as a result of each insight. This serves two purposes. It creates accountability for action, and it gives you a way to measure whether the change improved the thing you were trying to improve. Without this stage, the programme has no way to demonstrate its own value, and it will eventually be deprioritised.

The Customer Interview: Still the Highest-Signal Method Available

I’ve never found a substitute for talking to customers directly. Not surveys, not analytics, not social listening. A well-structured customer interview, even a short one, surfaces things that no passive data collection method will ever find.

The reason is simple. Customers don’t think in categories. They think in stories. When you ask someone to describe their experience of buying a product or solving a problem, they tell you things that no survey question would have thought to ask. They reveal the context, the alternatives they considered, the moment they decided, the thing that almost made them leave. That narrative is the raw material of good positioning and good messaging.

The discipline is in the interview structure. A few principles that hold up across every context I’ve used them in:

• Ask about behaviour, not opinion. “What did you do next?” is more useful than “What do you think of our onboarding?”
• Probe for context, not evaluation. “What were you trying to solve when you first started looking?” is more useful than “How satisfied are you with the product?”
• Let silence work. Most interviewers fill pauses. The best signal often comes after a pause, when the customer adds something they hadn’t planned to say.
• Record and transcribe everything. The language customers use is the output, not just the sentiment.

Ten well-structured customer interviews will usually surface the same three or four themes. That’s enough to act on. You don’t need statistical significance to know that four out of ten customers mentioned the same friction point unprompted. That’s a signal worth routing.

Where NPS Goes Wrong and What to Use Instead

NPS is the most widely used customer metric in marketing. It’s also one of the most misunderstood. The score itself is a blunt instrument. A number between minus one hundred and plus one hundred tells you almost nothing about what to do differently. The value, to the extent there is any, is in the open-text responses that follow the score question, and most companies barely look at those.

The other problem with NPS is that it’s a lagging indicator. By the time a customer gives you a low score, the experience that caused it is already in the past. You’re measuring the damage, not preventing it.

I judged the Effie Awards for several years, which meant reviewing hundreds of marketing effectiveness cases. The campaigns that demonstrated genuine customer understanding, the ones that showed evidence of real listening rather than assumed insight, consistently outperformed the ones built on category assumptions and demographic generalisations. The score wasn’t the differentiator. The depth of understanding was.

A more useful approach combines a lightweight satisfaction measure with a specific follow-up question tied to the stage of the experience. At onboarding: “What’s one thing that would have made getting started easier?” At renewal: “What’s the main reason you decided to stay?” At cancellation: “What would have needed to be different for you to continue?” Each question is designed to produce actionable language, not a score.

Managing risk in how you interpret and act on customer data is worth thinking about carefully. BCG’s work on risk management offers a useful frame for thinking about how organisations weigh evidence before committing to a course of action, a discipline that applies as much to customer insight as to financial decisions.

How VoC Connects to Messaging and Positioning

The commercial value of a VoC programme isn’t in the insight reports. It’s in what happens to the messaging when you actually use the language customers give you.

Most marketing copy is written from the inside out. The company decides what it wants to say about itself, then finds ways to say it. Customer language works the other way. You start with what customers say about their problem, their situation, their alternatives, and you build messaging that meets them there.

The practical application is straightforward. Take the most common phrases from customer interviews and open-text survey responses. Compare them to the language on your homepage, your ads, your email subject lines. Where the language matches, you’re probably in good shape. Where it diverges significantly, you have a messaging problem that no amount of media spend will fix.

This applies to conversion copy as much as brand messaging. Unbounce’s research into button copy is a useful illustration of how small language choices, informed by an understanding of what customers are actually thinking at the moment of decision, can have a measurable effect on conversion rates. The principle scales. If the micro-copy on a button matters, the macro-framing of your value proposition matters even more.

One thing I’ve learned from running agencies is that most messaging problems aren’t creative problems. They’re listening problems. The brief was written without enough customer language in it, so the creative team built something that sounds good internally but lands flat externally. A functioning VoC programme fixes the brief before it fixes the creative.

Building a VoC Cadence That Doesn’t Collapse Under Its Own Weight

The most common reason VoC programmes stall is that they’re designed to be comprehensive rather than sustainable. A team launches a full listening programme, runs it intensively for a quarter, produces a detailed report, and then quietly abandons it because the operational overhead is too high.

A sustainable cadence looks more like this:

• Continuous: Passive signal collection from review platforms, support tickets, and social mentions. Low effort. High volume. Reviewed weekly or fortnightly by a named owner.
• Monthly: A small batch of customer interviews, four to six, focused on a specific question the business is trying to answer. Transcribed and tagged within the same week.
• Quarterly: A structured survey to a larger segment, with results reviewed alongside the qualitative themes from the monthly interviews. Used to validate or challenge what the qualitative signal is showing.
• Annual: A deeper segmentation exercise that revisits the customer profiles the business is targeting and tests whether the assumptions still hold.

The cadence is less important than the consistency. A lightweight programme that runs every month for two years will produce more actionable insight than an ambitious programme that runs once and produces a report nobody reads.

When I grew an agency from twenty people to over a hundred, one of the things that made the difference was building a rhythm of customer conversations into the operating model rather than treating them as a project. Account managers had a standing brief to bring one piece of direct customer language to every internal planning meeting. It sounds small. Over time, it changed how the whole team thought about the work.

For more on how research and intelligence functions sit within a broader marketing strategy, the Market Research and Competitive Intel hub covers the full range of methods and frameworks worth knowing.

The Uncomfortable Truth About What VoC Reveals

A well-run VoC programme will, at some point, tell you something you don’t want to hear. That the product has a flaw the team is aware of but hasn’t fixed. That the onboarding experience is confusing in ways that support tickets have been quietly flagging for months. That the pricing model creates friction that customers tolerate but don’t forgive.

This is where the organisational politics of VoC get complicated. Marketing teams are often reluctant to surface findings that implicate product or operations, because doing so creates friction with other functions. So the insight gets softened, or buried in a report, or framed in a way that doesn’t require anyone to do anything uncomfortable.

I’ve believed for a long time that if a company genuinely delighted customers at every opportunity, that alone would drive growth. Marketing is often a blunt instrument used to prop up businesses with more fundamental problems. A VoC programme that’s working properly will surface those problems. The question is whether the organisation has the appetite to act on them.

The marketing teams I’ve seen get the most value from VoC are the ones that treat it as a business intelligence function rather than a marketing function. The insight belongs to the whole organisation. The routing protocol reflects that. Product gets product insights. Operations gets service insights. Marketing gets messaging insights. And the leadership team gets the picture of where the customer experience is falling short of the promise being made in the market.

That’s what a voice of customer framework is actually for. Not to produce a satisfaction score. To tell the business what it needs to hear.

About the Author

Keith Lacy is a marketing strategist and former agency CEO with 20+ years of experience across agency leadership, performance marketing, and commercial strategy. He writes The Marketing Juice to cut through the noise and share what works.

Frequently Asked Questions