# Voice of the Customer (VoC): Frameworks and Research Methods (2026)

> What Voice of the Customer (VoC) is, the main research methods and frameworks (needs hierarchy, Kano, journey maps, CTQ), sample sizes and analysis.

Source: https://www.pulselake.co/blog/voice-of-the-customer-voc-frameworks-and-research-methods-2026
Published 2026-10-26 · by Venkat Chandra · PulseLake

Voice of the Customer (VoC) is the structured practice of capturing what customers need, expect and struggle with, in their own words, and turning it into priorities a team can act on. The usual toolkit is a handful of collection methods (interviews, surveys, support and review data, observation) plus a framework for organising the results, such as a needs hierarchy, a Kano classification, a journey map or a critical-to-quality (CTQ) translation. A common recommendation is to start with qualitative interviews to find the needs, then use quantitative work to size and rank them.

We make PulseLake, a research platform, and mention it once near the end. The rest of this guide is method-first and applies with any tools.

![Customer comments grouped into a short list of needs.](https://www.pulselake.co/blog/img/production/80a3d70ef23333360a105a361212bf8f96e77121-1200x840.png?w=1600&fit=max&auto=format)

*Voice of the Customer (VoC): Frameworks and Research Methods (2026)*

## What is Voice of the Customer (VoC)?

Two definitions are in common use, and it helps to know which one a person means.

The research definition comes from product development. In a note summarising the 1993 paper by Abbie Griffin and John Hauser, Hauser describes the voice of the customer as a hierarchical set of customer needs, each with a priority, where a need is a description "in the customer's own words" of the benefit to be fulfilled ([Hauser, MIT Sloan](https://stuff.mit.edu/~hauser/Papers/NoteontheVoiceoftheCustomer.pdf)). A need is not a solution (a specific monitor type) and not an engineering measure (the number of breaks in a line). It is how the customer wants the result to look or work.

The programme definition comes from quality management and customer experience. Quality-One describes VoC as the needs, wants and expectations of the customer, stated and unstated, and as a continuous process of collecting, documenting and analysing customer feedback ([Quality-One](https://quality-one.com/voc/)). Most vendor guides use this broader sense: an ongoing programme that combines several feedback sources.

Both agree on the core idea. VoC is not a satisfaction score. A score tells you how people rated something; VoC tells you what they were trying to do and why it went well or badly.

## What are examples of VoC?

Examples of VoC data, from least to most structured:

- A customer says in an interview, "I never know whether my export finished." That sentence is raw VoC.
- A support ticket reads, "Can't find where to change the billing contact."
- A product review praises setup speed but complains about reporting.
- An open-ended survey answer explains why a respondent gave a 6 out of 10.
- A recorded sales call where a prospect lists what they currently work around.

Examples of VoC outputs, after analysis:

- A short list of primary needs (Hauser describes roughly two to ten), each with sub-needs written in customer language.
- A ranked list of problems by how many customers raise them and how much they matter.
- A journey map showing where customers get stuck.
- A set of measurable targets (CTQs) such as "export confirmation visible within a stated time".

A hypothetical example: a software team hears "reporting is confusing" in four calls and sees it in many reviews. Interviews reveal two different needs hidden in that phrase, finding last month's numbers and sharing a view with a manager. The fix, and the metric, differ for each.

## What are the main VoC research methods?

Methods split into asking customers, listening to what they already say, and watching what they do. A fair way to choose is by the question each answers.

| Method | Best for | Main limitation |
| --- | --- | --- |
| One-on-one interviews | Discovering needs and the reasons behind behaviour | Slow to run and analyse |
| Focus groups | Reacting to concepts and wording | Shared air-time; dominant voices |
| Surveys (CSAT, NPS, custom) | Measuring how common something is and tracking change | Closed answers miss the "why"; response bias |
| Open-ended survey and review text | Customer language at scale | Needs coding to be usable |
| Support tickets and call recordings | Unprompted problems | Only reflects customers who make contact |
| Social listening and public reviews | Unsolicited opinion | Noisy; skewed toward extremes |
| Behavioural and usage data | What people actually do | Shows what, not why |
| Observation and diary studies | Needs people cannot easily articulate | Time-intensive |

Practitioner guides group these in similar ways. Thematic lists surveys, interviews, focus groups, social listening, online reviews and behavioural data ([Thematic](https://getthematic.com/insights/voice-of-customer-research)), and Help Scout lists ten techniques that also include listening to recorded calls and talking to customer-facing employees ([Help Scout](https://www.helpscout.com/blog/voice-of-customer/)).

### Should you use qualitative or quantitative methods?

Use both, in sequence. SurveyMonkey's guidance is to use qualitative interviews first to find the right questions, then validate with a quantitative survey ([SurveyMonkey](https://www.surveymonkey.com/learn/customer-feedback/voc-research/)). Qualitative work answers "what do customers need and why"; quantitative work answers "how many, and how much does it matter". Doing only the second means measuring needs you never discovered. For more on combining them, see our piece on [mixed methods research](https://www.pulselake.co/blog/mixed-methods-research-done-right).

### How should you ask about needs?

Hauser's note makes a point that many guides skip: customers often have difficulty listing their needs when asked directly, so it is better to infer needs from experiential interviews or observation. In an experiential interview the customer describes a real or imagined use, and the interviewer probes until no new needs surface ([Hauser, MIT Sloan](https://stuff.mit.edu/~hauser/Papers/NoteontheVoiceoftheCustomer.pdf)). "Walk me through the last time you did this" usually produces more than "What do you want?"

## How many customers do you need to interview?

The most-cited evidence comes from Griffin and Hauser. In one study, 30 potential customers of portable food-carrying products were interviewed and the needs merged across analysts. Interviewing 20 of them identified over 90% of the needs that all 30 produced. Hauser adds that with modern interviewing methods about twenty respondents should be sufficient in most product categories, and that if segments are quite different, 10 to 20 interviews per segment may be needed ([Hauser, MIT Sloan](https://stuff.mit.edu/~hauser/Papers/NoteontheVoiceoftheCustomer.pdf)).

Two cautions. The result is one study in one product category, and it describes finding needs, not measuring how common they are. For measurement, survey sizing follows normal statistics: SurveyMonkey's guide gives about 370 responses for a 10,000-customer base at a 5% margin of error and 95% confidence, and notes you must invite more people than you need because not everyone responds ([SurveyMonkey](https://www.surveymonkey.com/learn/customer-feedback/voc-research/)).

Hauser's same note compares formats: in a study of office equipment, two one-hour interviews found about as many needs as one two-hour focus group (51% versus 50%), which suggests one-on-one interviews are at least as efficient as groups for discovery.

## What are the main VoC frameworks?

Methods collect the voice; frameworks give it shape. Four are used often, and they answer different questions.

### The needs hierarchy (Griffin and Hauser)

Discussions with customers typically produce 100 to 200 phrases that could be needs, which is too many for a team to work with. The classic approach structures them into primary (strategic), secondary (tactical) and tertiary (operational) needs, with primary needs numbering roughly two to ten, each elaborated into three to ten secondary needs. The team then measures the priority of each ([Hauser, MIT Sloan](https://stuff.mit.edu/~hauser/Papers/NoteontheVoiceoftheCustomer.pdf)).

Structuring can be done by the team in an affinity chart, or by customers sorting need cards into piles. In Hauser's food-carrying example, managers judged the customer-sort hierarchy clearer than the team's affinity chart. Use this framework when you are defining a product, a positioning or a service from scratch.

### The Kano model

Kano classifies features by how they affect satisfaction: must-be (expected; absence causes strong dissatisfaction), one-dimensional (more is better), attractive (unexpected delighters), indifferent and reverse. It originates in a 1984 paper by Noriaki Kano and colleagues. Respondents answer a pair of questions per feature, how they would feel if it were present and how they would feel if it were absent ([Wikipedia, Kano model](https://en.wikipedia.org/wiki/Kano_model)). A feature's category can shift over time as competitors improve, so re-test periodically. Our guide to [Kano analysis](https://www.pulselake.co/blog/kano-analysis-explained) walks through the method. Use Kano when deciding which features to build.

### Customer journey mapping

Nielsen Norman Group describes a journey map as a visual timeline of a persona's goals, actions, thoughts and emotions across phases, with touchpoints and the insights, pain points and owners that follow from them ([NN/g](https://www.nngroup.com/articles/customer-journey-mapping/)). NN/g's advice is that it should support a known business goal. Use it when the problem is spread across teams and no one sees the whole experience.

### Critical to quality (CTQ)

In Six Sigma, VoC is translated into measurable requirements. One Six Sigma guide describes separating needs from wants, converting feedback into measurable requirements and using tools such as Pareto charts to prioritise ([Six Sigma Institute](https://www.sixsigma-institute.org/Six_Sigma_DMAIC_Process_Define_Phase_Capturing_Voice_Of_Customer_VOC.php)). Quality-One describes a related step: linking customer "whats" to design "hows" in a House of Quality matrix ([Quality-One](https://quality-one.com/voc/)). Use CTQ when you must turn "it feels slow" into a number a process can be held to. Using VoC inside Six Sigma projects, and building a standing VoC programme, are separate topics we cover in their own guides.

### The collect-analyse-act loop

Thematic describes a five-stage programme model: collect, analyse, share, act, monitor ([Thematic](https://getthematic.com/insights/voice-of-customer-research)). It is less a research framework than an operating cycle, and it is where many programmes stall. SurveyMonkey's guide, for example, ties surveys and qualitative deep dives to a cadence, from continuous transactional surveys to quarterly qualitative work ([SurveyMonkey](https://www.surveymonkey.com/learn/customer-feedback/voc-research/)).

### Which framework should you choose?

| Your question | Start with |
| --- | --- |
| What do customers need from a new product or segment? | Needs hierarchy |
| Which features should we build or cut? | Kano |
| Where does the experience break down end to end? | Journey map |
| How do we set measurable targets from complaints? | CTQ |
| How do we run this continuously? | Collect-analyse-act loop |

Pick one primary framework and add others as the programme matures; they are complementary, not competing.

## How do you analyse VoC data?

Collection is rarely the weak point. Analysis is. For open text, the standard approaches are thematic coding, sentiment analysis (best read alongside themes) and keyword or entity extraction ([SurveyMonkey](https://www.surveymonkey.com/learn/customer-feedback/voc-research/)).

A practical sequence:

1. Keep the customer's wording. Code from the original phrase, not a paraphrase.
1. Build a codebook with a definition and an example per theme, and write down exclusions.
1. Have two people code a sample and compare. Disagreement usually means a theme is vague. Agreement checks like this are easy to skip and worth keeping.
1. Count themes, but also weight them: a rare complaint from your highest-value segment can outrank a common one from a low-value segment.
1. Separate observed feedback from your interpretation in the report.

If you use AI to code at scale, treat it as a first pass and check a sample against human coding; see [qualitative coding at scale](https://www.pulselake.co/blog/qualitative-coding-at-scale).

## What are common VoC mistakes?

- Treating NPS or CSAT as the whole programme. A score without the reason cannot be acted on.
- Listening only to customers who contact you. Support data omits silent churners and non-buyers.
- Asking about solutions ("Would you like a dashboard?") instead of experiences.
- Collecting without a decision attached. Each study should name the decision it informs.
- Skipping consent and data-handling checks for recorded calls and transcripts.
- Not closing the loop. Telling customers what changed encourages future feedback.

## Where does PulseLake fit?

PulseLake supports traditional research, AI-led interviews with real participants, synthetic research and simulation, with a persistent study context that keeps objectives, methods and evidence together, and evidence provenance for answers drawn from study data. In VoC terms, that is most relevant to the discovery and analysis steps: running interviews, coding responses and keeping the link between a theme and the quotes behind it. It does not replace a support desk or a case-management system, and you should evaluate it against your own interview and analysis process.

## Frequently asked questions

### What does VoC stand for?

VoC stands for Voice of the Customer. It refers to the process and the resulting body of evidence about customer needs, expectations and experiences, usually in the customers' own words.

### What is the difference between VoC and customer satisfaction research?

Customer satisfaction research measures how customers rated past performance. VoC explains why: it captures needs and reasons, so scores can be interpreted and acted on. SurveyMonkey frames VoC as a diagnostic and satisfaction research as a measurement of past performance ([SurveyMonkey](https://www.surveymonkey.com/learn/customer-feedback/voc-research/)).

### What are stated, implied and silent customer needs?

Stated needs are those customers can put into words. Implied needs are taken for granted (a washing machine should clean clothes). Silent or latent needs are ones customers cannot easily express and that you infer from context and observation. Sources differ on the exact terms. Hauser uses basic, articulated and exciting needs for a similar split ([Hauser, MIT Sloan](https://stuff.mit.edu/~hauser/Papers/NoteontheVoiceoftheCustomer.pdf)).

### How often should you collect VoC?

It depends on the method. SurveyMonkey suggests continuous transactional surveys, relationship surveys every 6 to 12 months and qualitative deep dives quarterly or per cohort ([SurveyMonkey](https://www.surveymonkey.com/learn/customer-feedback/voc-research/)). Treat that as a starting point, not a standard, and adjust to how quickly your product and market change.

### Who owns VoC in a company?

There is no single owner in the sources reviewed. In practice it is shared: research or insights runs the studies, support and sales supply raw feedback, and product or operations decides what to act on. Name one person accountable for the loop, or findings will not reach decisions.

### Can synthetic respondents replace real customers for VoC?

No. VoC is about what real customers say and do. Synthetic respondents can help test an interview guide or explore hypotheses before fieldwork, but their output should be labelled and validated against real customers.

## Related reading

- [Kano analysis explained: how to prioritize features](https://www.pulselake.co/blog/kano-analysis-explained)
- [Qualitative coding at scale](https://www.pulselake.co/blog/qualitative-coding-at-scale)
- [Mixed methods research done right](https://www.pulselake.co/blog/mixed-methods-research-done-right)