What Is Voice of the Customer?

Voice of customer (VoC) research is the standing practice of collecting what customers say, prompted or unprompted, alongside evidence from what they do, and turning it into decisions about product, pricing, packaging, messaging and service. Many teams still hear from customers only once a quarter, in a report that arrives after the decision it should have informed; a standing program listens every week.

Tropicana’s 2009 packaging redesign shows what that gap can cost. Consumers complained in writing by letter and email and also phoned the company. Ad Age reported that sales fell 20% between January 1 and February 22, and The New York Times reported that the company discontinued the design. The feedback came, but only after launch. A voice of customer program exists to reverse that order.

What is voice of the customer?

Abbie Griffin and John Hauser’s 1993 Marketing Science paper “The Voice of the Customer” defined voice of the customer as a hierarchy of customer needs, each described in the customer’s own words and ranked by importance, elicited through interviews to feed product design.

Forrester describes every VoC program as a closed loop of four activities: Listen, Interpret, Act and Monitor. A standing program keeps that loop turning; a quarterly survey opens it once and closes it in a slide deck.

What counts as voice of the customer data?

Gartner’s 2017 taxonomy sorts VoC data into three kinds (direct, indirect and inferred) by whether the customer knows the brand is listening, is talking about the brand to someone else, or is saying nothing and doing something.

Direct feedback

Direct feedback is what consumers give when they know the brand is asking: surveys, complaints and support tickets sent to the company, interviews, focus groups, panels. A conversation returns the reasoning in the consumer’s own sentences, and it can follow a surprising answer somewhere the questionnaire never anticipated. Human-moderated in-depth interviews and focus groups collect this kind of input. AI-moderated interviews collect it too, as live voice-to-voice conversations with real participants; what differs is who asks the follow-up question and how many conversations can run at once.

Indirect feedback

Indirect feedback is commentary customers volunteer without being asked, including online reviews, forum threads and social posts. Indirect signal is candid and uninvited, which makes it a good early warning and a poor sample. The customers who post felt strongly enough to type, so complaint volume measures intensity better than prevalence.

Inferred signals

Inferred signals are what customers do: purchase history, product usage, app journeys, abandon rates, wait times. Tropicana’s sales drop was an inferred signal, but a sales line can only confirm that shoppers walked away. It cannot say whether they failed to recognize the pack or disliked it. Nor can it show whether they assumed the juice inside had changed. An insights team uses inferred data to decide where to look and direct conversations to learn what it is looking at.

Why does voice of customer research matter for consumer brands?

For a consumer brand, VoC is the difference between learning about a packaging misfire from shoppers and learning about it from the sales report weeks later. Insights teams serve brand, strategy and commercial colleagues who decide which of three pack designs to launch and where to price a new size, and who need to know why trial buyers churn after month two.

Forrester’s 2024 US Customer Experience Index release reported that customer-obsessed organizations saw “41% faster revenue growth, 49% faster profit growth, and 51% better customer retention” than organizations that were not (Forrester does not name the sample behind these figures). If reversing one of these decisions after the retail reset will be expensive, the insights team should hear customers’ reasons before the brand team commits.

Which metrics measure voice of the customer, and what do they miss?

Three scores carry most VoC dashboards:

• Net Promoter Score (NPS): Bain’s NPS asks customers how likely they are to recommend the brand on a zero to ten scale and subtracts the percentage of detractors (0 to 6) from the percentage of promoters (9 to 10), for a range of −100 to 100.
• Customer Satisfaction (CSAT): CSAT has no single owner or formula; the common practitioner version asks for a rating on a 1 to 5 scale and reports the share scoring 4 or 5 as a percentage.
• Customer Effort Score (CES): CES, developed at CEB and now part of Gartner, asks customers to rate agreement with “The company made it easy for me to handle my issue” on a 1 to 7 scale, and Gartner scores it as “the percentage of customers that at least ‘somewhat agree’ (those who give a 5 or above).”

An independent 2026 retail satisfaction study of about 31,000 surveys, one of the few independent benchmarks, puts online retailers at 79 and supermarkets at 78 on its 0 to 100 scale.

Vendor benchmarks vary by source and sample, so a brand should track movement against one chosen source rather than compare figures across publishers.

Each score records that a customer feels a certain way and nothing about why. One study in the University of Cambridge repository found that 42% of customers scoring 9.5 or higher out of 10 still complained.

When the question is how many customers feel something, the score answers it; when the question is why and what to change, insights teams need the conversation, run on the same cadence as the score.

How do teams turn voice of the customer into decisions?

For an insights team, Forrester’s loop becomes three jobs:

• Listen: Keep direct and indirect feedback channels open. Track inferred signals alongside them, so a tracker movement can be matched to conversations with the customers behind it that week.
• Analyze: Code open-ended feedback into themes and count how many customers share each one. Attach the verbatim that proves it.
• Act: Route each theme to the team that owns the fix, such as a recognition problem after a pack redesign to the pack design team, and tell the customers who raised it what changed.

Nielsen Norman Group describes AI coding of open-ended feedback as “a preliminary first pass” that can omit data or “manufacture insights.” In practice, a theme in a readout should open to the transcript and clip that produced it.

Bain defines closing the loop as sharing feedback with the employees responsible, following up with customers, fixing individual problems and starting on systemic ones. Bain’s Logitech example shows the systemic side: verbatim analysis on the Logitech MX 5000 found unreliable Bluetooth connectivity and a hard-to-read LCD. Customers also found recharging difficult. The next model’s average NPS came in 27 points higher. If no team owns a theme by the next weekly review, the theme has not been acted on.

Which tools can teams use for voice of the customer research?

Qualitative platforms provide interview depth; survey and NPS platforms track scores. CX feedback suites capture multichannel feedback, while social listening tools track public commentary; the table compares interview depth, scale, analysis, source links, cadence and service level:

Qualitative research platforms

A qualitative research platform lets a team run interviews to learn the reasons behind its scores. Strella, an AI-native research platform, runs AI-moderated interviews as voice-to-voice conversations with real people. The moderator asks the follow-up question a researcher would ask, based on what the participant said, and hundreds of those conversations run at once.

• Researchers receive thematic analyses and source-linked verbatim quotes through automated synthesis, plus video highlight reels organized around the research objectives. They can add past interviews and customer calls to the repository and ask questions across every study.
• Teams choose AI, human or hybrid moderation study by study on the same platform, and human-moderated interviews are live. Human moderation is valuable where cultural nuance or emotional weight is high, and these sessions are led by the team’s own moderators. AI moderation for research is its own method, sitting between a survey and an in-depth interview, while a seasoned moderator brings the ability to read a room.

Ritual ran AI-moderated pricing tests on Strella, and its published case study reports a 24% increase in net revenue per user. To hear how the moderator follows up, try an interview.

Listen Labs runs AI-only interviews in which an AI moderator asks real participants follow-up questions across voice, video and text in 120+ languages, with a global participant panel, emotion and hesitation detection, and a research library that compounds findings across studies; it is built for high-volume interview programs, and Forbes reported 1M+ interviews conducted. Conveo runs AI-led video interviews with facial coding and embedded quant methods (MaxDiff, conjoint, Kano and Van Westendorp), links every insight to source data, and is aimed at enterprise CPG insights teams. Both run AI-only interviews, while Strella runs AI-moderated interviews with real participants, alongside human-moderated and hybrid interviews, on one platform, with levels of service from self-serve to fully managed.

Survey and NPS platforms

Survey and NPS platforms such as Qualtrics send surveys on a trigger or on a schedule and track NPS, CSAT and CES, and some have added AI analysis.

CX feedback suites

Medallia, a CX feedback suite, captures surveys, voice, chat, web behavior, social and employee feedback, and announced that Forrester named it the only Leader in its Q3 2026 Wave for customer feedback management and analytics. In practice these suites route feedback and feed dashboards; teams that need interview depth pair them with a qualitative platform.

Social listening tools

Social listening tools are a category of software in which the tool collects public posts from social networks, forums, news and review sites and alerts the team to shifts. Brandwatch processes half a billion posts a day from more than 100 million sources across X, Instagram, TikTok, Reddit, forums, news and review sites, with alerts on shifts, and other tools cover 30+ channels with live alerts and unified NPS and CSAT reporting. These tools capture indirect signal at a volume no interview program can match, and none can ask the poster a follow-up question.

If teams need a count for the decision, a survey platform or CX suite fits. If they need the reason at tracker scale, they should first evaluate a qualitative platform that links every finding to its source interview.

How is voice of the customer different from related concepts?

VoC is one of several “voices” the quality tradition names:

Meyer and Schwager’s 2007 HBR article separates CX from CRM: customer experience management is a practice in which the company collects what customers think of it, including through voice of customer research, and treats that as leading information, while CRM is a practice in which the company records what it knows about the customer, such as service requests and returns, and treats that as lagging. VoC is the listening component of CX.

Want to try out Strella for VoC research? Schedule some time with our team:

FAQ

What is a good voice of customer score?

A good score beats the brand’s own last reading and sits at or above its category on the same source, because publishers use different scales and samples. Movement against one chosen source means more than an absolute number.

How should an insights team present voice of the customer findings to executives?

Lead with the business outcome, then the evidence behind it. Forrester advises that the C-suite cares about revenue growth and cost reduction rather than NPS on its own. Profitability also matters to those executives. Pair the number with a customer’s own sentence or clip; insights leads find a clip gives a theme a face the number alone lacks.

What methods collect voice of the customer data?

Teams collect voice of customer data through:

• Surveys and NPS trackers
• Human-moderated and AI-moderated interviews
• Focus groups
• Support tickets and contact-center transcripts
• Reviews and social listening
• Behavioral analytics

Teams use surveys to count and interviews to understand why. They monitor feedback through transcripts and listening tools, alongside behavioral analytics; in a standing program, they draw on at least one of each.

What does closing the loop look like in practice?

Bain reports that Schwab and Apple’s retail division contact negative raters within 24 hours, and teams meet in 15-to-30-minute huddles to turn callbacks into structural fixes.

Teams without bandwidth can use Strella’s fully managed service, where Strella’s research team scopes the study, recruits participants, runs the AI-moderated interviews and synthesizes the findings. Bring your research question to us and we will talk it through. See Strella in action