Voice of customer (VoC): how to run a program and where interviews fit

Voice of customer (VoC) is the practice of capturing what customers need, want and complain about, in their own words, and getting it in front of the people who make product, sales and service decisions. A VoC program does this on a schedule: it collects from a few steady sources, sorts what comes in the same way every time, and uses live conversations to explain the patterns.

The term comes from product development. Abbie Griffin and John Hauser's 1993 paper "The Voice of the Customer" describes it as the part of Quality Function Deployment that identifies customer needs, structures them and sets priorities among them. Their work used one-on-one interviews and focus groups as the raw material. Many software teams now run VoC from support tickets, surveys and call notes as well, but the job is the same: turn scattered comments into ranked needs someone can act on.

A worked example

Suppose you run a payroll product for companies with 20 to 200 employees and have about 300 customers. This example is hypothetical. Support gets a steady flow of tickets, customer success has quarterly check-ins, and sales hears objections every week. Nobody can say which problem matters most, because each team remembers different customers. We will use this company in the sections below.

Start with sources you already have

You probably collect more feedback than you use. Before launching a new survey, list what already arrives:

Support tickets and chats. Customers write these unprompted, about problems they hit in real work.
Sales and customer success notes. Objections, feature asks, reasons for delay and renewal concerns.
Cancellation reasons and lost-deal notes. Short, often vague, but tied to money.
Public reviews and community posts. Useful for the words customers use about you and competitors.

Support tickets are often the richest of these. A team at the UK Government Digital Service described how it analyzes tickets for several products: agree what you want to learn, have a product manager, designer and tech lead each categorize the same 50 tickets, compare their tags to build one shared list of categories, then tag tickets as they arrive and report on volumes every two to three months. The team noted that tickets give feedback that is not steered by the team's own roadmap or interview questions.

For the payroll company, a shared tag list might include "tax filing error," "off-cycle payment," "new hire setup" and "integration sync." After a quarter, you can see which tags grow.

Add one or two structured measures

Structured questions give you a number you can track. Keep the set small and ask each one at a sensible moment.

Net Promoter Score. Bain defines NPS as the percentage of promoters (9 or 10 on "How likely are you to recommend us to a friend or colleague?") minus the percentage of detractors (0 to 6). If 120 payroll admins answer and 54 are promoters and 18 are detractors, NPS is 45% minus 15%, or 30 (hypothetical figures).
The product-market fit question. "How would you feel if you could no longer use the product?" works well for newer products; see our guide to the product-market fit survey.
A one-question check after a key event, such as the first payroll run or a support ticket closing.

A score tells you where things stand and whether they are moving. It cannot tell you why. NN/g's overview of research methods puts it plainly: qualitative methods suit questions about why or how to fix a problem, and quantitative methods suit how many and how much.

Where live interviews fit

Interviews are the part of a VoC program that explains the signals. Use them in three situations:

A tag or score moves and you do not know why. If "off-cycle payment" tickets double in a quarter, talk to five or six admins who filed them. Ask them to walk through the last time they needed an off-cycle payment, what they did and who else was involved.
Before a roadmap decision. Tickets tell you where people get stuck. They rarely tell you what someone was trying to get done or what it cost them.
When the people you need are not in your feedback stream. Customers who left, deals you lost and people who never tried you do not file tickets. Churned customers and lost deals each need their own approach, and prospects in your market can be found outside your customer list.

GOV.UK's guidance for live services lays out the same division: review analytics and support tickets to find problems, run surveys for broader feedback, then use interviews, visits and usability testing to understand the problems users report.

Source	What it tells you	What it misses	Typical rhythm
Support tickets	Where people get stuck, in their words	People who never contact support	Tag as they arrive, review monthly
NPS or other surveys	Direction of sentiment across many customers	Reasons behind a score	Quarterly or after key events
Sales and CS notes	Objections and renewal risks	Detail; notes are secondhand	Weekly skim, monthly summary
Interviews	Why something happens and what it costs	How common it is	A few each month, triggered by the above

Turn the inputs into a program

A VoC program is mostly routine. Decide:

Who owns it. One person collects, tags and reports, even if many people contribute.
One tag list shared by support, sales and customer success, reviewed each quarter.
A monthly summary with the top tags, any score change and what interviews found. Our research repository template is one simple way to store it.
Where each finding goes: product backlog, help docs, onboarding or sales training.
Closing the loop. Tell customers what changed because they spoke up.

Watch for the usual distortions

Who speaks up. Feedback forms and surveys hear from people who choose to answer; NN/g describes customer feedback as coming from a self-selected sample. Griffin and Hauser also found a self-selection bias in the satisfaction measures commonly used in QFD, so do not treat one group's answers as the whole customer base.
The loudest account. One large customer's requests can crowd out a pattern across fifty small ones. Count accounts as well as mentions.
Mentions as importance. A need that comes up often may be mildly annoying; a rare one may be why people leave. Ask about impact in interviews before ranking.

Your next step

Pull last quarter's tickets, tag 50 of them with two colleagues, and pick the tag that surprises you most. Then book five conversations about it. If the people you need are outside your customer list, such as payroll admins who use a competitor, search for people by describing their role, or start from the operations professionals in payroll software directory page. Instant Expert finds people whose work matches your question. You review them, it sends the invitations, and you pay for each call that gets booked.