Voice of customer (VoC): a guide for software teams
7 min read
Voice of customer (VoC) is the practice of capturing what customers need, in their own words, and turning it into priorities a product team acts on. The term comes from quality function deployment in manufacturing, where researchers described three tasks: identifying customer needs, structuring them and prioritizing them. A software team already has most of the raw material in support tickets, feature requests and usage data. A VoC program is the habit of collecting it in one place, reading it on a schedule and showing customers what came of it.
| Source | What it captures | What to watch for |
|---|---|---|
| Support tickets | Problems, described the way customers see them | Skews toward users who are stuck right now |
| Feature requests and in-app feedback | Ideas and asks, often with a proposed fix | An ask is a solution; the need sits behind it |
| Customer interviews | Context, motivation, workarounds | Small samples, and the interviewer’s framing |
| Surveys | Satisfaction scores and structured answers | Only answers the questions you thought to ask |
| Usage data | What people actually do | Shows what happened, never why |
| Sales and churn notes | Why deals are won and accounts leave | Secondhand, filtered through whoever wrote it down |
What is voice of customer?
It is a structured way of listening: gather what customers say from every channel, translate it into needs, rank those needs, and feed the result into product decisions. The point is to make the customer’s view a regular input rather than something that reaches the roadmap only when a big account complains.
The idea grew out of quality function deployment (QFD). Hauser and Clausing’s 1988 Harvard Business Review article The House of Quality describes the house of quality as “the basic design tool” of QFD and traces it to Mitsubishi’s Kobe shipyard in 1972, after which Toyota and its suppliers developed it further. Griffin and Hauser’s 1993 paper The Voice of the Customer calls QFD a process “in which the ‘voice of the customer’ is deployed throughout the R&D, engineering, and manufacturing stages of product development,” and defines the VoC part as “the tasks of identifying customer needs, structuring customer needs, and providing priorities for customer needs.” Those three tasks still describe the work well for a software product.
Where does VoC data come from in a software product?
From every place a customer already talks to you, plus the behavior they leave behind. The table above lists the usual six; many teams have all of them and read only one or two.
Each source is biased in a different direction, which is the reason to combine them. Support tickets over-represent people who are blocked today. Feature requests come from users engaged enough to write in. Interviews go deep with a handful of people. Usage data covers everyone and explains nothing. A need that shows up in three sources is far more trustworthy than one that shows up loudly in one. Support tickets vs. feature requests compares the two most common sources in detail.
Collection works best when it happens where customers already are. An in-app feedback box catches the thought at the moment someone has it. In Changeloop, the widget has a Feedback tab, and submissions are filed as GitHub issues in your connected repository, labelled bug or feature-request, with a priority: label and from-widget, so customer input lands next to the code that will answer it. How to ask for customer feedback covers timing and wording for the other channels.
How do you turn customer feedback into needs?
Rewrite each piece of feedback as the problem behind it, then group the problems. “Add a CSV export” is a proposed solution; the need might be “get this data into our finance system every month,” which a scheduled integration could meet better.
Griffin and Hauser describe structuring needs into a hierarchy of “primary, secondary, and tertiary needs.” In practice that means a small number of broad themes, each with more specific needs under it, and the original quotes attached at the bottom. Keep the customer’s words; a theme called “reporting” loses the detail that one called “my accountant can’t read the export” keeps.
Merge duplicates as you go, because the same need arrives in many phrasings across channels. Duplicate feature requests covers how to merge them without losing who asked. Griffin and Hauser’s paper also raised the question of whether frequency of mention can stand in for importance, and it is worth keeping that doubt: a need mentioned often by small accounts and a need mentioned once by every churned customer are different signals.
How do you prioritize what customers say?
Score needs on how much they matter to customers and what meeting them costs you, and be explicit about which customers you are weighting. A ranked list nobody can explain will be overruled by the next escalation.
Two methods fit VoC data well. The Kano model sorts features by how customers react when they are present or absent, which separates the basics you must fix from the extras that delight. RICE scoring puts reach, impact, confidence and effort into one number that can be compared across requests. Prioritizing feature requests covers running RICE on a real backlog.
How do you run a voice of customer program?
Give it an owner, one place for input, and a fixed review rhythm. Without those three it becomes a spreadsheet someone built once.
- One inbox. Route tickets, requests, interview notes and survey comments to a single place, or at least a single tag across tools. Feature request tracking covers the plumbing.
- An owner. Usually product management, with a named person who reads everything each week.
- A regular review. Weekly triage of new input, and a monthly look at themes with support, sales and engineering in the room.
- A visible output. A short summary of the top needs and what the team decided about each, including what it decided not to do. Declining feature requests covers saying no well.
- A closed loop. Tell customers what happened to what they said; the next section covers how.
How do you close the loop with customers?
Tell the people who raised a need when you ship something for it. This is the step programs skip most often, and it is the one that keeps customers telling you things.
The changelog is the natural place for it, because every shipped change already gets an entry there. In Changeloop, when widget feedback became a GitHub issue and the merged pull request closes it, approving the changelog entry posts a “Shipped” comment on that issue and shows the submitter the entry in the widget. Issues filed by hand get no comment, so the habit still matters. Closing the customer feedback loop walks through the steps from request to announcement.
How do you measure a VoC program?
By what changes because of it. Volume of feedback collected says little; decisions and outcomes say a lot.
Useful numbers: the share of new feedback triaged within a week, the time from a need being recorded to something shipping for it, the share of shipped work that traces back to customer input, and whether requests on a theme fall after you ship for it. If the last number does not move, the team probably solved a different problem than the one customers described.
FAQ
What is the difference between voice of customer and customer feedback? Customer feedback is the raw input: a ticket, a request, a survey answer. Voice of customer is the program that gathers that input from every channel, turns it into ranked needs and feeds it into decisions.
Is NPS a voice of customer method? It is one input. A satisfaction score tells you how customers feel overall, but without the written reasons behind it, it cannot tell you what to build, so pair it with open questions and the other sources.
Who should own a VoC program? Product management usually owns it, because the output is product decisions. Support and sales contribute most of the input, so they should see the summaries and the decisions that come from them.
How is voice of customer different from a customer feedback loop? VoC is the whole program, from collecting input to ranking needs to deciding what to build. The feedback loop is the part that reports back to each person who spoke up, and a VoC program without it gets less input over time, because customers stop writing when nothing seems to happen.
The technical claims in this article have not been independently reviewed. If something here is wrong, tell us and we will correct it.