Voice of Customer Research in India: Sources, Coding and a Monthly Rhythm

How to run voice of customer research in India: which feedback sources to use, one coding frame for all of them, rates versus counts and a monthly VoC rhythm.

In shortVoice of customer (VoC) research puts two kinds of feedback side by side: what customers say when you ask them, through surveys, and what they say when nobody asked, in complaints, call notes, reviews and chats. Code both into one set of themes so a refund problem counts the same wherever it shows up. Read surveys as rates and complaints as counts, because mixing the two is the most common mistake. Then give every theme an owner. Hercules Works adds the outside view, category users and competitors' customers on its Indian panel, starting with 30 free responses.

Fig.A matrix of verified respondents with a sampled subset highlighted
Contents

Feedback You Ask For, and Feedback You Do Not

Most companies are not short of customer feedback. They are short of one place where it adds up. Survey scores live in one tool, complaints in another and call notes in a third, and each team reads its own slice. Ask which problem matters most this month and you will get three different answers.

Voice of customer research is the work of joining those slices, which makes it a programme rather than a survey. Below: which sources to start with, how to code them in one frame, why a complaint count is not a share of your customers, and a monthly routine that turns themes into fixes. Where we quote real numbers, they come from a study we have published. Everything else is labelled as an illustration.

What Counts as Voice of Customer Research

Feedback comes in two kinds. Asked feedback answers questions you chose, from a group you defined, so you can compare it month to month. Unasked feedback comes from people who decided to speak up. It is messier, and it is also where you find the problems you never thought to ask about.

SourceAsked or unaskedWhat it tells youMain bias
Relationship survey (NPS)AskedWhether the overall relationship is improvingVery happy and very unhappy customers answer most
Touchpoint survey (CSAT, effort)AskedWhich step in the journey is failingCovers only the steps you chose to ask about
Complaints and support ticketsUnaskedProblems serious enough to reportOver-represents customers who complain easily
Call-centre and chat notesUnaskedThe words customers use, in their own languageAgent summaries lose detail
App-store and marketplace reviewsUnaskedThe public view of product and deliveryExtremes, and some reviews are not genuine
Social media postsUnaskedFast-moving issues and toneSmall, vocal groups
Sales and exit interviewsAskedWhy people buy or leaveSmall samples and imperfect recall

You do not need to connect all seven sources at once. Start with one relationship survey and the complaint log, then add another source when a team is ready to review and act on it. Without a clear owner, even a useful source can end up in a dashboard no one checks. For the survey side, see our guides to NPS methodology and CSAT surveys.

Rates Versus Counts: A Worked Example

A common mistake in VoC reporting is treating a survey rate and a complaint count as if they measure the same thing. A survey gives you a share of a group you defined. Complaints give you the number of people who chose to report a problem. The figures may look comparable on a slide, but they answer different questions.

An illustration. Say you code 1,200 complaints in a month and 180 of them mention a refund delay. That is 15% of complaints. It is not 15% of customers, and it tells you nothing about the people who waited and stayed quiet. To size the problem, you have to ask. Survey 400 customers who requested a refund, and if 60 say it took longer than promised, that is 15% of refund requesters, give or take about 3.5 points in a simple random sample. The complaint count told you where to look. The survey told you how big the problem is.

Complaint volume also needs context. If orders double, complaints may double even if the experience has not changed. Track each theme per 1,000 orders to see whether its rate is actually rising.

One Coding Frame Across Every Source

A coding frame is a fixed list of themes, each with a one-line definition. Use the same frame for survey comments, tickets, call notes and reviews, so "refund delay" is counted once whichever channel it arrived through. The template below uses invented comments to show the shape.

CodeDefinitionExample comment (illustrative)Owner
Refund delayRefund not received by the promised date"Paisa abhi tak nahi aaya" ("The money has not come yet")Payments
Delivery damageItem arrived broken or already opened"Box was torn and the seal was broken"Logistics
Login failureCannot sign in, or the OTP does not arrive"OTP nahi aa raha" ("The OTP is not coming")Product
Staff behaviourCustomer reports a rude or dismissive agent"The agent cut the call on me"Support
Price changePrice felt too high or changed without notice"It was cheaper last week"Category team
OtherDoes not fit any code yetReview these monthlyVoC lead
  • Code in the language people wrote in. Customers in India write in Hindi, Hinglish, Tamil and much else. Translate for the report, not before coding, or the detail is lost.
  • Check that two people agree. Give two coders the same 100 comments, compare, and tighten any definition they split on.
  • Watch the "Other" bucket. If it keeps growing, a new theme is hiding inside it.
  • Review the frame each quarter, and map old codes to new ones so your trends survive the change.

Our gut health study shows how this works in practice. We coded 1,879 open answers to the question, “What is your current remedy missing?” An answer could be assigned to more than one theme, so the shares add up to more than 100.

ThemeShare of coded answers
Unclear or miscellaneous32.2%
Targeted digestive ingredients19.5%
Faster, lasting relief17.2%
Diet and routine14.8%
Better format and convenience8.8%
No missing feature8.4%
Guidance and support3.7%

The biggest bucket was the vague one. That was a finding, not a failure: people were broadly content with what they used and unsure what would make it better. So the report recommended a follow-up study to separate the two clear asks, targeted ingredients and faster relief, before anyone built a product around either. Keep the verbatims next to each code, too. One answer filed under targeted ingredients asked for "comprehensive, targeted enzyme support for specific, hard-to-digest food types". A line like that does more to move the team that owns the fix than any percentage.

Our open-ended analysis guide shows how to build and test a frame, and the multilingual survey guide covers collecting answers in Indian languages.

Adding the Outside View

These feedback sources mostly tell you about people who are still your customers. They do not explain why someone chose a competitor or left without giving a reason. Those perspectives can help explain why growth has stalled.

Two studies close that gap. A category survey asks people who buy in your category, competitors' customers included, how they choose and what lets them down. A churn study asks former customers why they left. Run the category survey every quarter and code its open answers in your usual frame. Then you can tell whether a theme is your problem or the whole category's, and that changes what you do about it. See competitor benchmarking surveys and churn surveys.

Customer Data, Consent and Vendors

VoC data is personal data: names, phone numbers, order details and what people said about you. India's Digital Personal Data Protection Act, 2023 governs how you collect it and how you share it with vendors who process it for you. So before you pipe tickets or call notes into a new tool, ask your compliance team which duties apply to you, and get the vendor's storage location, sub-processors and deletion process in writing. Our customer satisfaction platform checklist lists the questions to ask. This is not legal advice.

A Monthly VoC Rhythm

WhenWhatWho
WeeklyRead new low scores and complaints. Call back customers who asked to be contactedTouchpoint owners
MonthlyUpdate each code's count per 1,000 orders and its survey rate. Pick the top three themesVoC lead with owners
QuarterlyReview the coding frame. Run the outside-view survey of category usersVoC lead
After a fixRe-measure the same code and the same survey questionThe owner who made the fix

For each theme, report its count per 1,000 orders, its survey rate and response count, and the action taken. Assign an owner to each theme so the findings lead to follow-up. For a broader view of the customer journey, see customer experience surveys.

Running This Study on Hercules

Our platform fits the outside view: category users and competitors' customers in India. Surveys of your own customers and your unasked feedback stay in the tools that hold your customer list. This is a quarterly outside-view study at list price.

StepPlanResponsesPrice per responseCost
Test the questionnaireFree30₹0₹0
Small pilotPlus100₹150₹15,000
Quarterly outside-view wavePro400₹120₹48,000

Pro is sold in bundles of 400 to 1,000 responses and costs 20% less per response than Plus. Enterprise plans come with a dedicated research team and are priced on request.

The SuperJ network has 2 crore Indian respondents. The survey-eligible panel is a subset, and availability depends on your targeting. Members complete Aadhaar-based KYC once, and each study receives a zero-knowledge proof of verification. Surveys run in as many Indian languages as the study needs. Quality checks flag speeders, straight-lining and inconsistent answers. Read /panel-methodology and pricing.

The app reaches smartphone users only. To include households without smartphone access, use face-to-face fieldwork.

Start free with 30 responses

Frequently asked questions

What is voice of customer research?

It is a programme that combines asked feedback, such as surveys, with unasked feedback, such as complaints, call notes and reviews, coded into one set of themes with an owner for each. Our customer experience survey guide covers the survey side.

How is VoC different from a customer satisfaction survey?

A satisfaction survey is one source. VoC joins it with every other source, so you can see whether the problem a survey shows also appears in complaints and reviews. See the CSAT survey guide for that one source.

Which sources should a VoC programme start with?

Start with one relationship survey and the complaint log, both coded in the same frame. Add reviews, call notes and chats once each has an owner. The NPS methodology guide explains the relationship survey.

How do I analyse VoC feedback in Indian languages?

Code each comment in the language it was written in, check that two coders agree, and translate only for the report. If a large share lands in an unclear code, rewrite the question or follow up with interviews. The open-ended analysis guide shows how.

How many survey responses does VoC research need?

For a theme near 15%, 100 responses give a margin of about ±7 points and 400 give about ±3.5 points in a simple random sample. Use the sample size calculator for your own numbers.

Can a consumer panel survey my own customers?

No. A panel does not hold your customer list, so surveys of your own customers go through your own channels. A panel suits the outside view: category users and competitors' customers. See competitor benchmarking surveys.

Sources and further reading

Ready to get real consumer insights?

2 crore Indian respondents. Free to start.