Customer feedback analysis: how to turn comments into counted themes
A customer feedback analysis method: build a codebook, code every comment, count themes by people and accounts, and back each one with real quotes.
Free app. Runs on your Mac with the Claude or Codex plan you already have.
What customer feedback analysis is
Customer feedback analysis is the work of reading what customers have said, in surveys, tickets, reviews, interviews and sales calls, and turning it into a short list of themes with counts and quotes that a team can act on. Done well, every comment is counted once, under a clear heading, and the result answers a specific decision. Done badly, it is a word cloud, or a slide of the three most vivid quotes chosen to support a plan already made.
The standard approach is thematic analysis: read the comments, label each with short codes, group the codes into feedback themes and count them. The method below is how an experienced user researcher does it by hand. Each step says what to do and why, then how 10folds carries it out.
What it does for the business
- Decide from evidence, not the loudest voiceCounting how many distinct customers raise each theme stops one persistent customer, or one memorable call, from setting the roadmap. The team argues about what to do, not about what customers said.
- See problems behind the requestsSeparating what customers struggle with from the features they ask for shows the underlying problem. Often one fix answers several feature requests.
- Know who is affectedSplitting themes by plan, source or segment shows whether a complaint comes from your largest customers, your newest ones or one channel.
- Keep what customers valuePraise is coded too. Knowing what customers like protects it from being changed by accident while fixing something else.
Customer feedback analysis, step by step
Start from the decision and gather the feedback
Feedback answers a question better than it answers everything.
What to do
- Write the decision the analysis must inform, such as what to fix first next quarter.
- Collect feedback from every source for the same period into one table.
- Keep columns for record ID, date, source, person, account, a segment such as plan, and the text.
- Split long interview or call notes into one row per distinct point.
- Remove personal details from anything you will quote.
Why
A stated decision tells you which themes matter and how deep to go. Combining sources gives a fuller voice of the customer than any one channel, and keeping person and account columns makes honest counting possible later. Splitting long notes into single points matters because one interview can raise five separate issues, and each deserves its own code.
Read everything and build a codebook
If two people would code a comment differently, the definition is not clear enough yet.
What to do
- Read a broad sample first and note the recurring points in customers' own words.
- Write each theme as a short name, a one-line definition, what is included and what is excluded.
- Give each theme a kind: problem, request, praise or question.
- Keep a theme for minority and negative points, and an unclear code.
Why
A codebook, the written list of themes and their definitions, is what makes coding consistent. Separating problems from requests matters: the app crashes when I photograph a receipt is a problem, and add a receipt scanner is a request. Treating them as one theme hides what customers actually need. An unclear code and a minority theme keep awkward comments visible instead of letting them disappear into the nearest big theme.
Code every record and test the codebook
Every comment counts once under each point it makes, and never more.
What to do
- Assign each record one or more codes from the codebook.
- Code comments that fit nothing as unclear rather than forcing a fit.
- Recode a sample from the codebook alone and compare with your first pass.
- Tighten any definition that produced a different answer, then recode.
Why
Coding is where bias creeps in: a tired analyst starts filing everything under the themes already found. Recoding a sample is a simple check of consistency, and it often reveals two themes that overlap or one that is really two. Allowing several codes per record reflects how people write: one comment can praise a feature and report a bug in the same sentence.
Count by people and accounts, and by segment
Ten messages from one customer are one customer.
What to do
- For each theme, count mentions, distinct people and distinct accounts.
- Order themes by distinct people first, then mentions.
- Split each theme by segment: plan, source or customer type.
- Note the first and last date each theme appears.
Why
Mentions alone reward repetition. Counting distinct people and accounts shows how widely a problem is felt. Segment splits show who is affected and who is not, which often changes the decision more than the overall count. First and last dates show whether a theme is new, fading or constant, which matters when deciding how urgent it is.
Pick the quotes, find the dissent, write the brief
A vivid quote is not a trend; show how many records sit behind it.
What to do
- Choose two or three verbatim quotes per main theme, with their record IDs.
- Add at least one quote that shows the theme's limits or a customer who disagrees.
- Open the brief with the theme to act on first and why it beats the next one.
- For the top themes, give options, your recommendation and your confidence.
- State who is missing: customers who never write, and those who already left.
Why
Verbatim quotes make a theme concrete, and the dissenting ones keep the brief honest. A brief that names its gaps tells the reader how far to trust it. Feedback describes the people who wrote, never every customer. Giving options with a recommendation and a confidence level lets the decision maker see the reasoning and disagree with it openly.
Common mistakes
- Counting mentions onlyOne customer writing ten times is still one customer.
- Merging problems and requestsThe fix customers suggest is often not the best answer to their problem.
- Forcing every comment into a themeAn unclear code is more honest than a wrong one.
- Choosing quotes to fit a planShow the quotes that disagree as well, so readers can judge the theme's strength.
- Speaking for all customersFeedback describes those who wrote, not those who stayed silent or already left.
What a good result looks like
- A one-sentence answer to the decision, with counts.
- A codebook with a definition for every theme.
- Themes ordered by distinct people, with mentions, accounts and segments.
- Two or three verbatim quotes per main theme, including dissent.
- A note on what the feedback cannot tell you, and which customers are missing from it.
And then what? Using the result
- Take the top theme to the product team with its options and recommendation.
- Rank the feature requests against the problems they would solve.
- Reply to the customers behind the top theme once a decision is made.
- Run the analysis again next period with the same codebook, so the counts can be compared.
When to run it
- Before roadmap planningCounted themes give the team a shared view of what customers struggle with before priorities are set.
- After a launch or a pricing changeCoding the feedback from that period shows what customers noticed and how widely.
- When feedback sits in several toolsCombining surveys, tickets, reviews and call notes in one table gives a single picture instead of several partial ones.
- When the team disagrees on what customers wantCounts by people and accounts, with quotes, settle the debate with evidence.
Questions people ask
How do you do customer feedback analysis?
Gather feedback into one table, build a codebook of themes with definitions, code every comment, count each theme by mentions, people and accounts, and write a brief with quotes that answers a specific decision.
What is thematic analysis?
Thematic analysis is a method for finding patterns in text: you read the material, label passages with codes, group codes into themes and check each theme against the original words. Counting the themes afterwards shows how widely each one is shared.
How do you categorize customer feedback?
Use a codebook of themes, each with a short name, a definition and a kind such as problem, request, praise or question. A comment can carry more than one theme.
How do you prioritize customer feedback?
Start with themes raised by the most distinct people and accounts, weigh them by the segments that matter to your decision, and check the quotes behind them. Problems that affect many accounts in an important segment usually come before single requests, however loudly they are made.
Does 10folds put customer names in the report?
No. Quotes carry a record number, not a name. Personal details stay in your files.
Does it tell me what all my customers think?
No. It describes the feedback you supplied and says so. Customers who never wrote are not represented.
More jobs in Product
Summarise customer feedback into themes, on your own Mac.Free app. Runs on your Mac with the Claude or Codex plan you already have. Your files stay on your computer.
Download for Mac