Skip to content

How to categorize customer feedback for product decisions

Build a practical customer feedback taxonomy with clear tag rules, worked examples and a way to handle overlapping themes without double-counting demand.

Blue, pink and yellow sticky notes arranged on a whiteboard
Photo: Jakub Żerdzicki / Unsplash · Unsplash License

To categorize customer feedback, give each comment a product area and a specific problem label, then record what kind of response it needs. Keep the original message attached. A useful category lets a teammate find related evidence and decide what to investigate next.

Start with the feedback waiting for a decision. A label such as "user experience" is too broad to explain why people cannot finish an invitation flow. "Invitation delivery" and "role selection" give the team different questions to investigate.

Choose categories around a product question

Imagine a project management app whose team wants to understand why customers struggle to bring contractors into a project. The messages and numbers below are fictional examples.

The team could start with the product area "Collaboration" and inspect the comments inside it. Some concern missing invitations. Others concern what a contractor will be allowed to see. Both belong to collaboration, but putting them in one undifferentiated pile hides the work involved.

For this question, a small tag dictionary might look like this:

Invitation delivery

Use this label when an invitation has not arrived or cannot be opened. Keep difficulty choosing a recipient's role separate.

Role clarity

Use this label when someone cannot tell what a role permits. A known permission that does not meet their needs belongs under guest access.

Guest access

Use this label when someone needs limited access for a person outside the team. Keep trouble delivering the invitation itself separate.

These are starting definitions for this example, not a universal list. A billing product will need different labels. Add a category when it separates work the team would investigate or handle differently.

Keep the problem, feedback type and decision status distinct

"Guest access" describes the subject. "Feature request" describes how the customer presented it. "Under review" describes what the team is doing. Each answers a different question, so keep them in separate fields or clearly named groups of tags.

For feedback type, use a short list that matches your intake: a request for a capability, a reported malfunction, difficulty using an existing feature, a question, or praise. Keep an "unclear" option for messages that need a reply before you can classify them.

Calling something a reported malfunction leaves room to investigate. "The invite never arrived" does not establish whether the app failed to send it, the address was wrong, or a mail filter blocked it. Record the report without converting a guess into a confirmed bug.

Priority needs a separate decision too. A frustrated tone can tell you how someone feels; the consequence of the problem tells you what is at stake. Ask what they could not do and whether they found a workaround before assigning urgency.

Work through an ambiguous comment

Consider this fictional message:

I couldn't tell which role would let our contractor see only this project. I sent the invite anyway, but she says nothing arrived.

There are two observations to preserve: the sender could not understand the role options, and the recipient reported a missing invitation. Apply "Role clarity" and "Invitation delivery" to the source message. If those issues need separate investigation records, link both records back to that message and its author.

Do not add "Guest access" yet. The message does not establish that the required permission is missing; the sender may simply be unable to find or understand it. A useful reply asks which role they selected and what access they expected.

GOV.UK's research analysis guidance recommends recording individual observations separately from interpretations, then grouping related observations into themes. That distinction is useful when reviewing written feedback too: keep what the customer said visible beside your interpretation.

For a longer conversation, use a follow-up interview to reconstruct the customer's last attempt. Correct the labels when you learn more, and retain a short explanation if the change affects an earlier decision.

Test the definitions with another person

Before tagging the whole backlog, choose a small batch that includes straightforward comments and awkward cases. Ask two teammates to classify it independently with the same dictionary, then compare where they disagreed.

If one person uses "Guest access" whenever a contractor is mentioned, the definition needs a clearer boundary. Add a positive example and a near miss: a request for a limited guest role belongs there; a question about an existing role belongs under "Role clarity" until clarified.

Disagreement can also reveal missing information. "Sharing is broken" is not enough to choose between delivery, permissions and an expired link. Keep it marked for clarification and ask the customer what happened. Creating more tags will not recover a missing description.

Keep the dictionary somewhere the people reviewing feedback already work. Give one person responsibility for resolving naming conflicts, so "invite email", "missing invitation" and "invitation delivery" do not become three competing labels.

Count people and accounts separately from messages

Suppose eight messages mention invitation delivery. Five came from one customer account; the other three came from three different accounts. That is eight messages from four accounts. If you split one message into two investigation records, it still represents the same source account.

Record the unit beside the number: messages, distinct people, or distinct accounts. Choose the unit that fits the decision, and retain source identifiers so another person can check the count. Where identity is unknown, say so instead of assuming each anonymous message represents a new person.

Overlapping labels need similar care. In a fictional review of ten accounts, four mention invitation delivery and three mention role clarity. If two accounts mention both, five distinct accounts mention either issue: 4 + 3 - 2 = 5. Adding the category totals would overstate that total.

Those five accounts belong to the reviewed feedback set. Without a suitable sample, the count does not tell you what share of all customers has the problem. Use it to identify people to follow up with and evidence to examine during feature prioritization.

Change the taxonomy without losing earlier feedback

Rename a label when the wording is unclear. Merge labels when reviewers repeatedly use them for the same problem. Split a broad label when its contents lead to distinct investigations, as with invitation delivery and role clarity.

Check the existing items during each change. Keep a simple record of the old name, new name and reason. When splitting a category, reclassify the historical items you can support; leave ambiguous ones marked for review. An old chart and a new chart will not be directly comparable if the category's meaning changed.

Review "unclear" items alongside the normal feedback review. Some need a customer reply, while others reveal a gap in the dictionary. Give each unresolved item an owner so it does not disappear into an unattended bucket.

Turn a category into the next investigation

For the collaboration example, the next review should answer a concrete question: are invitations failing to arrive, or are customers unable to choose the right access? Pull the relevant source messages and record what remains unknown for each issue. Assign someone to check delivery evidence and someone to test the role wording with affected users, if those are the investigations the evidence supports.

In Productroad's feedback boards, tags help group requests while comments retain the discussion. Keep the tag definitions in your team's existing documentation, and use the request status to communicate actual progress. Categorizing a request does not mean the team has committed to building it.

Once the team makes a decision, return to the relevant conversations with an explanation. That follow-up connects the classification work to a customer feedback loop: people who raised the problem can understand what happens next and tell you whether a later change helped.

Productroad helps you collect user feedback, prioritize it, make a roadmap of a project and inform customers via changelog about new cool features.

Free 14 day trial. No credit 💳 card required.