Create a feedback table, define a few useful labels, review the coded comments, and report themes with counts and limitations. End with an investigation question and an owner rather than treating sentiment as a product roadmap.
In this guide
How do five comments become six theme assignments?
This is a training set, not a representative customer survey. Setup instructions appears in 2 of 5 comments, packaging in 1, parts guidance in 1, and smooth setup in 2: six assignments across five comments. The sample suggests checking the setup card and spare-part instructions. It cannot establish how common either problem is among all customers.
| Comment | Practice text | Labels |
|---|---|---|
| C01 | The setup card skipped the first step. | Setup instructions |
| C02 | Setup was unclear and the box arrived crushed. | Setup instructions; packaging |
| C03 | Where does the spare clip fit? | Parts guidance |
| C04 | Setup was straightforward. | Smooth setup |
| C05 | Easy setup on my second order. | Smooth setup |
Collect a bounded set of comments
Choose a period and sources that answer the question in front of the team, such as what prevented first-time buyers from setting up a product. Retain a stable comment ID, date, source channel, product context, and the original text you are permitted to use. Remove unnecessary identifying details from the working material.
Track repeat contacts about the same event separately from independent customer experiences. A table of messages and a table of customers answer different questions. Explain which unit you are counting before presenting frequency or comparing channels.
Choose the next investigation using evidence and impact
A theme becomes a candidate for action when the team understands what happened and what work it obstructs. Add reach within the observed set, severity of the task failure, available workaround, and confidence in the evidence. Keep these dimensions separate before deciding priority.
For a useful handoff, name someone who will inspect the affected instruction, product behavior, or support response. Record what evidence would confirm the explanation. This gives the following reporting period a concrete question instead of another generic sentiment chart.
Move from comments to reviewed themes
- Read a varied sample and define labels around customer tasks or obstacles, such as finding instructions or locating a missing component.
- Write an inclusion rule and an exclusion example for each label. Allow a comment to carry more than one label when it genuinely covers several topics.
- Ask AI to propose labels with supporting text, then review the assignments and preserve uncertainty.
- Group the checked comments, count the stated unit, and retain contrary or successful experiences alongside the problems.
- Prepare a report with the theme, evidence IDs, observed scope, impact question, and proposed next investigation.
Ask for themes supported by original comments
Analyze this bounded feedback set. QUESTION, PERIOD, AND SOURCES: [state] COUNTING UNIT AND DUPLICATE RULE: [state] COMMENTS WITH IDS: [paste permitted text] CODEBOOK: [labels, inclusion rules, exclusions] Return proposed labels with supporting passages and uncertainty. Preserve comments that contradict a theme. Count comments or customers according to the supplied rule, and distinguish counts from representative rates. Suggest questions for investigation. Do not infer customer demographics, causes, or market demand beyond the supplied records.
Connect the finding to the appropriate work
Use interviews when the comments need more context, and turn a repeated documented question into a help article when the answer is already known. Keep the original evidence links with either output.