A folder of customer comments can contain several different things: one message about two issues, a repeated copy of another message, a specific service concern and a broad preference. Counting every line as another unhappy customer produces a confident-looking result that the records cannot support.
This guide organises one defined batch of voluntary café feedback for an operational review. It starts after comments have arrived. It does not recruit research participants, resolve an individual complaint or measure customer retention. The example is original fiction, with no real guest messages or personal details.
Define the batch before choosing the themes
Name the outlet, included channels, receipt window and review cutoff. State whether inclusion follows the time a comment arrived or the service date it describes; they can differ. Keep a source manifest of every received item in scope, including later-excluded copies, so the review can explain what it used.
Use anonymous source references in the shared sheet. Keep original messages and any identifying details in the appropriate restricted system. A summary should preserve the meaning of the comment without copying names, contact details, order secrets or sensitive information. Record an event date only if the source supplies it; do not invent one from the receipt date.
Immediate service or sensitive concerns follow the established responsible process when they arrive. A themed review is not a reason to delay that response. This worksheet records a protected referral reference where needed; it does not become the complaint, medical or incident record.
Resolve copies without assuming who wrote them
Give each received item a reference and retain its channel, receipt time, source location and inclusion decision. If the same platform message was exported twice, link the second copy to the first with the evidence for that decision. Preserve both receipt records while including the underlying comment once.
Similar wording alone does not prove duplication. Two comments about waiting may describe different visits. Conversely, two different comments may come from the same person. Unless the existing records establish identity for an appropriate purpose, count source items rather than distinct customers. Mark unresolved duplication questions and keep them visible before using any tally.
Separate what was reported from what you infer
Break a comment containing several relevant points into separate observations, each linked back to its source item. Label a paraphrase as reported feedback. A guest saying the order status was unclear establishes what was reported, not the actual preparation time or the cause of the experience.
Keep three fields distinct: the observation, the theme assigned during review and the interpretation or question it raises. A statement about loud music can support a sound-level theme; it does not establish that the volume exceeded a standard or that all visitors dislike the atmosphere.
Write short definitions for the themes and the boundaries between them. Preserve the coding version and reviewer. If a theme changes, retain the earlier mapping and record the reason instead of silently moving difficult observations to a different category.
Worked example: six source items, eight observations
In fictional batch FB-025, Outlet North reviews feedback received through its approved comment channels from 4 to 6 November 2026, with a review cutoff of 7 November at 12:00 IST. The manifest holds seven received records. C07 is an evidenced second export of C02, so six underlying source items are included. Their authors are not counted as six different people.
C01 reports unclear order status and loud music. C02 also reports unclear status. C03 describes the music as welcoming. C04 reports unclear portion wording. C05 reports that order status was clear on that visit. C06 reports an uncomfortable seat and loud music. Splitting the two multi-point items creates eight observations, all retained below.
| Theme (fictional) | Observation references | Source items represented | Interpretation limit |
|---|---|---|---|
| Order-status communication | O01, O03, O06 | C01, C02, C05: 3 items | Two unclear reports and one clear report; keep the contrast |
| Music experience | O02, O04, O08 | C01, C03, C06: 3 items | Two loud reports and one welcoming report; no measured sound level |
| Portion wording | O05 | C04: 1 item | One report remains useful; not discarded for being isolated |
| Seating comfort | O07 | C06: 1 item | A reported experience, not a finding about every seat |
Read the counts as a description of this batch
The theme counts total eight because some source items contribute to more than one theme. They do not describe eight customers. They also do not mean that three out of every six café guests had an order-status problem: this is voluntary feedback, the identities are unknown and guests who supplied no comment are not represented.
Keep conflicting and isolated observations beside the repeated ones. C05 matters even though it challenges the emerging order-status concern. C04 remains visible even though portion wording appears only once. Record exclusions with reasons, but do not remove an observation simply because it weakens the preferred story.
Frequency can help describe the batch, but it does not by itself establish severity, priority or cause. The responsible team considers the specific issue, existing controls and evidence needed. A single relevant concern may require attention before a frequently mentioned preference.
Turn a theme into a checkable next question
For order-status communication, the next question could be whether guests receive a clear explanation at comparable service points. The shift lead can plan an appropriate observation under the existing operating process, recording the scope, evidence to collect and review date before starting. That is a proposed check, not a conclusion that staff failed or a promise of improvement.
Keep any individual service response linked to its own case. An operational action being completed does not show that a guest is satisfied. A later batch with fewer comments does not prove the action worked either; channel use, trading volume and logging practices may have changed.
Close this analysis review only when its source manifest, observation links, coding decisions and next actions have been checked. Keep the action outcomes open in their own records. Later comments belong in a new dated batch or an explicitly recorded extension, with the original review preserved.
Where the organising method comes from
GOV.UK’s user-research guidance separates observations, grouped themes, findings and actions. Those distinctions inform this original feedback record. This guide does not adopt its timing recommendations or advice to discard isolated notes, and it makes no claim that note-taking removes bias. The café workflow, references and example are original.
Put it into practice
Your action checklist
Your progress saves in this browser. No sign-up needed.
0 of 10 complete
From reading to doing
Feedback Themes
Open Feedback Themes for a fully manual review of one defined feedback batch. Record the outlet, channels, receipt window, cutoff and source manifest; keep duplicate decisions and excluded records linked to their evidence. Separate observation rows retain anonymous source references, reported meaning and evidence type. A versioned theme dictionary and mapping register preserve conflicting and isolated comments. Manual theme summaries distinguish observations from source items and never calculate a customer percentage. Interpretation, coverage gaps, next checks, ownership, corrections and dated review decisions have separate fields. The FB-025 example keeps seven received records, six included source items and eight observations distinguishable. Nothing is scored, sent, approved or closed automatically.
Open the Feedback Themes tab in the launch workbook. Keep a copy for your own business.
Download the free workbook See what’s inside the toolkit →Sources & further reading
- Analyse a research session (GOV.UK Service Manual) ↗
Retrieved 8 November 2026. Used only for separating observations, grouped themes, findings and actions. Timing recommendations, claims of unbiased notes and advice to discard isolated notes are not adopted. FB-025 and all comments, references and counts are original fiction.
Published by Keystonne Grow. Product links connect to the Keystonne ecosystem; check the linked product or service for its current scope.
