Define the question and collect suitable records
Decide what the analysis should help you do. “Understand why delivery feedback is worsening” is more useful than “analyse everything”. Choose the relevant locations, products, sources and dates, then obtain records through an appropriate permitted route.
Keep source links or references, publication dates, ratings and original text. Retain experience dates, languages and product versions only when supplied. Record missing data rather than inventing it.
Download the review analysis working template
The plain-text download contains fields to copy into your working document and clearly labelled synthetic examples. Remove those examples before using real records. Its fields are explained below.
Create a simple theme framework
Start with categories that map to business decisions: product quality, waiting, communication, delivery and value, for example. Read a small sample and adjust definitions before tagging the rest. One review can contain several themes.
Separate the subject from its sentiment. “The meal was excellent, but nobody explained the wait” contains positive food feedback and negative communication feedback. A single overall label would lose that distinction.
| Synthetic review | Theme | Aspect sentiment |
|---|---|---|
| Good food, long wait for the bill | Food; payment waiting | Positive food; negative waiting |
| Delivery was on time, packaging leaked | Delivery timing; packaging | Positive timing; negative packaging |
| Helpful team, unclear booking confirmation | Staff support; booking communication | Positive staff; negative communication |
Count consistently and show the denominator
Suppose 12 of 80 text reviews mention waiting. The waiting mention rate is 12 ÷ 80 × 100 = 15%. If a review repeats “waiting” three times, count that review once for this theme. Different theme rates may sum above 100% because one review can mention more than one subject.
This is a worked example, not a measured ReviewsBlender industry finding. A mention rate describes the selected review set; it does not estimate the proportion of all customers who experienced the problem.
Use the template to separate evidence from action
| Template fields | Purpose |
|---|---|
| Record ID, source, source reference, dates and location | Keep the record traceable and define the analysis window |
| Rating, scale, language and original text | Preserve source context |
| Themes, aspect sentiment and confidence | Record the interpretation and uncertainty |
| Evidence excerpt and operational hypothesis | Separate what was said from what might explain it |
| Action, owner, due date and follow-up measure | Make investigation and improvement trackable |
Avoid personal details that are unnecessary for the task. Secure access to working files and check reuse permissions before publishing review excerpts. Do not paste confidential patient, employee or customer records into a general-purpose tool without an appropriate handling basis.
Validate the interpretation and decide what happens next
Re-read examples from every important theme, check translation ambiguity and compare labels if more than one person is coding. Keep unclear records visible rather than forcing a confident interpretation.
Ask an operational owner to test the proposed explanation. A recurring queue complaint suggests an investigation; staffing, equipment, expectations and process can all contribute. Record the intervention and compare suitable follow-up periods. For prioritisation, use the action-planning guide.
If you would prefer the evidence organised into a report, the existing example shows the output ReviewsBlender provides.
Frequently asked questions
Does the template calculate sentiment automatically?
No. It provides a transparent working structure. Labels and conclusions need checking, whether a person or an AI tool assists with them.
How should we handle rating-only reviews?
Retain them for appropriate rating summaries, but do not assign textual themes when there is no text. State the text-review denominator separately.
Sources and further reading
- ReviewsBlender example operations intelligence report
- ICO: data protection and monitoring workers
- About ReviewsBlender and editorial standards
External product and policy information checked on 1 November 2026. Product descriptions are vendor documentation; recommendations and interpretations are ReviewsBlender’s. Linked policies take precedence if they change.
Find out what your reviews are telling you
ReviewsBlender produces a human-validated intelligence report for $99 one-off, delivered within 24–72 hours. Ongoing monitoring costs $59/month and delivers weekly intelligence reports with review response suggestions. Coverage is checked for your business before analysis begins.
No setup fees. Cancel monthly monitoring any time. At least 3 actionable insights in your report, or a refund.