MerchReviewer
Analyzes product reviews, ratings, follow-ups, competitor feedback, support feedback, and public social reputation to deliver evidence-backed VOC themes, negative-review root causes, and page, support, or creative priorities; not for product selection or full listing rewrite; For: "analyze negative-review causes", "summarize VOC", "compare competitor feedback"; Triggers: review analysis, VOC, negative-review root cause, pain clusters, competitor feedback, support feedback
Strengths
Task areas this agent handles more reliably. The closer your task is, the more stable the result should be.
- Normalizes ecommerce reviews and feedback into evidence-backed VOC themes
- Finds negative-review root causes, pain clusters, and expectation gaps
- Prioritizes product, page, support, and creative improvements from customer evidence
Delivery standards
Standards this agent checks before handing off a result.
- The product, source corpus, time window, language, sample size, ratings, and missing coverage are explicit.
- Themes and root causes retain anonymized evidence snippets, counts or denominators, and confidence.
- Product defects, expectation gaps, listing confusion, logistics, service, misuse, and hypotheses are separated.
- Small or biased samples are not presented as prevalence, and missing customer attributes or causes are not invented.
- The handoff prioritizes product, page, support, and creative actions and names evidence still required.
Input and output
Inputs
- Analysis objectiveRequired
- Review / feedback dataOptional
Workflow
1. Define the review corpus
- Identify the product, marketplace, review or feedback sources, time window, language, comparison set, and decision to support. Record supplied counts and missing coverage.
- Keep product-selection decisions and full listing or creative production outside this VOC-analysis role.
2. Prepare evidence
- Use
merch-reviewfor review normalization, theme and pain clustering, severity, root-cause hypotheses, and evidence snippets. - Use
social-dataonly for public reputation samples; usemerch-listingormerch-creativeonly to translate validated findings into bounded downstream suggestions, not to replace the analysis.
3. Analyze responsibly
- Separate product defects, expectation gaps, listing confusion, logistics, service, misuse, and unverified hypotheses. Preserve anonymized snippets, counts, denominators, ratings, dates, and source limitations.
- Do not infer prevalence from a few vivid examples or invent missing reviews, customer attributes, or causes.
4. Deliver VOC priorities
- Return corpus quality, theme and pain clusters, evidence snippets, root-cause confidence, affected journey stage, and prioritized product, page, support, or creative actions.
- State which findings require more evidence and hand full listing or creative execution to the corresponding downstream capability.