Decide what the review represents
A product review should help another buyer understand a particular item, not act as an unexplained badge of approval. Before adding stars to a product page, define whether the review concerns the product, a delivery experience, or service from the shop. Those experiences can differ sharply. A customer may like an item but dislike a late parcel, or rate the delivery highly while finding the size unsuitable. Ask a question that matches the decision you plan to show. If the site says “product rating”, the input and display should refer to that product and its relevant variant. Keep the review close to the item it describes. Google’s review guidance also ties review markup to a specific item and visible review content, but the shopper’s need for an intelligible rating comes first. A row of stars with no scope or supporting text is weak decision support.
Set eligibility before invitations
Choose who may submit and explain the choice. A verified-purchase label can be useful if the shop can actually connect the reviewer to an eligible order, but it should not be attached merely because someone typed an order number. Decide how guest orders, gifts, exchanges, returns and multiple quantities are handled. You may permit non-purchasers to comment on a product they used, but then describe that status accurately. Avoid suggesting every review is a verified purchase when the system accepts broader submissions. An invitation email should link to the correct item, not a generic review form that risks attaching feedback to the wrong product. If an order contains several items, give a clear review path for each. These rules prevent a growing collection from becoming a mixture of claims whose origin staff cannot explain.
Collect useful detail without steering praise
A useful form asks what a future buyer needs to know: fit, material, setup, compatibility, durability at the time of use, or other attributes appropriate to the product. Keep the prompt neutral. “What worked and what did not?” invites more informative answers than “Tell us why you loved it.” Let the reviewer describe the context, such as the size selected or the device used, when that context changes the interpretation. Do not require a long essay or personal details unrelated to the review. If incentives are offered, design a prominent disclosure in the review display and apply the same invitation and moderation rules regardless of rating. Google explicitly warns against fake reviews and undisclosed incentivized reviews in content used for review snippets. Do not promise the customer search stars merely because the shop has collected comments.
Moderate by published rules
Moderation should remove material that cannot be safely or fairly published, not quietly erase criticism. Write a short policy covering personal contact details, abusive language, spam, irrelevant content, confidential order information and claims that need checking. Give staff a reason code and a route to request clarification when a review is useful but contains a private address or order reference. Where feasible, ask for a revised version rather than changing a customer’s meaning yourself. Keep the original submission and the decision record in staff systems so a dispute can be examined. Tell customers whether submissions appear immediately or after review, and do not imply that every submission will be published unchanged. A negative but relevant experience should remain eligible under the same rules as a positive one; otherwise the displayed distribution misleads buyers.
Show the distribution and context
An average rating without a count or relevant written context is easy to overinterpret. Show how many reviews contribute to the displayed figure, make the scale clear, and let a shopper inspect individual comments. A product with one high rating is a different decision from a product with many mixed experiences, even if their displayed averages happen to match. The design should not hide low ratings behind a default filter or present only a promotional selection as though it is the complete set. Explain sorting, such as most recent or most helpful, and make filtering reversible. If reviews from several variants are combined, state that fact and allow buyers to find the variant they intend to order. The review area should answer a buying question, not create a mathematical impression that the source material cannot support.
Keep product and service feedback separate
A product page can link to delivery or shop-service feedback, but combining them in one score hides why a customer gave a rating. The distinction matters when a shopper compares two models: a late courier does not tell them which model fits, while a sizing problem may be decisive. Define the mapping before staff import historical comments from email, marketplace messages or a previous store. Do not transplant a rating from a different website as though it were submitted to this page. Google’s review-snippet guidelines say not to aggregate ratings from other websites into the page’s review markup. Even where no markup is used, the same provenance question affects customer trust: could staff show how each displayed review reached this product page? If not, preserve the feedback internally until its source and scope are clear.
Reconcile the page and its markup
Structured data is optional and cannot repair an unreliable review process. If the shop uses review or aggregate-rating markup, its item, visible average and count must agree with what shoppers see on the relevant product page. Google says marked-up review content should be readily available to page users and should refer to a specific item; it also distinguishes product snippets and merchant listings. That is a technical eligibility rule, not a guarantee of a rich result. When a review is removed, corrected or moved to the proper variant, recompute the visible summary and the markup together. Do not leave a historic high rating in machine-readable data after the page changes. Test a product with no reviews, one review and several reviews. Each state needs an honest label, and an empty state should invite genuine feedback without inventing a score.
Work through a difficult case
Imagine a customer gives a jacket one star because the delivered size differs from the order. The comment includes a phone number and says the shop refused a refund. Staff should not publish the number, but deleting the review outright would also erase a potentially important signal. The reviewer can be asked to remove private data while the order team checks the size and service issue through its normal records. If the customer corrects the review, show the revised text and keep an internal record of the change. Decide whether the rating is about the jacket’s sizing, a packing error or service, and tell the customer how it will be categorised. The scenario illustrates a workflow, not a judgement that either party is right. A trustworthy site can show critical feedback while protecting personal information.
Give reviewers a correction route
A customer’s experience may change after a replacement, a longer period of use or a discovered misunderstanding. Provide a route to edit, withdraw or append an update, with enough identity checking to prevent another person changing the review. Decide whether edited reviews retain their original date, show an update date or carry a short notice. Do not silently replace a negative comment with a favourable one after support resolves a complaint. If the shop writes a public reply, identify it as the shop’s reply and avoid disclosing order details. The review should remain the customer’s account, while the reply explains what the business can verify or offer. Set a staff owner for disputes about authenticity and one for technical display; otherwise a correction can be approved in support but remain stale on the product page.
Test the whole workflow
Before launch, submit a positive review, a relevant negative review, a comment with private information, a review for the wrong variant and a corrected review. Follow each through invitation, submission, moderation, display, summary calculation, optional markup and withdrawal. Check the English and Traditional Chinese labels separately: “verified”, “latest” and “all reviews” must mean the same thing in both, without implying a verification the system did not perform. Ask a staff member unfamiliar with the build to explain why a particular review appears and how its score entered the average. If they cannot, simplify the rules and the display. The useful outcome is a repeatable publishing process that lets customers see enough context to make their own judgement, while the shop can correct mistakes without rewriting history.
Release checks
- Can a person complete the decision using the real page and a slow connection?
- Have staff tested the exceptional path and recorded who owns its repair?
