Answer in brief
A review measures the gap between what your listing promised and what came out of the box. How to sort the feed by type of breakage, write replies aimed at the next buyer, and why a flawless rating works against you.
The short answer: a review is product data: A review measures the gap between the promise in the…
Handling reviews on a marketplace comes down to four moves repeated on a loop: collect the feed, sort it by which part of the business is broken, fix whatever repeats, and reply in public so the next buyer can weigh the risk before paying. This is product data work, not reputation defence.
The public reply is not really addressed to the person who wrote it. They already have the item and almost never revise their rating. It is read by the shopper who is still deciding and who is watching how a seller behaves when something goes wrong.
Buying reviews is off the table. Every major platform prohibits it in the seller rules, and several countries attach penalties of their own, which the eighth section sets out. The price of being caught is a suppressed listing or a closed account.
One counterintuitive point belongs up front: a listing carrying nothing but five stars often converts worse than one with a mixed rating and specific detail. The ninth section explains the mechanism, and it is easy to check against your own catalogue.
What a review actually measures
A review almost never measures product quality in the abstract. It measures the gap between the expectation your listing created and what the buyer pulled out of the box. The same item earns delight and fury because the expectations behind them differed.
That expectation is assembled from the photographs, the size chart, the material line, the stated dimensions, the delivery promise and the price. Price is a promise as well: it sets the bar the buyer holds the physical object against.
So the phrase poor quality in a review usually needs translating. Poor compared with what? Most often, compared with the version of the product the buyer built in their head out of your own listing.
The practical consequence is simple. A share of your negative reviews is fixed by copy and photography rather than by manufacturing, and that is the cheapest and fastest part of the work, which is why it goes first.
Four kinds of breakage in one feed
Sort a recent stretch of reviews into four buckets: listing, where the expectation missed; product, where there is a defect or a design decision; logistics, covering packaging, timing and damage; and service, covering support replies, returns and missing parts.
The buckets matter because each has a different owner and a different cost of repair. A listing is corrected by a content editor in a day. A product design is corrected by a factory over a quarter. Held in one undifferentiated task list, neither of them moves.
A single review can land in two buckets at once: arrived creased and runs small is logistics plus listing. Do not chase a clean taxonomy here. What matters is the address the task is sent to, not the label it wears on the way.
Write down separately the things that never show up in reviews but keep showing up in pre-purchase questions. A question is the same signal arriving earlier: it means the listing is missing an answer someone needed.
Repetition beats the star count
A one-off complaint is often not about the product at all. It broke in transit, the colour read differently on a screen, the buyer ordered the wrong variant. Answer it properly, and do not redesign anything around a single voice.
A repeated phrase is a different animal. When ten unconnected people independently write the same sentence about the zip, the strap or the smell, they are describing a property of the object rather than their mood on the day it arrived.
Test for repetition on words rather than on stars. Pull the frequent nouns and verbs out of the review text, such as tight, wobbles, smells and slips, and you have a task list ranked by the people who actually paid for the thing.
Repetition hides inside five-star reviews too. Smaller than I expected, but I love it is still a sizing signal. Reading only the one-star column throws away half of the data you already own and have paid nothing for.
Fix the listing before you fix the product
When the repetition points at a mismatched expectation, the listing is what changes: a measurement in centimetres, a photograph of the item in a hand or beside a familiar object, an honest material line, and an explicit line about what is not in the box.
That edit cuts returns as well as complaints, and a return usually costs more than an order you never received. A buyer who learns in advance that the item runs small will either size up or walk away, and both outcomes beat a return.
The logic runs in the other direction as well. If the listing is already accurate and the same complaint keeps arriving, the problem really is the product or the packaging. The task then goes to the supplier with a direct quotation attached rather than the phrase customers are unhappy.
Doing copy and photography first also saves months of guessing. A listing edit shows up in the next orders, while a design change shows up a production run later, which is a slow and expensive way to learn that you were wrong.
Write the reply for the next buyer
A public reply is not correspondence, it is a component of the listing. Hundreds of people who are still choosing will read it, so it has to work like a short paragraph of description: the fact, the cause, what changed, and what the reader should do now.
The working structure is short. Restate the specific complaint in your own words, give a checkable fact such as where the centimetre measurements sit, say what has been changed or how the case gets resolved, and close with an action and a channel.
Keep the tone level and drop the superlatives. One apology in the right place reads as sincere; an apology in every sentence reads as a script. Shoppers recognise a template instantly, and the template devalues even an answer that is factually right.
Reply to positive reviews too, but selectively. When someone praises a specific detail, a short reply pins that detail down for the next reader. A reply under every single review turns the thread into noise that nobody scans.
What a reply must never contain
Do not argue with the rating. The sentence you used the product incorrectly can be entirely true and will still lose, because the reader is imagining themselves in the reviewer's position rather than in yours.
Do not publish personal data: a full order number, a phone number, an address, or a name the buyer never used themselves. Moving the conversation to a private channel has to happen without spelling out the details on the shop front.
Do not offer money, a discount or a gift in exchange for removing or editing a review. Platform rules forbid it outright, and in several jurisdictions it is a prohibited practice in its own right, as the next section describes. Compensating a real failure is fine; buying a score is not.
Do not paste the same reply under different complaints. A repeated paragraph reads as an autoresponder and discredits every other reply on the listing, including the ones that took genuine work to write.
Bought reviews: platform rules and the law
Every major platform prohibits purchased reviews, self-bought orders placed to unlock a review, and review exchanges. The seller agreement spells out the consequences: suppressed listings, a reset rating, a closed account. That is contract, and it binds you wherever you sell.
Statute sits on top of the contract and differs by country. In the United States the Federal Trade Commission has adopted a rule covering fake and paid-for reviews and the suppression of negative ones. In the European Union the Unfair Commercial Practices Directive, as amended by the Omnibus package, bans submitting or commissioning fake reviews and requires traders to disclose whether they check that reviews come from real buyers. In the United Kingdom the DMCC Act 2024 treats fake reviews as an unfair practice.
Where no dedicated rule exists, manipulation is usually handled through general doctrines: unfair competition, misleading the consumer, and truthfulness requirements in advertising law. Wording and enforcement vary, so confirm the specifics with a lawyer in your own jurisdiction, because we do not provide legal advice.
Incentives are a separate question. Asking for a review in exchange for a gift is allowed in some places and almost nowhere if the reward depends on the score given. The safe frame fits in one line: you may ask for a review, you may not buy a rating.
Why an all-five-star listing converts worse
A listing without a single critical review raises a question rather than confidence. Experienced buyers know that every production run has defects and every product has people it did not suit. An unbroken wall of praise reads as filtered.
The second reason is more practical. Shoppers do not open reviews for confirmation, they open them for boundaries: what exactly is annoying, who it did not fit, what they will have to live with. A specific negative answers that; great, thanks answers nothing at all.
The third reason is your own reply. A handled complaint is the only place on a listing where a buyer can see how the seller behaves when something fails. With no negatives on the page, that demonstration has nowhere to appear.
So the working target is not a 5.0. It is an honest rating whose deductions are explained. A negative review with a clear answer under it removes an objection more effectively than one more anonymous five.
Asking for reviews, and when one can come down
Asking is allowed and worth doing, but ask every buyer the same way and attach no condition to the score. Leave a review if you liked it is a filter, and in several jurisdictions that kind of selection is treated as distorting the picture buyers see.
Timing matters more than wording. Too early and the buyer has not used the thing; too late and they have forgotten it. Anchor the request to the usage cycle rather than the delivery date, and use the platform's own request tool rather than a side channel.
A review comes down because it breaks a rule, not because it is unflattering: abuse, the wrong product, personal data, a competitor's advertising, or content unrelated to the purchase. That goes in as a report with the quotation and a screenshot attached.
Everything else comes down through fixing rather than reporting. If moderation refuses, reply anyway: the reply is visible to buyers and the moderator's decision is not. Never promise a customer that a review will be deleted, because the platform decides that and you do not.
Reviews on your own site: markup and accessibility
Marketplace reviews live on the platform, but the same kind of feedback usually sits on your own site as well. Schema.org defines review and aggregateRating on Product, which is how a machine learns that the page genuinely carries ratings.
The rule is simple: mark up only what the page displays. Describing reviews the page does not show, or importing someone else's marketplace scores, is a policy problem rather than a shortcut, and search engines apply their own extra conditions before showing a rating snippet.
Accessibility is not decoration here. Under WCAG 2.2, stars are non-text content that needs a text equivalent, graphical rating objects carry a contrast requirement, and the controls that filter reviews fall under the minimum target size of 24 by 24 CSS pixels.
The effect is direct. If the score exists only as an image, a screen reader cannot announce it and some parsers cannot read it either. Putting text next to the stars buys accessibility and machine readability in the same edit.
What we do and what it costs
Demand Audit is $50 over one to two working days, with one round of revisions on the findings document. It reviews demand and listings, and the review feed is read in the same pass for what repeats inside it. Access to your seller account is not included in the package.
Listing Refresh is $70 over two to three working days with one round of revisions: the listings are rewritten around what the reviews and the pre-purchase questions exposed. Photography, ad spend and account management are not included.
Marketplace + SEO System is $110 over three to five working days, with two rounds of revisions on listings and copy, and it connects the platform listings to search on your own site. Photography and paid placement are not included. Marketplace + SEO Express is the faster route at $190 within three working days with one round, also without photography or ad spend.
Managed Demand Growth is $200/mo on a monthly cycle with 30 days notice to stop, and it covers continuous work on listings and on the complaints that keep coming back. Advertising spend on the marketplace is not included. The terms of each package are set out on /services/marketplaces.
Practical checklist
- Export a recent stretch of reviews and sort them into listing, product, logistics and service.
- Count the frequent words in the review text instead of watching the average score.
- Start the fixes in the listing: measurements, a photograph for scale, a line about what is not in the box.
- Rewrite replies as if the next buyer is reading them rather than the complainant.
- Strip personal data and every offer made in exchange for editing a review out of your replies.
- Check how you are allowed to ask for reviews against the platform rules and the law where you sell.
Questions and answers
How quickly should we reply to reviews?
Faster is better, but accuracy beats speed. A reply written in an hour without checking the order is easy to get wrong, and the correction has to happen in public. Set an internal turnaround you can actually hold, and split the feed: complaints with specifics go to someone who knows the product, while thank-you notes can be answered less often and more briefly. Silence under a negative review is worse than a late reply, because the reader sees the question and no seller in sight.
What does it cost to work through reviews and rewrite listings?
Demand Audit is $50 over one to two working days with one round of revisions on the findings document; access to your seller account is not included. Listing Refresh is $70 over two to three working days with one round of revisions, and photography, ad spend and account management are not included. If the listings need to be tied to search on your own site, Marketplace + SEO System is $110 over three to five working days with two rounds of revisions on listings and copy.
Can we ask a buyer to change their rating?
Asking someone to add to their review once the problem is solved is usually acceptable, provided the platform allows it and the request is not tied to compensation. Never link money, a discount, a gift or a replacement to editing or deleting a score: platform rules forbid it and several jurisdictions treat it as a prohibited practice. Keep the two actions apart, resolve the problem either way, and leave the rating entirely to the buyer.
What should we do about a review that is factually wrong?
Check the order first, because a share of the wrongness turns out to be a different variant, a different production run or a misread instruction. If the claim genuinely does not hold, answer calmly and concretely: name the checkable detail and offer a channel to resolve it. Report it to the platform only on a formal ground, such as the wrong product or content unrelated to the purchase. Readers almost always score a public argument against the seller.
Should we reply to five-star reviews?
Not to all of them. Reply where the buyer named a specific detail, because confirming that detail serves the next reader better than a thank-you does. Positive reviews are also worth reading as a source of signals about sizing, smell, weight and what is in the box, since praise often carries the same information as a complaint with the sign reversed. Identical replies under every positive review read as automation and cost you trust.