Review requests: how to ask so people actually answer

A product page with thirty honest reviews outperforms the same page with none by a margin no amount of copywriting will close. Reviews also answer the precise questions buyers have, in buyers' own words, which happens to be exactly the unique text that makes a product page rank — the one thing a competitor pasting the same manufacturer description cannot copy.

Getting them is a flow, and it contains four decisions. Each one moves the response rate considerably more than the email design does, and three of the four are about restraint rather than effort. There is also a fifth consideration that has become a legal one rather than a commercial one, and it is the part most shops are quietly getting wrong.

When to ask, and why most shops ask too early

The single most common failure in this flow is asking before the customer has the product, and it is always a timing bug rather than a writing problem.

1. Ask after delivery, not after payment. A request that lands while the parcel is still in transit asks somebody to rate something they have not seen. Customers remember it, and it is the kind of detail that makes a shop feel careless in a way unrelated to the actual service.
2. Wait long enough for them to have used it. Ten to fourteen days after delivery suits most goods. The right number is however long it takes to form an opinion worth writing down.
3. Longer for anything whose value takes time to show. A skincare product, a tool, a mattress, a subscription — three to four weeks, because a review written on day two is a review of the packaging.
4. Shorter for consumables used immediately. A coffee or a candle can be reviewed within a few days and the memory is sharpest then.

Deriving the delivery date is the part that requires real data. Where the shop records fulfilment, use it. Where it does not, add a realistic shipping time per destination country rather than one global number — the same flat seven days that works domestically will fire three days early on every cross-border order, which is precisely the subset of customers most likely to have had a slow delivery and least likely to be charitable about being asked too soon.

One suppression that belongs here and is routinely missed: anybody whose order was refunded, cancelled, or whose parcel came back. Asking somebody who returned the goods how they are enjoying them is the worst possible version of this email, and it is entirely preventable by checking the order state at send time rather than at flow entry.

Ask one specific question instead of five stars

"How did we do?" is a question about you. "Did the size run true?" is a question about them, and it reliably gets several times the response.

The difference is cognitive rather than motivational. A generic five-star widget asks somebody to summarise an entire experience into a single number, which requires them to decide what they think before they can answer. A specific question asks them to report something they already know, which takes no effort at all.

• Take the question from your support inbox. Whatever people ask before buying is exactly what you want answered in reviews, because the next buyer will ask it too. Sizing, fit, strength, how long it lasts, whether it suits sensitive skin, whether it fits the older model.
• One question, one click to begin. A star rating rendered in the email itself, which lands on a page with that rating already applied, is the pattern that works. Every step between the email and a submitted review loses a share of the responses.
• Make the written part optional. A rating alone is worth having. Demanding prose before the form will submit loses most of the people who would have given you a number and a sentence if asked gently.
• Name the product and show its picture. Somebody who ordered three things needs to know which one you are asking about, and a review attached to the wrong product is worse than no review.
• Ask once, and once more at most. A single reminder a week later is reasonable. A third request is a complaint waiting to be filed.

What to do with the critical ones

The instinct to filter negative reviews is commercially wrong and, implemented in certain common ways, unlawful. Both halves of that are worth taking seriously.

Commercially, a page of nothing but five stars reads as fake, and experienced shoppers actively seek out the critical reviews before anything else. A three-star review saying "lovely, but runs small" does more good than harm: it prevents a return, it sets an expectation the product then meets, and it makes every positive review on the page more credible by demonstrating that you publish what you are sent.

Legally, the pattern of routing low scores to a private feedback form while sending high scores to the public page is the specific practice EU consumer rules have been tightened to address. Presenting a filtered subset as if it were all reviews is a misleading practice, and "we only publish verified happy customers" is not a defence if the filtering is on sentiment rather than verification.

1. Publish them, including the ones that sting.
2. Reply once, plainly, without defensiveness. The reply is read by every future buyer and does more work than the review does damage. "You are right, it runs small — we have added a sizing note" converts a complaint into evidence that you listen.
3. Never suppress by rating. Moderate for abuse, irrelevance and personal data; never for score.
4. Do not incentivise positive reviews. If you offer anything in exchange, it must be for any review regardless of content, and the incentive has to be disclosed on the review.

What the rules now require you to say

EU consumer law has tightened specifically around reviews, and three obligations are worth building into the flow rather than discovering in a complaint.

• Say whether you verify that reviews come from actual purchasers, and describe how. The obligation is not to verify — it is to be accurate about whether you do. A shop that publishes unverified reviews and says so is fine. A shop that implies verification it does not perform is not.
• Do not publish fake reviews or commission them. This explicitly covers buying them, generating them, and writing them internally, and it is listed among the practices considered unfair in all circumstances.
• Do not present a filtered set as complete. If you moderate, state the basis.
• Keep the marketing consent question separate. A review request sent to a customer about their own order sits comfortably within the existing relationship. Quietly adding everyone who responds to a newsletter list does not, and it is the kind of thing that turns a good flow into a complaint about a different flow entirely.

None of this requires legal review to implement. It requires one sentence on the reviews section of your product pages and a decision not to filter by score — both of which you can do this afternoon.

Getting the value back out

Collected reviews that sit in a widget are half-used. They are an asset with at least four uses, and three of them are free.

1. Product page copy. The phrases that recur across reviews are the words your buyers actually use, which are frequently not the words in your description. Feed them back into the title and the opening paragraph. This is the cheapest SEO work available and it comes with evidence attached.
2. Structured data, but only where the reviews are genuinely rendered on that page and are for that specific product. Marking up ratings that live elsewhere, or that belong to the brand rather than the item, is the most common cause of a structured-data manual action.
3. Email and ads. A quoted review beside a product consistently outperforms a written benefit, because it is somebody else making the claim. Rotate them rather than picking one favourite.
4. Internal product research. Read the two- and three-star reviews monthly. They are the most specific and least filtered product feedback a shop will ever receive, they arrive unprompted, and they routinely identify a sizing or durability problem months before the returns data makes it obvious.

That last one has no marketing value and is usually the most valuable. A recurring complaint in reviews is a return rate you have not yet paid for.

Measuring the flow

Response rate is the obvious metric and it is the right one to optimise, but on its own it will mislead you — it is trivially raised by asking more often, which costs you elsewhere.

• Response rate per request sent, split by how long after delivery you asked. This is the number that tells you whether your timing is right, and it is worth re-deriving once a year.
• Reviews per product, ranked. The products with revenue and no reviews are your priority list, and they are usually not the products you would have guessed.
• Conversion rate on product pages above and below a review-count threshold. Crude, confounded, and still the clearest internal argument for spending time on this.
• Complaints and unsubscribes on review requests, which should be near zero. Anything else means you are asking too often, too early, or of people who returned the item.
• Average rating over time, watched for drift rather than level. A slow decline is a product or supplier problem announcing itself quietly.

What not to measure: the average rating as a target. The moment it becomes a number somebody is accountable for, the pressure to filter returns, and that is the practice this entire article is arguing against.