TechmadeazyFor twenty years reviews did one job. They reassured a human who was already on the page and nearly ready to buy. Star average near the top, a few testimonials, job done.
That job has not gone away. But reviews now do a second one, and it happens much earlier, somewhere the brand cannot see. When an assistant is deciding which three products to name, reviews are part of the evidence it reads. Not the average. The sentences.
That changes which reviews are valuable and what to do about the bad ones.
The star rating is the least useful part
A 4.6 tells an assistant almost nothing it can put in an answer. Every serious product in a category has something between 4.3 and 4.8.
What gets used is the content. Who bought it, what for, whether it worked, what went wrong. A review saying "perfect for my 4 year old's lunchbox, dishwasher has not warped it in six months" carries information that no amount of stars does, because it maps onto a question somebody is going to ask.
This is why review volume alone is a weak target. Two hundred reviews that all say "great product, fast delivery" give an assistant nothing to recommend the product for.
Recurring complaints are a content brief
Open the reviews, sort by lowest rating, and read a hundred of them in one sitting. It is a miserable afternoon and it is the highest value research available in most quarters.
The thing to look for is the complaint that appears again and again. Almost every product has one. The chew is too big for small breeds. The oil smells strong and takes two washes to come out. The pan needs more seasoning than the instructions suggest.
Now look at the product page. In nearly every case, it says nothing about it.
That silence does damage twice over. The buyer who cares finds out from a stranger instead of from the brand, and the assistant, reading a page that never addresses the most discussed issue with the product, has no answer to give when somebody asks.
Answer the objection on the page
The instinct is to bury it. The better move is to state it and handle it.
If the chew is large, say so, say which breeds it suits, and explain how to split it. If the oil is strong smelling because of the actual active ingredient, say that, and say how to wash it out. If the pan needs extra seasoning, give the method.
Two things happen. Buyers who would have returned the product do not buy it, which improves the reviews from that day forward. And assistants start describing the product accurately, including its trade-offs, which is precisely the kind of answer people trust.
Honest handling of a known objection does not reduce sales. It reliably reduces returns.
Recency, specificity, volume, in that order
If effort has to be prioritised:
Recency first. A product that was reformulated, repackaged or improved eighteen months ago is still being described by reviews from before the change. Current reviews are how that gets overwritten.
Specificity second. Reviews that name a use case are worth several that do not.
Volume third. It matters, but only once the first two are working. Two hundred vague reviews is worse than forty detailed recent ones.
Ask better and you get better
Most post-purchase review requests are a star widget and the word "review". They produce exactly what they asked for.
Ask a question instead. One question, in the email, before the widget. "What were you using this for, and did it do the job?" For a considered product, ask again at thirty days rather than at three, because the useful observations are not available on day three.
For marketplace sellers, the review programs available at launch are worth using properly, particularly for a new variant. A new listing with no reviews is invisible to an assistant regardless of how good the product is.
And reply to reviews, including the good ones, in a way that adds information rather than thanks. "Glad it worked. For anyone reading, the small size fits a standard 500ml bottle" is a sentence an assistant can use. "Thank you for your feedback!" is not.
What does not work
Review gating, incentivised batches and anything that produces a wall of identical five-star sentences. It lifts the average, says nothing, and reads as suspicious to a system comparing one review body against a competitor's.
Also, responding to a recurring complaint with a template apology while changing nothing. The complaint keeps arriving, the thread of identical apologies becomes its own signal, and the effort buys nothing.
Where to start
- Read the last hundred negative reviews and write down the top three recurring themes.
- Add a plain paragraph to the product page addressing each one.
- Change the review request email to ask one open question.
- For any product that was reformulated or improved, push specifically for new reviews that describe the current version.
