Ask everyone, automatically, or you will forget
The reason new stores have no reviews is almost never that customers refuse. It is that nobody asked, or asked once by hand and stopped.
Set up a review request as an automatic flow triggered by fulfilment, so every order produces a request without anyone remembering. This is a one time setup and it is the single highest return action on this list.
Make the first interaction happen inside the email. A star rating the customer can click directly in the message, which then takes them to a page with the rating already selected, removes the biggest drop-off point. Asking someone to click through, find the product and start from a blank form loses most of them.
Send one reminder if there is no response, and then stop. A third and fourth request annoys people and produces the reviews you least want.
And start on day one. Reviews accumulate slowly and they compound, so a store that begins asking with its first order will have a useful body of them within months, while one that waits until it feels established starts from zero much later.
Time the request to when the product can actually be judged
Sending the request the day after delivery is the default and it is wrong for most products. Ask before the customer has formed an opinion and you get a review about the packaging.
Work out how long it takes to have an experience worth describing, and set the delay per product type.
Something consumed quickly, such as food or a small accessory, can be asked about within days. Clothing needs a wear and usually a wash. Skincare and supplements need weeks before anyone can say anything meaningful. Furniture and appliances need enough time to be assembled and lived with. Tools need a job done with them.
If your platform allows different delays by product type or collection, use it. If it does not, choose a delay that suits your main category rather than an average that suits nothing.
Also wait until delivery is confirmed rather than counting from the order date, because a request that arrives before the parcel does reads as careless and can generate a poor rating about shipping.
And suppress the request entirely for orders with an open support ticket or a return in progress. Asking those customers for a public rating at that moment is a predictable mistake.
Ask questions that produce reviews the next buyer can use
A five star rating with no words persuades nobody and helps nobody choose. What converts is a review that answers the question the next shopper has.
So ask specific questions alongside the rating, and keep it to two or three. Which size did you order and how did it fit. What did you use it for. What surprised you. How does it compare with what you used before.
Structured fields are worth adding where they apply. Height and size ordered for apparel. Skin type for beauty. Room size for furniture. Vehicle or model for parts. These become filters and summaries that help every subsequent shopper, and they reduce returns because the next person can find someone like themselves.
Make the free text box small rather than large. A big empty box asks for an essay and gets abandoned. A small one with a specific prompt gets filled.
And publish the details alongside the review. A comment saying the fit runs small, from someone who states their usual size, is worth more than fifty entries saying great product.
That is also the content that helps your product pages get found and quoted, because it is specific.
Ask for photos separately, and put reviews where the decision happens
Photos from customers do more for conversion than text, and they are the thing people are least likely to volunteer. Ask for them as a distinct step after the written review rather than as a field in the same form, and say why they help.
Where reviews appear matters as much as how many you have. A block of reviews at the bottom of the page is seen by a fraction of visitors on a phone.
Put the rating summary near the price, put relevant review snippets next to the decision they inform, and let people filter by the structured fields you collected. A sizing comment beside the size selector changes behaviour. The same comment three screens down does not.
Surface reviews on collection pages too, so the rating is visible while someone is choosing which product to open.
And use reviews elsewhere. Quotes in email, in ads where permitted, and on the pages that answer buying questions. A customer explaining in their own words why the product suited them is more persuasive than anything written in house, and you already have it.
What not to do, and how to handle the bad ones
Do not buy reviews, write your own, or ask friends and family to post as customers. Beyond being against the terms of every review platform, presenting fake reviews as genuine is against consumer protection law in many countries, and the reputational damage when it is discovered is far larger than the benefit.
Do not offer payment or a discount in exchange for a positive review. Rules differ on incentives, but the consistent principle is that any incentive must not be conditional on the review being favourable, and it must be disclosed. Read the rules for your platform and your country before running anything like this.
Do not hide or delete negative reviews. A page of only perfect scores reads as untrustworthy, and a small number of critical reviews raises the credibility of the rest.
Do reply to the poor ones, in public, briefly and without defensiveness. Acknowledge the problem, say what you have done, and offer to sort it out. Prospective buyers read the replies more carefully than the complaints, and a good reply converts better than the complaint costs.
And feed the criticism back into the product page. A repeated complaint is a description problem you can fix.
Worked example, not a result
A review flow for a store's first year
- Trigger
- delivery confirmed
- Delay, apparel
- 14 days
- Delay, skincare
- 30 days
- Reminders
- one
- Rating clickable in the email
- yes
- Structured questions
- size ordered, usual size, use case
- Photo request
- separate step after submitting
Worked example, not a result