Ecommerce product review statistics and what shoppers use to decide By Checkout Reply, Checkout Reply. Sources checked 2026-11-11. Compare observed review and conversion patterns with shopper survey data, and learn which product reviews can actually answer a checkout question. ## Quick answer Northwestern’s Spiegel Research Center found that purchase likelihood for products with five reviews was 270% higher than for products with none in its retailer analysis. In a separate survey of 8,153 US consumers, PowerReviews reported that 68% read reviews on brand sites and 45% said they would not buy a product with no reviews. Northwestern Medill Spiegel Research Center, How Online Reviews Influence Sales (https://spiegel.medill.northwestern.edu/wp-content/uploads/sites/2/2021/04/Spiegel_Online-Review_eBook_Jun2017_FINAL.pdf) · PowerReviews, The Ever-Growing Power of Reviews (https://www.powerreviews.com/power-of-reviews-2023/) ## Key findings and sources ### Five reviews compared with none Northwestern’s Spiegel Research Center reported 270% higher purchase likelihood for a product with five reviews than with none in its 2017 retailer analysis. This is a relative difference for the studied products. Northwestern Medill Spiegel Research Center, How Online Reviews Influence Sales (https://spiegel.medill.northwestern.edu/wp-content/uploads/sites/2/2021/04/Spiegel_Online-Review_eBook_Jun2017_FINAL.pdf) ### How the relative result changed by price group In Spiegel’s gift retailer analysis, conversion rates were reported to be 190% higher when reviews were displayed for lower-priced items and 380% higher for higher-priced items. The sample included approximately 15.5 million page views across 1,800 products. Northwestern Medill Spiegel Research Center, How Online Reviews Influence Sales (https://spiegel.medill.northwestern.edu/wp-content/uploads/sites/2/2021/04/Spiegel_Online-Review_eBook_Jun2017_FINAL.pdf) ### Where US shoppers reported reading reviews In PowerReviews’ April 2023 survey of 8,153 US consumers, 91% reported reading reviews on retail websites, 70% on search engines and 68% on brand websites. Respondents could use more than one source. PowerReviews, The Ever-Growing Power of Reviews (https://www.powerreviews.com/power-of-reviews-2023/) ### Shoppers who say they will skip products without reviews In the PowerReviews April 2023 survey of 8,153 US consumers, 45% said they would not buy a product with no reviews available. It is a stated response rather than an observed conversion effect. PowerReviews, The Ever-Growing Power of Reviews (https://www.powerreviews.com/power-of-reviews-2023/) ### A written review versus a star rating alone In the PowerReviews survey fielded in April 2023, 56% of shoppers said that a star rating alone was less trustworthy than a star rating accompanied by written review text. The survey covered 8,153 US consumers overall. PowerReviews, The Ever-Growing Power of Reviews (https://www.powerreviews.com/power-of-reviews-2023/) ### Verified purchase and stated likelihood in the historical review analysis The 2017 Spiegel report analyzed about 65,000 verified-buyer reviews and 57,000 anonymous reviews in its reviewer analysis. It reported a 15% relative increase in purchase likelihood when reviews were identified as written by a verified buyer versus an anonymous reviewer. Northwestern Medill Spiegel Research Center, How Online Reviews Influence Sales (https://spiegel.medill.northwestern.edu/wp-content/uploads/sites/2/2021/04/Spiegel_Online-Review_eBook_Jun2017_FINAL.pdf) ### Review-panel engagement changes the denominator In a worked review placement example, ten extra orders divided by 1,000 randomized eligible visits equals ten per 1,000. Dividing by 250 panel opens gives 40 per 1,000, using a denominator the original randomization did not define. Northwestern Medill Spiegel Research Center, How Online Reviews Influence Sales (https://spiegel.medill.northwestern.edu/wp-content/uploads/sites/2/2021/04/Spiegel_Online-Review_eBook_Jun2017_FINAL.pdf) ### The rounded survey results imply substantial cross-site review use PowerReviews’ 2023 review-destination shares imply that roughly 62% to 68% of respondents could have selected both Amazon and brand websites. The bounds use the reported 94% and 68% shares, without assuming the two responses are independent. PowerReviews, The Ever-Growing Power of Reviews (https://www.powerreviews.com/power-of-reviews-2023/) Same multi-response question on reported review destinations in the 2023 survey. The overlap is at least the sum of the two marginals minus 100%, and at most the smaller marginal. ### Most brand-site review readers could also use Amazon Using the rounded 2023 destination shares, the smallest feasible overlap represents about 91.2% of respondents who selected brand websites for reading reviews. The exact observed overlap was not reported in the displayed table. PowerReviews, The Ever-Growing Power of Reviews (https://www.powerreviews.com/power-of-reviews-2023/) Respondents selecting brand websites in the 2023 review-destination question. Divide the lower-bound overlap of 62% of all respondents by the 68% brand-site share. ## Compare the results Figure description: Where surveyed shoppers read product reviews. Exact values are also in the study data. Where surveyed shoppers read product reviews. ## Five reviews is a result from the studied stores, with a specific comparison In the Spiegel analysis, the jump in purchase likelihood was steep early. Products with five reviews had 3.7 times the purchase likelihood of products with no reviews, equivalent to 270% higher. The incremental association flattened after the earliest reviews. The result describes the retailer data and does not tell a merchant what five reviews would add in their own store. That matters when deciding where to work. A high-traffic item with no relevant feedback is a different problem from a bestseller with many reviews that all discuss the same superficial feature. The first needs reliable coverage. The second might need a way for buyers to find reviews about fit, setup or durability quickly. Copying the same star rating into a checkout reply does little to resolve a specific question. Use an inventory of which variant-level questions cause hesitation. Mark whether reviews cover them, whether those reviews are recent and whether the feedback is about the same version. Check the evidence gap on the product that generates the question before expanding reviews everywhere. ## A high-price result does not reveal the absolute increase in orders The study’s gift retailer analysis compared approximately 15.5 million page views for 1,800 products across a year. The reported conversion difference when reviews were displayed was 190% for lower-priced gifts and 380% for higher-priced gifts. These are relative increases and their starting conversion rates may differ. The report does not support a claim that higher-priced products gained twice as many orders. For example, a relative increase of 100% would move a 1% starting conversion rate to 2%. The same relative increase applied to a 3% baseline would take it to 6%. Those are hypothetical values. For the merchant decision, compare review exposure within similar products and traffic conditions, and report the initial rate, absolute order difference and assigned group sizes. The same report describes purchase likelihood peaking below a perfect 5.0 in the categories studied. That is a reason to preserve representative evidence, including relevant negative experiences. It is not a target for altering ratings. An honest review about a narrow fit issue may be more useful to the buyer than a glowing review that omits the condition they care about. ## Where shoppers read and what they want the reply to answer PowerReviews asked 8,153 US consumers in April 2023. Respondents reported reading reviews on retail sites at 91%, brand sites at 68% and search engines at 70%. Those are overlapping answers. The fact that someone reads on a retailer site does not tell you where they purchase or whether the brand’s reply made a sale more likely. A shopper asking whether a bag fits under a particular airline seat needs the measured dimensions of the variant and the relevant airline limits. Reviews may explain how another person used it, but they do not replace specifications. Conversely, feedback from the same variant can be helpful when the question is subjective, such as whether the material feels stiff. State whose experience it was and point to the actual product evidence. A good reply uses the right kind of proof for the question. Keep specifications, policy terms, and verified customer experiences distinct. If the store has no evidence for the variant, say what information can be checked and invite the shopper to supply the one missing detail. ## Measure the answer and the checkout outcome separately A merchant can test whether making relevant review excerpts available changes the number of questions answered accurately, the need for clarification, or checkout completion. A test of the reply should hold the original item and offer constant, record the shopper’s initial question, and audit whether the evidence genuinely matched it. If conversion rises, check returns or customer contacts about the very property the reply addressed. A fast reassurance can make a checkout easier to complete and still create a bad purchase if it overstates what the customer will get. The review data support a better way to decide what to verify. The store’s own experiment answers what the intervention added. ## Verified purchase status answers a different question from the star rating The Spiegel report separately compared anonymous and verified-buyer reviews. It analyzed approximately 57,000 anonymous reviews and 65,000 from verified buyers. That gives a verified share of about 53.28% within those two analyzed groups. The authors reported 15% higher relative purchase likelihood when a review was identified as verified rather than anonymous. Verification can establish that the reviewer bought an item through the tracked route. It cannot establish that the buyer used the same variant, sized the product correctly, or had the same delivery conditions as the person asking for help. The report also describes different paths for prompted and unprompted review submission. The two groups can differ in who chooses to post and the information they see before writing. A useful checkout response uses verification and relevance together. If someone asks about fit, quote a verified buyer only when their experience actually addresses the same model and condition. Describe the limits of that experience and use the product specification for any measurable requirement. Keep relevant negative feedback available so the shopper can judge what the product will and will not do. ## Track who could see the reviews before measuring who opened them A collapsible review section gives you two counts. The first is everyone assigned to a version of the page. The second is the smaller group that opened the reviews. Opening the panel can reflect an existing intention to buy or a question about fit. Comparing openers with everyone else will mix those motivations into the result. Imagine 1,000 eligible visits in each version, with 30 purchases in the review version and 20 in the control. If 250 people opened the panel, dividing the difference by the openers gives 40 extra orders per 1,000 panel impressions. On the randomized denominator, the same difference is 10 extra orders per 1,000 eligible visits. The first number assigns the entire page-level change to the people who opened the panel, a claim the experiment did not establish. Decide in advance how you will count eligible visitors. Track panel opens, questions about the product and review provenance as diagnostics. If an answer resolves a concern without requiring a panel open, that can still benefit the overall experience. Judge the test on the result for the group that could receive the change, and use the opening records to understand why it performed that way. ## A brand-site review audience can also be a marketplace audience PowerReviews reports that 94% of respondents used Amazon to read reviews and 68% used brand websites in its 2023 question about review destinations. Those are separate selections from the same question. Even without the raw joint counts, at least about 62% would have selected both if the rounded shares are taken as written. The highest possible overlap is about 68%. That means roughly 91% to 100% of the brand-site review audience could also use Amazon under those marginals. The study cannot tell a retailer which page the shopper visited first, which reviews they read for the same product or whether either visit produced a sale. It does establish a reason to check consistency between the product evidence shoppers may see in different places. A headline about where reviews are most popular misses that shared audience. ## Use reviews to resolve the constraint the shopper has today If a buyer asks whether a jacket runs small, a star average from a listing with several fits will not settle the question. Find reviews for the right variant where possible, identify the fit pattern and describe the conditions that matter. A shopper comparing another product line still needs the correct size guide and returns terms. Measure whether the answer resolved the sizing question before attributing every ensuing sale to the review content. Choose a small set of high-consideration questions and have a reviewer trace each reply to its evidence. Keep two separate results: whether a reply accurately characterized the review evidence and whether the shopper completed checkout. When the answer depends on a manufacturer specification, use the specification for the claim and reviews for how it felt in use. That keeps a review from being cited as proof of material composition or compatibility. ## Study data Table: Northwestern Medill Spiegel Research Center, How Online Reviews Influence Sales Measure | Value | Context Relative purchase likelihood difference, five reviews versus none | 270% | Relative difference, not conversion points Relative conversion difference for lower-priced gifts when reviews were shown | 190% | Relative conversion increase as reported Relative conversion difference for higher-priced gifts when reviews were shown | 380% | Relative conversion increase as reported Page views in the gift retailer analysis | 15.5 million | Approximate number of page views Users in the gift retailer analysis | 7.8 million | Approximate count of users Products in the gift retailer analysis | 1,800 products | Distinct products SKUs in the lower-priced analysis | 111,460 SKUs | SKUs covered by the lower-priced retailer analysis Categories in the lower-priced analysis | 22 categories | CPG categories Anonymous reviews in the separate reviewer analysis | 57,000 reviews | Anonymous review count Verified buyer reviews in the separate reviewer analysis | 65,000 reviews | Verified buyer review count Relative purchase likelihood difference for a verified-buyer review | 15% | Relative purchase likelihood as reported Table: PowerReviews, The Ever-Growing Power of Reviews Measure | Value | Context Read reviews on Amazon | 94% | Stated preference or shopping behavior as published Read reviews on retail websites | 91% | Stated preference or shopping behavior as published Read reviews on search engines | 70% | Stated preference or shopping behavior as published Read reviews on brand websites | 68% | Stated preference or shopping behavior as published Seek out sites with ratings and reviews | 77% | Stated preference or shopping behavior as published Use reviews to learn about new-to-them products | 74% | Stated preference or shopping behavior as published Say reviews affect whether they purchase | 93% | Stated preference or shopping behavior as published Say they will not buy a product with no reviews | 45% | Stated preference or shopping behavior as published Trust a star rating alone less than a rating with a written review | 56% | Stated preference or shopping behavior as published ## What changes when you compare the figures Table: Calculations from the published values above. Comparison | Result | Calculation Five-review purchase likelihood relative to no reviews | 3.7 times | = 1 + 270 / 100 Relative conversion difference between the two gift price groups | 190 percentage points | = 380 − 190 Survey share reading at retail sites minus brand sites | 23 percentage points | = 91 − 68 Survey share reading at search engines minus brand sites | 2 percentage points | = 70 − 68 Verified share of the two review groups analyzed | 53.28% | = 65000 / (65000 + 57000) * 100 Difference in number of verified and anonymous reviews analyzed | 8,000 reviews | = 65000 - 57000 Worked example, baseline purchases per 1,000 eligible visits | 20 purchases | = 1000 * 0.02 Worked example, review panel impression-only denominator | 40 additional orders per 1,000 panel impressions | = (30 - 20) / 250 * 1000 Worked example, same numerator on randomized eligibility | 10 additional orders per 1,000 eligible visits | = (30 - 20) / 1000 * 1000 Minimum Amazon and brand-site overlap from rounded shares | 62% | = 94 + 68 - 100 Maximum overlap from smaller marginal | 68% | = 68 Lower-bound Amazon overlap among brand-site review readers | 91.176% | = 62 / 68 * 100 ## Methodology and sources We transcribed the Spiegel 2017 report’s separately named retailer, gift retailer and reviewer analyses, and PowerReviews’ April 2023 US consumer survey. Relative purchase and conversion differences are reproduced as published. Survey questions are shown on their own stated bases and are not pooled with observed sales. Ratios and percentage-point gaps are arithmetic on rounded source values. The chart shows places respondents said they read reviews, with overlapping choices. [Download the source data (CSV)](decision-table.csv) [Download the calculations (CSV)](calculations.csv) [Download the citation notes (CSV)](citation-notes.csv) [Download the evidence behind the new findings (CSV)](finding-evidence.csv) ## Related reading AI shopping recommendation accuracy statistics (https://checkoutreply.com/guides/ai-shopping-recommendation-accuracy-statistics/) AI shopping price and availability error statistics (https://checkoutreply.com/guides/ai-shopping-price-availability-statistics/) Product search statistics on exact matches and substitutes (https://checkoutreply.com/guides/ecommerce-product-substitute-statistics/) What to text when a shopper leaves checkout (https://checkoutreply.com/guides/first-imessage-abandoned-checkout/) How to answer “Will it get here by Friday?” before they buy (https://checkoutreply.com/guides/delivery-question-before-checkout/) Browse checkout recovery research (https://checkoutreply.com/research/) ## Sources and definitions Northwestern Medill Spiegel Research Center, How Online Reviews Influence Sales (https://spiegel.medill.northwestern.edu/wp-content/uploads/sites/2/2021/04/Spiegel_Online-Review_eBook_Jun2017_FINAL.pdf) PowerReviews, The Ever-Growing Power of Reviews (https://www.powerreviews.com/power-of-reviews-2023/) The source data includes the population, period, denominator and limitations for every figure. Checkout Reply · checkoutreply.com