Product Recommendation Revenue Statistics
A sourced roundup of product recommendation statistics. Visits where the shopper clicked a recommendation make up 7% of traffic but 26% of revenue (Salesforce), the average store attributes 12% of sales to recommendations (Barilliance), and a typical widget gets a 9% click rate (Wisepops).
Product recommendation revenue statistics
Sourced figures on how much revenue, conversion and order value online stores get from product recommendation widgets.
Product recommendation revenue statistics measure what recommendation widgets ('You may also like', 'Frequently bought together', 'Recently viewed', cart cross-sells) contribute to a store's results. The main metrics are the share of revenue and orders from sessions where a shopper clicked a recommendation, the widget's click rate, the conversion rate of shoppers who click, and the effect on average order value.
Most published figures come from recommendation vendors' own customer data. That makes them useful for setting expectations, but they measure attributed revenue, not incremental revenue. Shoppers who click recommendations are usually high-intent to begin with. Read every number below with that bias in mind, and check it against a holdout test on your own store.
Key statistics
Visits where the shopper clicked a product recommendation made up 7% of visits but drove 24% of orders and 26% of revenue (Salesforce, study of 250M visits published in 2017). On average, Barilliance customers attribute 12% of sales to recommendations. The best performers reach up to 31% (Barilliance, Q4 2014 study). 35% of what consumers buy on Amazon comes from product recommendations (McKinsey, 2013). A typical store gets about a 9% click rate on its product recommendations. The low end is around 6.5% and the best performers are above 20% (Wisepops, 2026). Roughly 4% of recommendation clicks turn into an order (Wisepops, 2026). Session conversion rose 288% after a single recommendation engagement, from a 1.02% baseline (Barilliance). Above-the-fold recommendations get more than twice the click-through and 1.7x the sales of below-the-fold ones (Barilliance, 2014). Shoppers who engage with AI-powered recommendations have a 26% higher average order value (Salesforce). 49% of shoppers have bought something they didn't plan to after seeing a personalized recommendation (Salesforce Connected Shoppers Report).
Most of these figures trace back to just two large datasets, from Salesforce and Barilliance, and both are several years old. They still get quoted because nobody has published a larger, independent study since. Below, each figure carries its publisher and year so you can cite it accurately and judge how well it applies to your store.
How much revenue comes from product recommendations
Share of orders and revenue linked to product recommendations
| Metric | Figure | Scope | Source & year |
|---|---|---|---|
| Share of visits with a recommendation click | 7% | 250M visits on ecommerce sites | Salesforce, 2017 |
| Share of orders from those visits | 24% | Same dataset | Salesforce, 2017 |
| Share of revenue from those visits | 26% | Same dataset | Salesforce, 2017 |
| Comparison: site-search visits, share of visits / share of revenue | 10% / 25% | Same dataset | Salesforce, 2017 |
| Average share of sales attributed to recommendations | 12% | 300 randomly selected Barilliance customers | Barilliance |
| Attributed revenue share, best-performing sites | Up to 31% | Barilliance customer base | Barilliance, Q4 2014 |
| Share of Amazon purchases from recommendations | 35% | Amazon (single retailer) | McKinsey, 2013 |
The Salesforce split, 7% of visits driving 26% of revenue, works out to about 3.7x more revenue per visit for recommendation clickers than their share of traffic would suggest. Site search shows almost the same pattern: 10% of visits produce 25% of revenue. Both tools attract shoppers who are already looking for something specific.
For a store doing €1M-€15M a year, the useful benchmark is Barilliance's 12% average, not the 31% ceiling. Stores near the top usually have deep catalogs, a lot of repeat purchasing (beauty refills, apparel basics) and recommendations on every key template. A store with 40 SKUs and mostly one-off purchases will sit well below 12%, and that's normal.
The Amazon figure of 35% is the most quoted number in this category, and the hardest to apply to your own store. It describes one retailer with an enormous catalog and years of purchase data. Use it as a sign that recommendations can matter a lot, not as a target.
Click-through and conversion rates for recommendation widgets
Product recommendation click rate: benchmarks and named stores
Wisepops' 2026 benchmark puts the typical recommendation click rate at around 9%, with the low end near 6.5% and the best performers above 20%. Both fashion case studies in the dataset, Pierre Hardy at 11.6% and Maison Lejaby at 15%, got most of their clicks on mobile. A widget designed for desktop leaves engagement on the table.
Clicks only matter if they lead to orders. Wisepops finds roughly 4% of recommendation clicks become an order, so judge a widget on orders per click as well as click rate. Barilliance reports session conversion rising 288% after one recommendation engagement, from 1.02% to about 3.96%, and says engaged shoppers are 5.5x more likely to complete a purchase.
Salesforce's data adds a device split. Desktop shoppers who clicked a recommendation converted at 9.8%, stayed for about 15 minutes on average, and 24% of them created a basket. Mobile shoppers who clicked recommendations completed orders at a higher rate than desktop shoppers (25% on mobile). Later Salesforce Shopping Index coverage puts recommendation clickers' conversion at about 4.5x that of non-clickers.
Placement, personalization and order value effects
Recommendation performance by placement and format, and the effect on order value
| Placement / effect | Figure | What it measures | Source & year |
|---|---|---|---|
| Above-the-fold vs below-the-fold widget | More than 2x CTR, 1.7x sales | Relative performance | Barilliance, 2014 |
| Product page widget below the description | 3-5% CTR | Typical click-through range | GrowthSuite, 2026 |
| Product page sidebar widget (desktop) | 1-2% CTR | Typical click-through range | GrowthSuite, 2026 |
| Recommendation popups, all stores / Shopify | 4.65% / 8.11% | Average conversion rate | Wisepops, 2025 |
| Targeted bundle popup (case study) | 22.9% CTR, 14.7% of clickers ordered | Click-through, then click-to-order | Wisepops, 2025 |
| Personalized vs generic best-seller recommendations | 2.2x as effective | Relative effectiveness | Barilliance |
| AOV, shoppers engaging with AI recommendations | +26% | Lift vs non-engaged shoppers | Salesforce |
| AOV, shoppers who click a recommendation | About +10% | Spend per order vs non-clickers | Salesforce via Hello Retail, 2026 |
| AOV, sessions with no recommendation engagement | $44.41 | Baseline order value | Barilliance |
| Shoppers who made an unplanned purchase after a personalized recommendation | 49% | Self-reported behaviour | Salesforce Connected Shoppers Report |
Placement has the most consistent effect. Barilliance found above-the-fold widgets more than doubled click-through and produced 1.7x the sales of below-the-fold ones. Vendor guidance puts a product page widget below the description at 3-5% CTR and a desktop sidebar at only 1-2%. Personalized widgets were 2.2x as effective as generic best-seller rows.
The order-value figures don't agree: Salesforce material cites a 26% AOV lift for engaged shoppers, while summaries of other Salesforce data put it nearer 10%. The difference comes down to how 'engaged' is defined and which retailers are in the sample. For planning, a 5-15% AOV lift among clickers is a cautious working range for a mid-size apparel or beauty store, not a sourced figure.
Attributed revenue isn't incremental revenue
Every 'share of revenue' figure on this page counts orders from sessions where a shopper clicked a recommendation. Many of those shoppers would have bought anyway, because people already comparing products click more. The 26% Salesforce figure and Barilliance's 12% are attribution numbers. The only way to measure the extra revenue a widget actually creates is a holdout test: hide the widget from a random 10-20% of traffic and compare revenue per visitor.
How to use these numbers
Start by comparing your own split with Salesforce's. In GA4 or your recommendation app, find the share of sessions with a recommendation click and the share of revenue from those sessions. If 7% of sessions bring in well under 20% of revenue, the widget is probably showing the wrong products or sitting in the wrong place. That's a PDP Optimization problem, not a reason to buy a new engine.
Next, test the levers the data rates highest: move the widget above the fold on mobile product pages, swap generic best-sellers for personalized or 'frequently bought together' logic, and add a complementary cross-sell to the cart page. Run each change against a holdout and track revenue per visitor, not click rate. Watch page weight too: a recommendation script that adds 300ms to product page load can cancel out the gain.
Finally, give the test enough time. With a 9% click rate and roughly 4% of clicks turning into orders, a store with 50,000 monthly product page sessions gets only about 180 recommendation-driven orders a month. Plan for a test window of three to six weeks, and carry over your historical baseline instead of starting from zero.
Product recommendation statistics: FAQ
Published figures range from about 12% to 31% of revenue. Barilliance reports a 12% average across its customers, with best performers reaching up to 31%. Salesforce found that visits with a recommendation click made up 7% of traffic but 26% of revenue. All of these are attributed figures, not proven incremental lift.
Wisepops' 2026 benchmark puts the typical click rate at around 9%, with the low end near 6.5% and top performers above 20%. Click rate depends heavily on placement: a desktop sidebar widget often gets only 1-2%, while a well-placed product page widget does much better. Judge success on orders per click, not clicks alone.
Yes, in most published data, though the size of the effect varies. Salesforce material cites a 26% higher AOV for shoppers who engage with AI recommendations, while other Salesforce summaries put the lift nearer 10%. 'Frequently bought together' and cart cross-sells usually affect AOV the most.
Prioritize the product page above the fold on mobile, the cart or cart drawer for complementary items, and the homepage for returning visitors. Barilliance found above-the-fold widgets produced 1.7x the sales of below-the-fold ones. Treat placement as part of PDP Optimization and test one location at a time.
Only as an upper bound. McKinsey's 2013 estimate that 35% of Amazon purchases come from recommendations reflects a huge catalog and years of purchase data. A store with a few hundred SKUs should compare itself with the 12% average from Barilliance instead.
On average, yes. Barilliance found personalized recommendations were 2.2x as effective as generic best-seller rows. Best-sellers still work for first-time visitors with no browsing history, so many stores show best-sellers to new visitors and personalized rows to returning ones.
Much better, partly because they were already high-intent. Barilliance reports engaged shoppers are 5.5x more likely to complete a purchase, and Salesforce Shopping Index coverage puts it at about 4.5x. In Salesforce's earlier data, desktop clickers converted at 9.8%.
Run a holdout test: hide the widget from a random share of visitors, usually 10-20%, and compare revenue per visitor between the two groups. Attribution reports in recommendation apps credit every order that included a click, which overstates the real effect. Let the test run until the revenue-per-visitor difference is statistically significant.
They can. Many apps load their own JavaScript plus extra product images, which can add hundreds of milliseconds on mobile. Check load times before and after installing, lazy-load widgets below the fold, and compare any speed loss with the revenue the holdout test shows.
They use the same purchase and browsing data but reach shoppers in different places. Email blocks such as 'recommended for you' or replenishment reminders bring shoppers back, while on-site widgets work during the session. Measure each separately so a returning email click isn't counted twice as on-site recommendation revenue.
Test ideas before you ship them
Run unlimited A/B tests, attach hypotheses to outcomes, and build a searchable archive of what works — and what doesn't.