Do Product Recommendations Increase Shopify Conversion Rate?

Do product recommendations increase Shopify conversion rate? Barely, and almost never in the way the app store listing promises. They move average order value. That is a different lever, worth real money, and it is not the one you were sold.

I believed the big number for longer than I want to admit. A dashboard tells you the recommendation widget "influenced" a quarter of your revenue, and it feels like you found free money. Then you turn the widget off for two weeks and revenue does not move.

Here's what is going on underneath, why Baymard found 58% of major sites build cross-sells wrong, and the five tests that separate a widget that earns its slot from one that just sits there looking busy.

What do the famous recommendation statistics actually say?

Four numbers get quoted in every pitch deck in this category. All four are real. Three of them are being read backwards.

The statistic	What it measures	What it does not prove
Recommendations drive up to 31% of site revenue	Sessions that touched a recommendation somewhere	That the revenue would have vanished without it
Clickers convert 5.5x higher than non-clickers	A comparison of two self-selected groups	That clicking caused the conversion
Conversion rises 288% after a single interaction	The same self-selection, expressed as a lift	That the widget created buying intent
Personalized suggestions raise view to purchase 15 to 30%	Tested personalization against a static control	Much, honestly. This one is the useful number

The industry statistics roundups put the revenue contribution at up to 31%, with shoppers who click recommendations generating about 26% of revenue and 24% of orders. Those are attribution figures. They describe traffic that passed through a widget on the way to a purchase.

Think about who clicks a recommendation carousel. Somebody browsing hard. Somebody comparing sizes. Somebody who has already decided they want something from your store and is now working out which thing. Of course they convert at five times the rate of a person who bounced off your page in nine seconds. They were a different person before the widget ever loaded.

Measuring clickers against non clickers tells you who your buyers are. It does not tell you what your widget did.

The one number in that table worth building on is the last one. When personalized suggestions get tested against a static control, the view to purchase improvement lands somewhere around 15 to 30%. Smaller than 288%. Also real.

So what is the honest answer?

Three things are true at once, and holding all three is the whole game.

One. Recommendations reliably raise average order value. Basket size grows, order value grows, and the effect shows up in controlled tests, not only in attribution dashboards. In one commonly cited case, average order value rose 17% and basket size grew 24%.

Two. Recommendations move conversion rate a little, and only when they solve a mismatch. A visitor landed on the wrong size, wrong price tier, or wrong style. A well placed alternative rescues that visit. That is a genuine conversion save.

Three. Recommendations can lower conversion rate. Put a carousel of eleven other products directly under the buy button of a page the visitor is already sold on and you have given them a reason to start over. I have watched that exact change cost a store money.

Here's the math on a store like this. Conversion rate 1.8%, average order value $70. That means revenue per visitor is $1.26. On 10,000 monthly visitors, that's $12,600.

Now suppose the supplements do their job and one visitor in nine adds the accessory. Conversion rate 2.0%, average order value $82. Revenue per visitor: $1.64. Same 10,000 visitors: $16,400.

That's $3,800 a month, and notice where it came from. The conversion rate moved two tenths of a point. The average order value moved twelve dollars. If you had been grading the widget on conversion rate alone, you would have called it a failure and uninstalled the thing that was paying you.

Why do 58% of stores build this wrong?

Because there are two completely different tools here and most stores ship one blended mush.

Baymard's product page research splits suggestions into two families. Alternatives help someone find the right product. Supplements help them finish the package. Their finding: 58% of major ecommerce sites either offer only one type, or dump both into the same element. Only 42% get it right. In the same body of work, up to 62% of leading sites rate mediocre or worse on product page usability overall, so this is not a small store problem.

	Alternatives	Supplements
The job	Find the right one	Finish the package
Example	Other sizes, a cheaper tier, a similar style	The filter, the case, the mat, the refill
Moves	Conversion rate	Average order value
Right place	Lower on the page, after the pitch has had its chance	Beside the add to cart action, and in the cart
Right label	Customers also considered	Customers who bought this also bought
Failure mode	Distracts a decided buyer	Feels like a shakedown at checkout

When you mix them, both jobs fail. The shopper who wanted a smaller size scrolls past a cable and a carrying case. The shopper who was ready to buy sees four other chairs and remembers they had not finished comparing.

Split them. Different sections, different labels, different places on the page. That single change is usually worth more than switching recommendation apps.

Where should recommendations go on a Shopify product page?

Below the buy box. Always. I will fight about this one.

Above the fold you have one job, which is to close the product this visitor chose. Everything competing for that space is a tax. The sequence that works on nearly every store I have taken apart:

1. The buy box. Price, the promise, the risk reversal, add to cart. Nothing else.
2. The pitch. Why this product, the mechanism, the proof, the objections. This is where a page earns its money, and it is the part most stores skip in favor of widgets.
3. Supplements, right after the second add to cart. The buyer is warm and has just committed. This is when "people who bought this also bought the wall mount" reads as helpful.
4. Alternatives, near the bottom. By now, a visitor still scrolling is telling you this product missed. Catch them before the back button does.
5. The cart. One supplement. One. Under $40, related, one tap to add.

The exception is a catalog with heavy variant confusion, like footwear or anything sized. There, an alternatives strip goes higher, because the mismatch is the most likely reason for the loss. Same logic I used in the comparison table breakdown: the fix belongs wherever the doubt lives, not wherever the theme put a slot.

What should the carousel say?

The label is doing more work than the algorithm.

Baymard's test subjects responded poorly to suggestions when the basis for them was unclear. People want to know why they are being shown this. "You may also like" answers nothing. It is the interface equivalent of a shrug.

Compare:

• You may also like → a shrug
• Customers who bought this also bought → a reason, and social proof
• Complete the setup → a reason, with a job to do
• Other sizes in this style → a reason, and a rescue
• Cheaper alternatives in this range → a reason, and it keeps a price sensitive visitor on your site instead of on somebody else's

Five words of label copy routinely outperforms a machine learning upgrade. That is an annoying sentence to write as somebody who builds with AI daily, and it is still true.

Nobody clicks a carousel because the algorithm is good. They click because the header gave them a reason to.

Which catalogs get paid, and which ones get nothing?

The answer changes a lot by what you sell, and this is the part the vendor case studies flatten into one average.

Catalog type	What recommendations do	The move that pays
Consumables (coffee, supplements, skincare)	Very little on conversion rate, a lot on order size	Quantity and refill supplements, not other flavors
Apparel and footwear	The biggest conversion rate effect of any category	Alternatives high on the page, because size and fit mismatch is the main loss
Electronics and accessories	Strong on average order value, weak on conversion rate	Compatibility supplements, stated as compatibility and not as "you may also like"
High ticket single purchases (a chair, a sauna, a mattress)	Close to nothing, and often negative	Nothing beside the buy box. Move the accessory to the post purchase page
One product stores	Nothing, by definition	Quantity tiers and a bundle, which is a different tool

Notice the pattern. Where the buyer's failure mode is "I picked wrong," alternatives earn their slot. Where the failure mode is "I forgot something," supplements earn theirs. Where the buyer is making one large, careful, once every eight years decision, the best recommendation strategy is an empty space and a stronger pitch.

A founder selling a $2,400 sauna asked me last quarter why their recommendation carousel converted at nothing. It converted at nothing because a person spending $2,400 on a box for their garage does not want to see three other boxes. They want one more paragraph about installation. Different problem, different tool.

What changes when the traffic comes from an AI assistant?

Worth flagging because it is moving fast in 2026.

More product research now happens in a chat window before anyone reaches your store. That visitor arrives having already been given a shortlist, which means they land on your product page with the comparison work finished. An alternatives carousel is close to useless for them, because the alternatives were already weighed somewhere you were not present.

The supplement widget, though, gets more valuable, not less. The assistant almost never tells someone which mat, which filter or which cable they will also need. That gap is yours to fill, and it sits at the bottom of your page where you already had a slot.

If a growing share of your traffic arrives pre-decided, weight your effort toward finishing the package and away from reopening the choice.

Five tests worth running

If you want to know what your recommendations do on your store, rather than what they do in a case study, run these in this order. Each one is a week, and each one answers a question the dashboard cannot.

Test 1: the holdout. Turn recommendations off for half your traffic for two weeks. Compare revenue per visitor, not clicks. This is the only test that answers the causation question, and almost nobody runs it. Expect a smaller number than the app promised. Sometimes expect zero.

Test 2: split the types. Alternatives in one labeled section low on the page, supplements in another beside the add to cart. Versus your current blended carousel. This is where most of the winners come from.

Test 3: the label. Same products, same slot, only the header changes. "You may also like" against "Customers who bought this also bought." Cheapest test on the list.

Test 4: count. Four suggestions against twelve. More options usually loses. A visitor who has to choose between twelve things often chooses none of them, and you have converted a decided buyer into a browser.

Test 5: the cart supplement. One item, under $40, one tap. Against no cart cross-sell at all. This is the closest thing to free average order value in ecommerce, and it is the one worth building properly rather than accepting the app's default. The mechanics overlap heavily with how to use product bundles to increase Shopify average order value.

Run them in that order because test 1 tells you whether to bother with tests 2 through 5.

How to run the holdout test without buying anything

People skip test 1 because it sounds like engineering. It is about two hours of work.

Pick your top three product pages by traffic, because a holdout on a page with 200 monthly sessions will tell you nothing for a year. Between them they need at least 8,000 sessions over the two week window, or the result will be noise you can read either way.

Split the traffic. A theme level toggle on a random 50% is cleanest, and if your setup will not allow it, run the calendar version instead: widget on for week one, off for week two, then repeat the pair to cancel out day of week effects. Cruder, still usable.

Then measure the right thing. Not clicks on the carousel, which only counts the people who were browsing anyway. Not conversion rate on its own, which will hide the whole average order value effect. Revenue per visitor for each group, calculated the plain way: total revenue from those pages divided by sessions to those pages.

Write down your prediction before you start. Everybody says they expected the result afterward.

Three outcomes and what each one means:

• Revenue per visitor is flat. The widget is decoration. Keep it if you like it, stop crediting it in reports, and put the effort into the page copy above it.
• Revenue per visitor is up, and average order value carries the gain. Working as designed. Now go run tests 2 through 5 to make it bigger.
• Revenue per visitor is down. Almost always placement. Something is competing with the buy box. Move it lower and rerun.

Two weeks. No new software. That result is worth more to your store than every statistic in this post, mine included.

What I would do first on a $40,000 a month store

Honestly? I would leave the recommendation widget alone for a month and go fix the page it sits on.

Here's why. A recommendation carousel can only rearrange the demand a page already created. If the page above it has not answered the buyer's real questions, adding suggestions to the bottom is decorating a leak. It is the same trap as adding a review widget to a page whose copy never made a claim worth reviewing.

The bedding brand we work with did the unglamorous version of this. Three hero products, rebuilt around what the buyer was actually deciding. Before: conversion rate 1.0%, average order value $125, revenue per visitor $1.25. After: conversion rate 3.5%, average order value $231, revenue per visitor $8.10. On 10,000 visitors, that's $12,500 before and $81,000 after. You can see the full case study numbers on our results page. Real client numbers, not typical results, and not a promise of what your store will do.

Nothing in that rebuild was a recommendation app. The average order value moved because the page made a bigger order make sense, which is the same reason a good supplement widget works and a bad one gets ignored. Same principle runs through the Shopify office chair product page optimization teardown: the accessory sells itself once the main decision feels safe.

Get the page right, then let the widget do the small job it is good at. In that order, the widget is worth having. In the other order, you are amplifying a leak.

The honest summary

Recommendations are a real tool with a modest, specific effect, oversold by attribution math that credits them for buyers they inherited.

• They move average order value dependably. Budget your effort there.
• They move conversion rate slightly, and only when they rescue a mismatch.
• They can cost you conversions when they interrupt a decided buyer above the fold.
• The label and the split between alternatives and supplements matter more than the algorithm.
• The holdout test is the only number you should trust about your own store.

If you install one thing from this post, install the holdout test. Two weeks, half your traffic, revenue per visitor as the scoreboard. You will learn more about your store than any vendor case study can teach you.

Frequently asked questions

Do product recommendations increase Shopify conversion rate? Modestly, and less than the marketing suggests. They move average order value far more dependably. The headline numbers, such as clickers converting 5.5 times higher, compare people who chose to click against people who did not, which measures intent rather than the effect of the widget.

What percentage of ecommerce revenue comes from product recommendations? Up to 31% of site revenue by common industry figures, with clickers generating roughly 26% of revenue and 24% of orders. Read those as influenced revenue, not added revenue.

Where should product recommendations go on a Shopify product page? Below the buy box. Supplements go next to the add to cart action and in the cart. Alternatives go lower on the page, where they catch a visitor who is drifting.

What is the difference between alternative and supplementary recommendations? Alternatives help someone find the right product. Supplements help them finish the package. Baymard found 58% of major sites show only one type or blend both into one element, which is why so many carousels do nothing.

Should I label my recommendation carousel? Yes, and say what the suggestions are based on. "Customers who bought this also bought" outperforms "You may also like" because the first one is a reason and the second one is a shrug.

How do I measure whether recommendations are working? Revenue per visitor on the pages carrying the widget, measured against a holdout group. Conversion rate 1.8% at a $70 average order value is $1.26 per visitor. Conversion rate 2.0% at $82 is $1.64. On 10,000 visitors that is $12,600 against $16,400.

Book Your Profit Audit

Most stores are running a recommendation widget that is quietly doing a third of the job it could do, on a page that is doing half the job it could do.

Get your free profit audit and we'll show you exactly where your revenue per visitor is leaking, then rebuild a high converting product sales page in less than 15 minutes so the suggestions at the bottom have something worth supporting.

Book Your Profit Audit →