Shopify A/B Testing Guide 2026: Rollouts, Apps, Sample Size

To A/B test on Shopify in 2026 you have two routes. Shopify's built-in Rollouts splits live traffic between two theme versions with no app, but the Help Center says experiments need the Grow plan or higher, and its analytics report rates rather than a verdict (read September 6, 2026). Apps such as Shoplift, Intelligems and ABConvert add price testing and a statistics engine from $69 to $99 a month. Whichever route you take, size the test first: a store converting at 2% needs about 80,700 visitors per version to detect a 10% lift, on the standard formula.

What A/B testing on Shopify actually is

An A/B test shows your current page to one random half of your shoppers and a changed version to the other half, at the same time, and measures which half earned more per visitor. The random split is the point. Same week, same ads, same weather for both groups, so whatever gap opens up belongs to the change and nothing else.

That is what separates a test from a before-and-after. Ship a new product page on Tuesday and see sales up by Friday, and you have learned almost nothing, because payday or a creator mention could have done that. Run the two pages side by side and the calendar cancels out.

Two words you will see all through this guide. The control is what you have now. The variant (Shopify's Rollouts calls it the treatment) is the version with the change. One change per test, so you know what caused the result.

Heatmaps, recordings and analytics tell you what shoppers do. A test is the only tool that tells you whether a change caused more revenue. The wider Shopify CRO playbook covers the finding part; this page covers the proving part.

One thing to clear up, because old guides still send people the wrong way: Google Optimize is gone. Google's own notice says Optimize and Optimize 360 are no longer available as of September 30, 2023. Any guide telling you to wire Optimize into your Shopify theme is out of date.

Step 1. Pick the first thing to test

Test the change with the strongest evidence behind it and the largest expected effect. Not the one that is quickest to build.

The order matters more than most guides admit, and the reason is arithmetic you will meet in step 3. The traffic a test needs grows with the square of one over the effect you want to detect. A change that might move revenue per visitor by 20% needs roughly a sixteenth of the visitors that a 5% tweak needs. On a store with modest traffic, that gap is the difference between a result in a month and no result at all.

So on a first test, skip the button colour and the headline wording. Go after structural changes on the pages every plan can test, which is everything before checkout. Candidates that fit that description on most stores:

• A sticky Add to Cart bar on mobile. A thin bar that pins the price and the buy button to the bottom of the screen while the shopper scrolls the gallery and the reviews.
• Price and Add to Cart above the fold on mobile. If the first screen of your product page is a hero image and a title, the first screen sells nothing.
• A shipping and returns line under the buy button. One true sentence: delivery window, free returns, whatever you actually offer.
• Star rating and review count beside the price. A price with no proof next to it reads as a gamble.
• Inline variant buttons instead of a dropdown. Size and colour as tappable chips, so a phone shopper never opens a native picker.
• A free-shipping threshold with a progress bar in the cart. "You are $12 from free shipping" gives a reason to add one more item. Watch the shipping margin.
• Product image order and type. Lifestyle first versus product-on-white first is one of the cleaner tests a store can run. We wrote up how to A/B test product images separately, including what to hold constant.

Write each candidate down with the evidence for it: a recording where shoppers scroll past the buy button, a support ticket about returns, a mobile conversion rate half the desktop one. Then rank by evidence, then by expected size, then by effort. The top of that list is your first test.

Step 2. Write the hypothesis and pick the metric

Write one sentence: "If we change X, revenue per visitor will improve, because Y." Then decide, in writing, that revenue per visitor is the metric the test will be judged on.

The metric choice is where most Shopify tests go wrong, and it goes wrong before the test starts. Conversion rate counts whether people buy. Revenue per visitor (RPV) counts whether they buy and how much they spend, because it equals conversion rate multiplied by average order value. A change can lift the first and shrink the second.

Here is the trap with worked numbers, so you can see the shape of it. Say you test a "10% off, next 10 minutes" banner against your normal product page. Control: 1,000 sessions, 30 orders, 3.0% conversion, $80 average order, $2.40 per visitor. Variant: 1,000 sessions, 36 orders, 3.6% conversion, a 20% lift. But the banner pulled shoppers toward smaller, discount-driven baskets, and average order value fell to $62. The variant's RPV is $2.23. Conversions went up. Money went down.

Control

3.0% conversion rate

$80 average order

$2.40 per visitor

Variant: "10% off, next 10 minutes"

3.6% conversion rate (+20%)

$62 average order

$2.23 per visitor

Illustrative numbers. A tool that judges on conversion rate ships this. A tool that judges on revenue per visitor does not.

This matters twice over on Shopify, because the native tool cannot help you here. The Help Center's rollout analytics page lists the metrics an experiment reports: conversion rate over time, bounce rate, reached checkout rate, add to cart rate and checkout conversion rate. Sales figures appear in the launch metrics, not the experiment metrics, and the page states you cannot customize the available metrics. So if you run your test in Rollouts, you will need to pull revenue by version from your own reports to judge it properly. Measuring revenue per visitor on a Shopify store walks through where to get it.

Keep conversion rate and average order value on screen as guardrails. They tell you how a change earned or lost. They do not get the deciding vote.

Step 3. Work out the sample size before you start

Most small Shopify stores cannot reach textbook significance on a small lift inside a sensible window. Better to know that on day one than on day ninety.

Four things set how many visitors a test needs: your current conversion rate, the smallest lift you want to be able to detect (the minimum detectable effect, or MDE), the confidence level, and the statistical power. The last two are usually fixed at 95% and 80%. So on a real store, two knobs do the work: how often people convert now, and how small a change you are trying to catch.

The table below is our arithmetic on the standard two-proportion formula, at 95% confidence and 80% power with a 50/50 split. Figures are per version, so double them for the whole test. They are a floor: revenue per visitor is noisier than conversion rate, and weekly cycles and novelty add variance on top.

Current conversion rate	To detect +5%	To detect +10%	To detect +20%	To detect +30%
1%	~637,000	~163,100	~42,700	~19,800
2%	~315,200	~80,700	~21,100	~9,800
3%	~207,900	~53,200	~13,900	~6,500
4%	~154,300	~39,500	~10,300	~4,800

Two-proportion z-test, two-sided, alpha 0.05, power 0.80. Relative lifts. Rounded to the nearest hundred. Double each figure for the total sessions the test needs.

Read across the 2% row. Catching a 10% lift takes about 80,700 visitors per version, so roughly 161,000 in total. Catching a 5% lift takes about 315,000 per version. Same store, half the ambition, four times the traffic. That is the squared rule doing its work, and it is why the first test should be a big swing.

Now turn visitors into weeks, because weeks are what you feel. A store doing 5,000 sessions a month at 2%, hoping to detect a 10% lift, needs about 32 months on those numbers. A store doing 20,000 sessions a month needs about eight months for a 10% lift and about two months for a 20% lift. A store doing 50,000 a month at 3% can settle a 10% lift in roughly two months.

5,000 sessions / mo

~32 months

20,000 sessions / mo

~8 months

45,000 sessions / mo

~3.6 months

A healthy test runs two to six weeks. If your number lands outside that, change the test (bigger change, bigger MDE) rather than the calendar.

Use the sessions on the page or template you are testing, not the site total. A product page with 1,800 visits a month plays by different rules from your homepage. Then run your own numbers through our statistical significance calculator, which tells you the sample size before it will show you a p-value, and says "not enough data" when that is the honest answer.

Halve the lift you want to catch and you need about four times the traffic. Small wins are not cheaper to prove. They are far more expensive.

Step 4. Choose the tool: Rollouts or an app

If you are on Grow or higher and your test is a whole-theme change, Rollouts is free and flicker-free. If you want to test a price, a shipping rate, a single section, or you need a statistics engine, you need an app.

Option A: Shopify Rollouts, step by step

Rollouts lives in the admin under Markets. The Help Center's requirements page, read September 6, 2026, carries the two sentences that settle the plan question: "Rollouts are available on the Basic plan or higher." "Experiments are available to stores on the Grow plan or higher." Advanced and Plus are not named. A launch is a rollout with no change-level traffic split; an experiment compares a control and a treatment. Only the experiment is an A/B test.

Shopify's pricing page localizes by visitor, so we cannot give you a USD figure we read ourselves. From Canada on September 6, 2026 it showed Grow at CA$99 a month on annual billing or CA$132 month to month, with Basic at CA$37 or CA$49. Check the page from your own location.

To set up an experiment, following the Help Center's create-a-rollout page:

1. Duplicate your published theme and make your one change in the copy. Do not touch the live theme.
2. In the admin, go to Markets, then Rollouts, and click Create rollout.
3. Give it a name. The Help Center's own example is "Summer Sale 2026".
4. Click Add changes to your store and choose the theme change (or a checkout and accounts configuration replacement; those are the only two change types).
5. In the Change section, click the line that reads 100% of eligible visitors will see this and set the percentage with a preset or the slider. For a clean A/B test, 50%.
6. Click Select launch date and time, then Add end date. Pick the end date from the sample size you worked out in step 3, not from a hunch.
7. Publish the rollout, then leave it alone until the end date.

What Rollouts cannot do, from the same requirements page: it supports online store checkouts only, not headless or custom storefronts; it cannot be applied to vintage themes; and "You can't change Liquid templates as part of a rollout." Prices and discounts are not a change type, so a price test is impossible here. There is no audience segmentation and no custom goal.

The bigger gap is in the analytics. The rollout analytics page lists five experiment metrics, all rates, and nowhere on any Rollouts page does the Help Center mention statistical significance, confidence, a winner, or revenue per visitor. Shopify's own blog describes the feature as "split traffic between your current theme and a variant, measure conversion rate, and publish the winning version". The winning part is your call, made with a calculator, on revenue you pulled yourself. Our Shopify Rollouts A/B testing breakdown goes through the plan table, the analytics screen and how to compute significance from what it gives you.

Shopify gives you the pipes. Deciding what to test, building the variant, and judging the result on money is the part you add.

Option B: an A/B testing app

Every price below was read off the named page on September 6, 2026. Entry is the cheapest paid tier; annual figures are what the listing prints, not our multiplication. Prices move, so check the listing before you install.

Tool	Best for	Price	Price tests	Statistics	Read on
Shopify Rollouts	Whole-theme and checkout-configuration splits, server-side	No add-on fee; experiments need the Grow plan	No	Rates only; no significance, no winner	help.shopify.com, Rollouts pages
Trident AB	A first cheap product, price or page test	Free (500 impressions, 1 test); paid $19.99 to $49.99/mo	Yes	Basic	apps.shopify.com/tridentab
Intelligems	Price, shipping, discount and offer tests scored on profit	Smart Content $69/mo ($660/yr); Unlimited $349/mo ($3,468/yr)	Yes, on Unlimited ($349)	Yes, profit per visitor	apps.shopify.com/intelligems
Shoplift	Template, theme and URL tests with segmentation	Core $99/mo ($888/yr, from $74 by visitors); Advanced $399; Pro $999	Beta, on Advanced ($399)	Yes, Bayesian	apps.shopify.com/shoplift
ABConvert	Price, shipping, theme and checkout tests metered by test orders	Starter $99/mo (1,000 test orders); Growth $199; Scale $399; Pro $599	Yes, from Growth ($199)	Yes	apps.shopify.com/a-b-convert-price-a-b-test
Convert	Full experimentation platform, client-side, you bring the hypotheses	Growth $399/mo or $299/mo annual; Pro $599 or $420 annual	No	Yes, mature engine	convert.com/pricing
VWO	Enterprise testing and insights suite	No public price; every plan says Schedule a Demo	No	Yes	vwo.com/pricing

All App Store listings state charges are billed in USD every 30 days. Trials: Intelligems, Shoplift and ABConvert 14 days; Trident AB 5 days on paid tiers; Convert 15 days. ABConvert meters by test orders and charges an overage per extra thousand ($59 on Starter, down to $24 on Pro). Shoplift meters by monthly unique visitors.

What the table compresses. Intelligems' $69 entry tier covers content and theme tests; price testing, the thing it is known for, sits on the $349 Unlimited tier, which its own website scales with order volume (our Intelligems pricing page explains how). Shoplift's "from $74" is the annual rate at the lowest visitor band. Trident AB's free plan, one test and 500 impressions a month, decides nothing but teaches the workflow.

For the full ranking with trade-offs and who each one fits, read the best Shopify A/B testing apps compared. For what a testing program costs at your traffic level once meters and overages kick in, Shopify A/B testing pricing does the sums.

Step 5. Build the variant without breaking the store

Build the variant in a duplicate theme, change one thing, and check it on a phone before a single shopper sees it.

The safe pattern on Shopify is the same whichever tool you use. Duplicate the published theme. Make the change in the copy. Preview it on mobile, on the slowest connection you can simulate, and click through to checkout once. Only then attach it to a test. This keeps a clean rollback: if anything goes wrong, you publish the original and the variant disappears.

Two mechanics decide how clean the test is:

• Server-side versus client-side. Rollouts decides which theme to serve before the page leaves Shopify, so the variant is there at first paint. Most apps decide in the browser and swap content after load. Done well, that is a few milliseconds of masking; done badly, shoppers see the control flash into the variant, which is called flicker, and it biases the test against the variant.
• Whole theme versus single section. Rollouts and Shoplift's theme tests swap the whole theme version. That is the right tool for a redesign or a new template. For a single change, an app that targets one section or one page keeps the rest of the store identical, which is what you want when you are trying to learn what caused the result.

Theme testing has its own traps, from settings that live in the theme file and silently differ between copies, to apps that inject into one theme and not the other. Our Shopify theme A/B testing guide lists the checks to run before a theme test starts.

Step 6. Run it clean

Randomize, keep each shopper on their version, run full weeks, and do not look at the scoreboard until the finish line you set.

Sticky assignment. A shopper who sees the variant on Monday must see it again on Wednesday. Apps handle this with a cookie or a customer ID. A hand-rolled test with two themes and a redirect usually gets this wrong.

Full business cycles. Weekend shoppers behave differently from weekday shoppers. Run at least two full weeks even if the sample size arrives sooner, and never end a test mid-week.

No promotions in the window. A sale or an ad-spend change hits both versions, which is fine, but it also changes who is arriving, which can change which version wins. If something big lands, note the date and read the result in two halves.

No peeking. This is the one that quietly ruins more Shopify tests than any other. Evan Miller's "How Not To Run an A/B Test" walks through the arithmetic: if you check the result after every observation at a 5% significance level and stop as soon as it crosses the line, your real false-positive rate is about 26%, more than five times what you thought you were accepting. His fix is the same as ours: "Decide on a sample size in advance and wait until the experiment is over before you start believing the 'chance of beating original' figures." If you want to monitor as you go, use a tool whose statistics are built for repeated looks, and understand that Rollouts' dashboard is not one.

Watching a test every morning and stopping when it looks good turns a 5% chance of fooling yourself into roughly one in four. Set the finish line first.

Check the split. Before you read any result, confirm the two groups are the size you asked for. If you set 50/50 and got 58/42, something is broken (a redirect, a cache, a bot filter) and the result is contaminated. This is called a sample ratio mismatch. Fix the cause and rerun.

The full checklist for an honest run, including how to pre-register the test so nobody moves the goalposts halfway through, is in how to run an A/B test honestly.

Step 7. Read the result honestly

A result is real when it survives a checklist, not when a dashboard turns green.

Start with what significance means, because almost everyone reads it backwards. When the variant comes out ahead, there are two explanations: it is better, or it got a lucky draw in the split. The p-value puts a number on the second one. A p-value of 0.03 says that if the two versions were truly identical, a gap this size would show up about 3% of the time by chance. It does not say there is a 97% chance the variant wins. It measures surprise, not odds.

A better habit than staring at the p-value is to report the range. Instead of "the variant lifted revenue per visitor 8%", say "+2% to +14%, most likely +8%". That is the confidence interval, and it is the honest answer, because a test never hands you one true number.

Grey is your current page, green is the variant. The width of each curve is the uncertainty. An honest test resolves to a range, never a single number and never a fake 100% win.

Then the checklist, in order:

• The split came out as designed (no sample ratio mismatch).
• The test ran at least two full weeks and reached the sample size you set in step 3.
• No sale, launch or ad-spend change landed inside the window, or you have read the halves separately.
• The confidence interval on revenue per visitor sits entirely above zero.
• The win holds on mobile and desktop separately, rather than a desktop win hiding a mobile loss.
• One whale order is not carrying the result. Cap each session's revenue at the 99th percentile and re-check.
• Average order value and margin, the guardrails, did not fall in a way that the RPV number hides.

Clear all seven and you ship the variant, write down the lift as a range, and put the next test on. Miss one and the honest label is "keep running" or "let it go". Neither is "publish it".

Price testing rules on Shopify

You can test prices on Shopify, with an app, and the rules are stricter than for a layout test.

Rollouts cannot do it. The Help Center lists theme changes and checkout and accounts configuration as the only change types a rollout carries, and a product price is neither. That leaves the apps. On the App Store listings read September 6, 2026, Intelligems unlocks price testing on its $349 a month Unlimited tier, ABConvert on its $199 Growth tier, and Shoplift marks price testing as beta on its $399 Advanced tier. Trident AB's listing is titled "Product and Price Testing" and runs from free to $49.99.

Four rules before you touch a price:

1. Judge on revenue or profit per visitor, never on conversion rate. A lower price almost always converts better. The question is whether the extra orders cover the margin you gave away on every order, and only a per-visitor money metric answers it.
2. The price a shopper sees is the price they pay. The variant price has to carry through cart, checkout and the order confirmation. Any tool worth paying for handles this; a hand-rolled test with two product listings does not.
3. Keep the shopper on one price for the whole session and, ideally, on return visits. A shopper who sees $39 today and $34 tomorrow has a reason to distrust you, and a screenshot to post.
4. Size it for the effect you expect. A price change usually moves conversion by more than a layout change does, so price tests often need less traffic. They still need the calculation.

Our Shopify price testing guide goes deeper on test design, how to read margin against volume, and which tools carry a variant price through checkout properly.

Checkout testing, and why the product page is the wedge

Checkout is the least testable part of a Shopify store, and the plan you are on decides how much of it you can touch at all.

The Help Center's checkout customization page draws the line plainly. Apps that customize the information, shipping and payment pages are available to Shopify Plus. Apps that customize the thank-you and order-status pages are available on Basic or higher. Rollouts adds one more route: it can swap a checkout and accounts configuration as a change type, so a Plus store with two checkout configurations can experiment between them.

The old routes are closed. Per shopify.dev, checkout.liquid and additional scripts were sunset for the thank-you and order-status pages on August 28, 2025, and script tags were sunset the same day for Plus stores and on August 26, 2026 for everyone else. A guide that tells you to edit checkout.liquid predates all of that.

Which is the practical argument for starting where this guide started: everything before checkout is testable on every plan with every tool in the table above.

A/B testing on a low-traffic Shopify store

Our rule of thumb, derived from the sample-size table above rather than any published threshold: at 10,000 sessions a month on the page and a 2% baseline, a 20% lift needs about 42,200 sessions in total, roughly four months. Below that, a classic 50/50 test on a modest lift will not finish in season. You change what you test and how you measure, rather than waiting.

Go back to the 2% row of the sample-size table. At 5,000 sessions a month, a 10% lift takes about 32 months to settle. A 20% lift takes about eight. A 30% lift, roughly four. The only lever that moves those numbers by a lot is the size of the change, so on a small store you only test changes you expect to move revenue per visitor by 15% or more: the sticky Add to Cart bar, the reviews block, the free-shipping bar, the product-page layout. Never the shade of a button.

Then change the measurement. Instead of starving both versions with a 50/50 split, ship the change to 80 or 90% of visitors and keep a 10 to 20% holdback on the old version as a living control. Compare the two, and compare the treated group against the same weeks before the change as a second read. The honest label on this method is "measured, not fully controlled": it caps your confidence at "likely" rather than "strong", and a person should approve before the change becomes permanent. "Likely, and shipped this month" beats "strong, and proven in two years".

One more lever: pool up a level. One product page may see 1,800 visits a month while the product template across every product sees 40,000, so test the template.

Shopify also offers a simulation route: SimGym, an app in AI Research Preview, runs AI shoppers over an unpublished theme and reports which version they preferred. It is a prediction, not a test; we looked at how accurate SimGym's AI shoppers really are before you lean on it. The full method for small stores, with a decision table by traffic tier, is in A/B testing on a low-traffic Shopify store.

The mistakes that make a Shopify A/B test lie to you

Every one of these is common, and every one produces a confident, wrong answer.

• Judging on conversion rate. The discount example above. Judge on revenue per visitor.
• Stopping when it looks good. About 26% false positives if you check after every observation (Evan Miller). Set the sample size first.
• Testing something too small to detect. A 3% lift on a 2% conversion rate needs traffic most stores do not have. Test bigger.
• Changing two things at once. You get a result and no idea what caused it.
• Letting a whale order decide it. One $600 order in a test of 40 orders is the result. Cap session revenue at the 99th percentile before reading.
• Reading Rollouts' rates as a verdict. Five rates per side, no significance and no sales. It is a data source, not a judge.

The verdict

For most Shopify stores in 2026, the honest route is this. Test the mobile product page or the cart first, with a structural change. Judge it on revenue per visitor. Size it before you start. Use Rollouts if you are on Grow and the change is a whole theme; otherwise start with Shoplift at $99 for theme and template tests, or ABConvert at $199 if the test is a price. Do not look until the end date, and read the result as a range.

Rollouts is a real gift, and also a half of one. It removed the flicker, the app fee and the theme hacking from the splitting half of a test. It did nothing for the other half: choosing what to test, building the variant, and deciding on money whether it won. That half is still the work, and it is where the lift actually comes from. Whatever tool you pick, that is the part to be disciplined about.

You already paid for the traffic. Testing is how you stop handing it back.

Disclosure: we are building StorePilot, a CRO tool for Shopify that runs this loop; it is pre-launch, and nothing on this page depends on it. Every price above was read off the vendor's own page on September 6, 2026 and will move.

Questions merchants keep asking

Does Shopify have built-in A/B testing?

Yes. Shopify Rollouts, under Markets in the admin, has a rollout type called an experiment that compares a control against a treatment on live traffic. The Help Center states that experiments are available to stores on the Grow plan or higher (read September 6, 2026). It reports five rates per side, including add-to-cart rate and conversion rate. It does not compute significance, name a winner, or show revenue per visitor.

How do I A/B test on Shopify without an app?

Use Rollouts if you are on Grow or higher. Duplicate your published theme, make one change in the copy, then go to Markets, Rollouts, Create rollout, add the theme change, set the visitor percentage to 50%, add a launch date and an end date, and publish. At the end, run the two conversion rates through a significance calculator, because Rollouts will not do that step for you. On Basic you can schedule a launch but cannot split traffic.

Which Shopify plan do I need for A/B testing?

Grow or higher for native experiments, per the Help Center's requirements page on September 6, 2026. Advanced and Plus are not named as requirements anywhere on the Rollouts pages. Shopify's pricing page localizes by visitor; from Canada it showed Grow at CA$99 a month on annual billing or CA$132 month to month. Third-party apps run on any plan, because they split traffic themselves rather than through Rollouts.

How much traffic do I need to A/B test on Shopify?

It depends on your conversion rate and the size of the lift you want to catch. On the standard two-proportion formula at 95% confidence and 80% power, a store converting at 2% needs about 80,700 visitors per version to detect a 10% relative lift, and about 21,100 per version to detect a 20% lift. Double those for the whole test. Those are our arithmetic, not a vendor figure. Our rule of thumb, taken from that same table rather than any published threshold: at 10,000 sessions a month on the page and a 2% baseline, a 20% lift needs about 42,200 sessions in total, which is roughly four months. Below that traffic, test bigger changes or use a holdback instead of a 50/50 split.

How long should a Shopify A/B test run?

Until it reaches the sample size you calculated before starting, and never less than two full weeks so both weekend and weekday shoppers land in both versions. A store doing 20,000 sessions a month at a 2% conversion rate needs about eight months to settle a 10% lift and about two months to settle a 20% lift, on our arithmetic. If the answer is longer than about six weeks, change what you are testing rather than waiting.

What is the best A/B testing app for Shopify?

It depends on the job. For theme and template tests with a statistics engine, Shoplift (Core $99 a month, App Store listing read September 6, 2026). For price and offer tests, Intelligems (price testing on the $349 Unlimited tier) or ABConvert (from the $199 Growth tier). For a first cheap test, Trident AB (free to $49.99). Rollouts is the free option if you are on Grow and only need whole-theme splits.

How much do Shopify A/B testing apps cost?

From $0 to $999 a month on the listings read September 6, 2026. Rollouts has no add-on fee but needs the Grow plan for experiments. Trident AB runs free to $49.99. Intelligems starts at $69 for content tests and $349 for price tests. Shoplift starts at $99, ABConvert at $99 with price testing from $199. Convert is $399 a month or $299 on annual billing. VWO publishes no price at all; every plan card on vwo.com/pricing says Schedule a Demo.

Can I A/B test prices on Shopify?

Yes, with an app, not with Rollouts. The Help Center lists theme changes and checkout and accounts configuration as the only rollout change types, and prices are neither. Intelligems unlocks price testing on its $349 a month Unlimited tier, ABConvert on its $199 Growth tier, and Shoplift marks price testing as beta on its $399 Advanced tier, all per their App Store listings on September 6, 2026. Judge a price test on revenue or profit per visitor, never on conversion rate, because a lower price nearly always converts better while earning less.

Can I A/B test the Shopify checkout?

Only partly, and mostly on Plus. Rollouts can swap a checkout and accounts configuration as one of its change types. Apps that customize the information, shipping and payment pages are Shopify Plus only, per the Help Center's checkout customization page; apps that touch the thank-you and order-status pages work on Basic or higher. Everything before checkout is testable on every plan.

What should I A/B test first on Shopify?

The change with the strongest evidence behind it and the biggest expected effect, which on most stores lives on the mobile product page or in the cart: a sticky Add to Cart bar, a shipping and returns line under the buy button, visible review counts near the price, a free-shipping progress bar. These are structural, reversible, and large enough to detect on modest traffic. Leave button colours until you have traffic to spare.