Break-even
Markup calculator
Markup and margin, plus the conversion rate a new price has to hold to earn the same gross profit.
A/B testing
A/B price testing: how to test prices without losing trust
A/B price testing shows visitors different prices, or the same price presented differently, to learn which earns the most. This guide covers the two kinds of price test, the metric that decides them, how much traffic they need, the legal and trust questions, and what to do when a live price test is the wrong tool.
What is A/B price testing?
A/B price testing is an experiment that splits visitors at random between two versions of a price, then compares how much revenue each version earns. One version is usually your current price or price display; the other changes the amount, or keeps the amount and changes how it is presented. The winner is the version that earns more per visitor.
People search for the same thing as split testing for pricing, pricing A/B testing or price experimentation. The method is the same as any A/B test: random assignment, one primary metric chosen before launch, a sample size planned in advance and a result read only when that sample is reached. What makes price tests different is the metric, the traffic they need and the trust you put at risk.
Price point tests and price presentation tests
There are two kinds of price test. A price point test changes what people pay: one price against another for the same product. A price presentation test keeps the price the same and changes how it is shown, such as the order of plans, a real former price crossed out, or a free shipping threshold. Most sites should start with presentation.
| Question | Price point test | Price presentation test |
|---|---|---|
| What changes | The amount charged for the same product | How the same price is framed, ordered or explained |
| Examples | One price against a higher price; a different discount level | Plan order, a recommended plan, monthly against annual framing, a bundle total, a free shipping threshold, a guarantee beside the price |
| Do two visitors pay different amounts for the same thing? | Yes, during the test | No |
| Legal and trust risk | Higher: see the legal section below | Lower, as long as every claim is true, such as a former price that was really charged |
| What you learn | How demand responds to price, for that product and audience | Which display earns more at the price you already charge |
A third route sits between them: test two different offers rather than one product at two prices. Two plans with different features, two bundle sizes or a new product launched at a test price give each buyer a fair, consistent price for what they chose, while still showing how people respond to price.
Why revenue per visitor, not conversion rate, decides a price test
Revenue per visitor decides a price test because it counts both how many people buy and how much each one pays. Revenue per visitor is total revenue divided by visitors, which equals conversion rate times average order value. A lower price usually converts better, so a conversion rate win can still lose money.
Here is a hypothetical example, with round numbers chosen only to show the arithmetic. Version A sells a product at $40 and converts 3.0% of visitors. Version B charges $48 and converts 2.6%.
| Measure | Version A ($40) | Version B ($48) | Winner |
|---|---|---|---|
| Conversion rate | 3.0% | 2.6% | A |
| Revenue per visitor (conversion rate x price) | $1.20 | $1.25 (0.026 x $48 = $1.248) | B |
| Gross profit per visitor, if each unit costs $20 | $0.60 (0.03 x $20) | $0.73 (0.026 x $28 = $0.728) | B |
| Conversion rate B needs to match A's revenue | $1.20 / $48 = 2.5% | ||
| Conversion rate B needs to match A's gross profit | $0.60 / $28 = about 2.14% | ||
Read on conversion rate alone, version A wins. Read on revenue or profit per visitor, version B wins. Work out the break-even conversion rate before launch, so you know what result would change your decision. For subscriptions, the first order is not the whole story: a higher price can bring fewer but longer-staying customers, or more cancellations, so track revenue over the months that follow as a guardrail.
Convertica's own published tests show why the two numbers have to be read separately. In the free shipping threshold test, a product page message telling shoppers how close their cart was to free shipping produced +88.3% rate of order completion and +175.4% revenue per visitor on desktop at 95% confidence; mobile showed no significant change. Test length was not published. The two figures differ because the threshold changed how much people bought as well as whether they bought.
How much traffic does a price test need?
A price test needs enough visitors to reach the sample size you calculate before launch, and that is usually more than a test judged on conversion rate. Revenue per visitor is noisier: most visitors spend nothing and a few spend a lot, so detecting the same relative change takes a bigger sample. There is no single traffic number that fits every site.
- Start from conversion rate. Enter your baseline conversion rate and the smallest change worth detecting into the A/B test sample size calculator. Treat its answer as a floor for a revenue test, not the full sample.
- Count only the traffic that sees the price. A pricing page or a single product page often gets far fewer visitors than the homepage.
- Convert the sample into days. Divide by daily visitors to the tested page and round up to whole weeks, as set out in how long to run an A/B test.
- Decide before launch what you will do if the test cannot reach its sample in a reasonable time. A test that would take many months is a sign to test presentation, a busier page or one of the alternatives below.
Price changes that matter to the business can still be small in percentage terms, and small effects need large samples. When the test ends, check the result with the statistical significance calculator for the conversion rate, and look at revenue per visitor and order value by device before you roll anything out.
How to A/B test prices in 7 steps
To A/B test prices, decide whether you are testing the price or its presentation, pick revenue or profit per visitor as the primary metric, work out the break-even conversion rate and the sample size, assign visitors at random, keep each visitor on one price everywhere, and run in whole weeks until the planned sample is reached.
-
Choose the question
Write down what you want to learn: whether a higher price holds up, which plan order earns more, or whether a shipping threshold raises order value. If the question can be answered by changing presentation alone, test that first.
-
Pick one primary metric and your guardrails
Use revenue per visitor, or profit per visitor when costs differ. Add guardrails you will watch but not optimize for, such as refunds, support contacts, cancellations and plan mix.
-
Work out the break-even point
Calculate the conversion rate the new version needs to earn the same revenue or profit, as in the example above. If that rate looks unrealistic, the test is not worth running in that form.
-
Plan the sample size and the end date
Use the sample size calculator, allow extra for a revenue metric, and write down the end date before launch. Do not stop early because one price is ahead.
-
Check the legal position and set your safeguards
Read the section below, ask your own counsel where it applies, and decide in advance how you will treat buyers who paid the higher price, for example by refunding the difference when the test ends.
-
Assign at random and keep the price consistent
Split visitors at random, never by personal data, and make sure each visitor sees the same price on the product page, in the cart, at checkout, in emails and in your structured data. A cart that shows a different price from the product page is a broken test and a broken promise.
-
Read the result, then roll out one price
When the planned sample is reached, compare revenue per visitor, check the guardrails and look at desktop and mobile separately. Then move everyone to the winning version, so the period with two prices ends.
Is it legal to show different prices to different customers?
In the sources checked for this guide, a random price test is not banned outright in the US or the EU, but some rules apply and they are changing. New York requires a notice when a price is set by an algorithm using personal data, and EU law requires a notice when a price is personalized by automated decision-making. This is general information, not legal advice: ask your own counsel.
What could be confirmed from official or primary sources when this guide was published:
- US price discrimination law is about business buyers. The FTC's guide to price discrimination under the Robinson-Patman Act describes a seller charging competing buyers different prices for the same commodity, and says the Act applies to commodities, not services. The page does not address sales to consumers, so it does not settle whether a consumer price test is allowed.
- The FTC is studying personalized pricing. In January 2025 the FTC published preliminary findings from its surveillance pricing study: details such as location or browsing history can be used to show individual consumers different prices for the same goods. The release describes the study as ongoing and makes no finding that anyone broke the law.
- New York requires a disclosure for algorithmic prices that use personal data. According to the New York Attorney General, the Algorithmic Pricing Disclosure Act took effect on 10 November 2025 and requires most companies that use algorithmic pricing to show "THIS PRICE WAS SET BY AN ALGORITHM USING YOUR PERSONAL DATA" near the price, with a penalty of up to $1,000 per violation. Whether a random test that uses no personal data is covered could not be confirmed here.
- The EU requires a notice for personalized prices. Directive (EU) 2019/2161 added a duty to tell consumers, where applicable, that a price was personalized on the basis of automated decision-making. Its recital 45 says this does not apply to dynamic or real-time pricing in response to market demand when there is no such personalization. Whether a random A/B split counts as personalization could not be confirmed here.
- Discount claims must be true in every version. If a variant shows a crossed-out former price or a list price, the FTC's Guides Against Deceptive Pricing still apply. The psychological pricing guide covers false reference prices, false deadlines and hidden fees.
Not covered here, because it could not be verified from an official source for this guide: state laws beyond New York, rules for regulated sectors such as insurance and credit, rules outside the US and the EU, and how privacy laws treat the data a testing tool uses to assign visitors. Laws on pricing are being written and challenged now, so check the current position where you sell before every price point test.
What happens when customers notice a price test?
Customers who find out they paid more than someone else for the same thing at the same time tend to see it as unfair, whatever the legal position. The best-known case is Amazon's DVD test in 2000, which the company called a mistake and answered by refunding the difference. Plan your safeguards before the test, not after someone notices.
In its own September 2000 statement, Amazon said the test varied discounts of 20% to 40% at random on 68 DVD titles over five days, denied that it used customer demographics, and said it had made 6,896 refunds averaging $3.10 to buyers who paid more than the lowest test price. It also promised that any future test would give every buyer the lowest test price once the test ended.
Safeguards that lower the risk
- Refund the difference to anyone who paid the higher test price, once the test ends.
- Test on new visitors, new products or new plans, not on existing subscribers who already know your price.
- Assign at random, never by location, device, income signals or any other personal data.
- Keep the test short and planned, with a fixed end date, then move everyone to one price.
- Prepare an honest answer for support staff, in case a customer asks why a friend saw a different price.
Price presentation tests to run before you change the price
Price presentation tests change how a price is shown without changing what anyone pays, so they avoid most of the legal and trust risk of a price point test. Anchors, plan order, discount framing, shipping thresholds and bundles can all change revenue per visitor at the price you already charge, which makes them the natural first price tests.
| Test | Example variants | Watch |
|---|---|---|
| Price anchor | Most complete plan shown first against cheapest first; a bundle shown against the total of its parts | Plan mix, revenue per visitor |
| Recommended plan | Highlight the middle plan against the higher plan | Plan mix, revenue per visitor |
| Billing default | Monthly selected against annual selected | Sign-ups, revenue per visitor, cancellations |
| Discount framing | Percent off against amount off, for the same real discount | Revenue per visitor, order value |
| Free shipping threshold | No message against "you are $X away from free shipping" | Order value, revenue per visitor |
| Shipping shown beside the price | Shipping cost revealed at checkout against stated on the product page | Order completion, revenue per visitor |
| Package layout on mobile | Packages stacked against a swipeable comparison | Revenue per visitor on mobile |
The psychological pricing guide explains why the first price a visitor sees changes how the rest are judged, and the discount calculator checks that a percent off and an amount off really describe the same discount.
Alternatives to A/B testing your prices
When traffic is too low or the risk is too high for a live price test, there are other ways to learn about price. Presentation tests, before and after comparisons, painted door tests and price surveys such as the Van Westendorp price sensitivity meter each answer part of the question, with weaker evidence than a randomized test.
| Method | How it works | What it tells you | Main limit |
|---|---|---|---|
| Presentation test | A normal A/B test of how the same price is shown | Which display earns more at today's price | Does not tell you the best price |
| Before and after | Change the price for everyone, then compare with the period before | A rough read on the new price, with one price at a time | Seasonality, campaigns and other changes mix into the result |
| New offer at a test price | Launch a new product, plan or bundle and compare its price with what you already sell | How people respond to price on something with no old price to compare | The offers differ, so the comparison is not like for like |
| Painted door test | Show a price or plan that does not exist yet; a click leads to an honest waitlist message, never a charge | How many visitors try to choose it | Measures interest, not whether people would pay |
| Van Westendorp survey | Four questions: at what price the product seems too cheap to trust, a bargain, getting expensive, too expensive to consider | A range of prices your audience accepts | Measures what people say, not what they buy |
| Gabor-Granger survey | Asks whether the respondent would buy at a series of set prices | A stated demand curve across those prices | Stated intent can differ from real buying |
The Van Westendorp price sensitivity meter was presented by the Dutch economist Peter van Westendorp at the ESOMAR congress in 1976. Plotting how many respondents call each price too cheap or too expensive gives an acceptable range and reference points inside it. Survey methods are best used to narrow the range of prices worth a live test, not to replace one.
Price presentation tests from Convertica case studies
Convertica's published case studies include three tests that changed how a price or its shipping cost was presented, without changing the price itself. Each was judged on revenue per visitor as well as order completion, and the figures below are as stated in the case studies.
| Case study | What was tested | Stated result |
|---|---|---|
| Free shipping threshold | A product page message telling shoppers whether their cart already qualified for free shipping, or how much more it needed | +88.3% rate of order completion and +175.4% revenue per visitor on desktop at 95% confidence; mobile showed no significant change. Test length not published. |
| Free shipping badge | A "Free Shipping Included" badge placed next to the price on product pages | +27% rate of order completion and +60.3% revenue per visitor on desktop; mobile showed no significant change. Test length and confidence level not published. |
| Mobile pricing table | A supplement store's stacked packages turned into a swipeable slider on mobile | +43.9% rate of order completion and +36% revenue per visitor on mobile at 98% confidence. Test length not published. |
Results like these depend on the site, the audience and the starting point, so treat them as examples of what to test rather than results to expect. More store tests are collected in the guides to ecommerce A/B testing, Shopify A/B testing and landing page A/B testing, and in the A/B testing examples.
Find out what to test on your pricing first
Before you test a price, check that visitors can see it, understand what it includes and trust it. Convertica's free CRO audit checks a page for people and for AI agents and gives you every finding, with why each one matters, so your first price test starts from a real problem rather than a guess.
If you want a team to plan and run the tests, Convertica quotes A/B testing one by one as a custom proposal: see A/B testing services to ask for an A/B testing proposal.
Free tools for planning a price test
Work out the break-even point, the sample and the result with Convertica's free calculators. Each shows its formula and a worked example.
Plan a test
A/B test sample size calculator
Visitors each version needs, and how many days that takes at your traffic.
Check a result
Statistical significance calculator
Whether a finished A/B test result is bigger than chance.
A/B testing services
Want your A/B tests run for you?
Convertica runs A/B tests for clients, quoted one by one as a custom proposal. Monthly traffic comes first, because it decides whether a test on your site can reach a clear result, and how long it takes.
The split tests in Convertica's case studies were run by its team for past clients, and each result is published with its metric, timeframe and source.
Short answer
A/B price testing splits visitors at random between two prices, or two ways of presenting a price, and compares the revenue each version earns. Judge it on revenue per visitor or profit per visitor, not conversion rate, plan the sample size first, and prefer presentation tests, because charging people different prices for the same thing raises legal and trust risks.
- There are two kinds of price test: the price point itself, and how the same price is presented (anchors, plans, discounts, shipping thresholds).
- Revenue per visitor equals conversion rate times average order value, so a price that converts worse can still earn more.
- Revenue metrics are noisier than conversion rate, so price tests usually need more traffic than a headline test.
- Showing different prices for the same product is not banned outright in the sources checked here, but disclosure rules exist in some places. Ask your own counsel before you test.
- When traffic is low or the risk is high, use a presentation test, a painted door test or a price survey such as Van Westendorp instead.
Questions about A/B price testing
Is A/B price testing legal?
It depends on where you sell and how prices are set. The sources checked for this guide do not ban random price tests outright, but New York requires a notice when a price is set by an algorithm using personal data, and EU law requires a notice for prices personalized by automated decision-making. Ask your own lawyer before you test.
What is the difference between price testing and price presentation testing?
A price test changes the amount people pay, such as one price against another. A price presentation test keeps the price the same and changes how it is shown: plan order, a crossed-out real former price, monthly against annual framing, a free shipping threshold or a bundle. Presentation tests carry far less legal and trust risk.
What metric should a price test use?
Use revenue per visitor, or profit per visitor if your costs differ between versions. Revenue per visitor is total revenue divided by visitors, which equals conversion rate times average order value. Conversion rate alone can mislead: a lower price usually converts better and can still earn less.
How much traffic do you need to A/B test prices?
Enough to reach the sample size you calculate before launch, and there is no single number that fits every site. Work it out from your baseline conversion rate and the smallest change worth detecting, then expect to need more, because revenue per visitor varies more from visitor to visitor than a yes or no conversion does.
Can you A/B test prices on Shopify?
Yes, with a testing app that changes the price shown and charged, or by testing two products or plans rather than one product at two prices. Check that the price in the cart and at checkout always matches the price the visitor saw on the product page, and plan the sample size before you start.
Should existing customers be included in a price test?
Usually not. Existing customers and subscribers already know your price, compare notes and are the most likely to notice a difference. Testing on new visitors, new products or new plans lowers that risk. If a test ever charges some buyers more, consider refunding the difference once it ends.
What is a painted door test for pricing?
A painted door test shows a price or plan that does not exist yet and measures how many visitors try to choose it. Clicking leads to an honest message, such as a waitlist, never a charge. It measures interest in an option, not whether people would actually pay, so treat the result as a signal rather than proof.
What is the Van Westendorp price sensitivity meter?
It is a survey method from 1976 that asks four questions: at what price the product would seem too cheap to trust, a bargain, getting expensive and too expensive to consider. The answers give a range of prices people accept. It measures what people say, not what they buy, so it narrows the range for a later test.
