# Price testing: definition, method and examples

Source: https://leverprice.com/glossary/price-testing
Updated: 2026-11-17

> Price testing means showing different prices for the same product to randomly assigned groups of visitors and keeping the price that earns the most. The deciding number is revenue per visitor rather than conversion rate, because a higher price can lose some buyers and still bring in more money.

A price test splits new visitors at random between two or more prices for the same plan and counts the money each group brings in. For example, half of the people who open your pricing page see Pro at $29 a month, the other half see $39, and after a few thousand visitors per price you compare what each half paid.

It is an A/B test scored in money, because the price changes both how many people buy and how much each of them pays.

## How to calculate it

Each price needs three counts from the same period: visitors who saw it, buyers and revenue.

```text
RPV             = revenue at a price ÷ visitors who saw that price
lift            = RPV of new price ÷ RPV of current price − 1
break-even loss = 1 − current price ÷ new price
```

Example: 4,000 visitors see $29 and 100 of them buy, so that price earns $2,900, or $0.725 per visitor. Another 4,000 see $39 and 84 buy: $3,276, or $0.819 per visitor. The $39 price converts worse (2.1% against 2.5%) and still earns 13% more per visitor.

The break-even line shows how much conversion a higher price can afford to lose. Moving from $29 to $39 pays off as long as fewer than 25.6% of buyers walk away (1 − 29 ÷ 39), and in this example 16% did. With 184 sales in total, though, a gap this size is still within the noise, which is why the next question is always [sample size](https://leverprice.com/glossary/sample-size).

*Figure: Chart: the share of buyers a price increase can lose before revenue falls. Buyers you can lose after a price increase and still earn the same revenue. From $29 to $39, that is 25.6%.*

## Why it matters for pricing

Many software prices are set once, by gut or by copying a competitor, and never checked. A test replaces the guess with what visitors actually paid. A survey such as Van Westendorp tells you which range feels acceptable; a test tells you which number inside that range earns the most.

The setup decides whether the result means anything. Assign the price on your server so nobody can edit it in the browser, keep each visitor on one price, enroll new visitors rather than existing customers, keep bots out of the counts, and use revenue your payment provider confirmed. [Leverprice](https://leverprice.com/price-testing) runs that loop for software products: it shows each visitor one price from your range, follows them to a paid order and moves traffic toward the price with the highest revenue per visitor.

For subscriptions, the first payment is only part of the answer. A price that wins on first purchases can lose over a year if its buyers cancel sooner, so check how each price's customers retain before you treat the winner as final.

## Common mistakes

- Picking the price with the best conversion rate. The cheapest price almost always converts best, and it often earns the least.
- Calling the test the day one price pulls ahead. A lead built on a few dozen sales flips easily, so set the stopping rule before you start.
- Showing the same person two prices. A visitor who sees $29 on Monday and $39 on Friday stops trusting the page, and their purchase lands in the wrong group.
- Counting crawlers, link previews and prefetch requests as visitors, which dilutes every price's revenue per visitor.

## Related terms

A price test is scored on [revenue per visitor](https://leverprice.com/glossary/revenue-per-visitor) and depends on [sticky assignment](https://leverprice.com/glossary/sticky-assignment) to keep the groups clean. [Statistical significance](https://leverprice.com/glossary/statistical-significance) tells you whether a gap is real, and a [multi-armed bandit](https://leverprice.com/glossary/multi-armed-bandit) moves traffic to the better price while the test is still running.

## Frequently asked questions

### How long does a price test take?

It depends on your sales volume and on how far apart the prices perform. In our simulation of the production engine, a product with 300 sales a month reached a decision in a median of 23 days when the best of three prices earned 20% more per visitor than the current one, and in 66 days when it earned 5% more. The [full simulation results](https://leverprice.com/blog/how-long-should-a-price-test-run) cover other volumes.

### Should existing customers see the test prices?

No. Run the test on new visitors and keep current subscribers on the price they signed up at. Changing what existing customers pay is a separate decision with its own churn risk.
