Average order value (AOV) and Revenue per visitor (RPV) at a glance
Average order value is total revenue divided by the number of orders in the same period. The AOV glossary entry covers the revenue conventions behind it and why a mean order can describe an order nobody actually placed. It is the metric most merchandising work aims at: bundlesSeveral products sold together as one item at a single price, used to raise order value; a kit is a bundle designed as a starting point for new customers., cross-sellsOffering a customer a related product alongside the one they are buying; upselling offers a higher-value version instead., delivery thresholds and volume pricing all try to raise it, and the AOV levers guide sets those levers out in detail. This comparison does not repeat them.
Revenue per visitor is total revenue divided by the number of visitors in the same period, where the visitor is a user or a session and the choice is written down. Because orders divided by visitors is conversion rateShare of visits or clicks that convert., and revenue divided by orders is average order value, revenue per visitor equals the product of the two. The revenue per visitor glossary entry explains its role as a primary metric or guardrail in experiments.
The difference between the two is who sits in the denominator. Average order value ignores everyone who did not buy, so it cannot see a change that stops people buying. Revenue per visitor counts every visitor, so it sees the net effect of a change on both how many buy and how much they spend.
Compare table
Average order value (AOV) vs Revenue per visitor (RPV)
| Row label | Average order value (AOV) | Revenue per visitor (RPV) |
|---|---|---|
| Numerator | Revenue in the period | Revenue in the period, under the same convention |
| Denominator | Orders | Visitors, as users or sessions |
| What it ignores | Everyone who did not buy, and so any change in how many buy | Nothing about conversion, but it cannot say whether a change came from order size or order count |
| Sensitivity to outliers | High: a few very large orders pull the mean up | Higher still in tests, because most visitors contribute nothing and a few contribute a lot |
| Use as primary metric | Rarely: it can rise while revenue falls | Strong for commercial tests: pricing, offers, thresholds, recommendations |
| Use as diagnostic | Strong: it explains whether a change in revenue came from basket size | Weak alone: it shows the net effect, not the cause |
| Effect of discounts | Usually lowers it, even when order count rises | Shows whether the extra orders outweighed the smaller baskets |
| Effect of bundles | Usually raises it, even when some buyers are put off | Shows whether the larger baskets outweighed any lost orders |
The rows that matter most are what each ignores and the effect of discounts and bundles. Almost every change that moves one part of revenue per visitor moves the other part in the opposite direction, and only the combined metric shows which effect won.
When to choose Average order value (AOV)
Use average order value to understand the basket. It answers questions about what buyers do once they have decided to buy: whether a cross-sell is being taken, whether a bundle is chosen over single items, whether a delivery threshold is pulling baskets up to it. Those are real questions, and a merchandising team needs the answer to them.
Average order value is also the right diagnostic when revenue per visitor moves. If revenue per visitor rose, the first thing to know is whether conversion rate, order value or both changed, and splitting the result into its two factors answers that. A rise driven by order value suggests the basket levers are working; a rise driven by conversion suggests the change made buying easier or more attractive.
Where average order value misleads is as the goal. Almost every lever that raises it can also lower conversion: a higher threshold, fewer discounts, bundles in place of single items. A test judged on average order value alone will tend to pick the variant that sheds its smaller buyers, and it will call that a win.
When to choose Revenue per visitor (RPV)
Use revenue per visitor as the primary metric for any store test or merchandising change whose purpose is commercial. Pricing displays, offers, delivery thresholds, product recommendations and checkout changes all touch both conversion and order size, and the business cares about the combined effect. The A/B testing guide covers how to choose the unit, the primary metric and the guardrails for an A/B TestControlled comparison of two variants., and names revenue per visitor as a guardrail when a test is judged on something else that could harm revenue.
Revenue per visitor is also the fairer basis for comparing traffic sources. Two channels with the same conversion rate can be worth very different amounts per visit if one brings larger baskets, and two with the same average order value can differ widely in how many visitors buy.
The cost is noise. Because most visitors spend nothing and a few spend a lot, revenue per visitor varies far more than conversion rate, and a test that makes it primary needs more traffic to reach a clear answer. The sample size and significance guide covers that planning, and the guide to writing hypotheses warns against using revenue per visitor as the primary metric for small, local changes, such as a form field, whose effect it would bury.
A ceramics shop tests a delivery threshold
Agatequill Ceramics, an online pottery shop, tests raising its free-delivery threshold from £40 to £60 against the existing threshold, with 20,000 visitors in each variant. In the control, 600 visitors buy at an average order value of £48, so revenue per visitor is £1.44. In the new variant, 480 visitors buy at an average order value of £57, so revenue per visitor is £1.37. Average order value rose by £9, which the merchandising team had set as its success measure, but revenue per visitor fell, because the higher threshold put off more buyers than it lifted. The shop keeps the £40 threshold, adds a prompt in the basket suggesting a small item to reach it, and makes revenue per visitor the primary metric for every pricing and delivery test that follows.
Decision table
| Situation | Use | Why |
|---|---|---|
| Testing a delivery threshold, bundle or price display | Revenue per visitor as primary | These changes move conversion and basket size in opposite directions |
| Explaining why revenue per visitor rose or fell | Average order value and conversion rate | The two factors show where the change came from |
| Judging a cross-sell widget once buyers reach the basket | Average order value, with revenue per visitor as guardrail | The widget acts on buyers, but must not cost orders |
| Comparing traffic sources on value | Revenue per visitor | It combines how many buy with how much they spend |
| A small change to a form or button | Conversion rate as primary, revenue per visitor as guardrail | Revenue per visitor would bury a small effect in noise |
| Planning test traffic | Whichever metric is primary, decided before launch | Revenue per visitor needs more visitors to read clearly |
Choosing the primary metric for a store test
Decide the primary metric before the test starts, in the hypothesis, and size the test for it. The question to ask is what the change could plausibly affect. If it can only act on people who are already buying, such as a recommendation shown in the basket, average order value can lead, with revenue per visitor as a guardrail to make sure no orders were lost. If it can change whether people buy at all, which covers most pricing, offer and delivery changes, revenue per visitor must lead, because average order value cannot see the people who left.
Then fix the conventions. Use the same revenue rule in both variants, including how discounts, delivery charges and refunds are treated, and use the unit the test randomises on as the visitor; the sessions versus users comparison sets out how the two visitor counts differ. Large orders can swing either variant, so agree in advance how outliers will be handled rather than deciding after the result is in. The comparison of A/B and multivariate testing covers which design suits a change, and the conversion rate calculator gives the first factor of revenue per visitor once orders and visitors are counted. Report the result as all three figures, revenue per visitor first, then conversion rate and average order value as its explanation.
Common mistakes
- Declaring a test won on average order value while conversion fell and revenue per visitor with it.
- Using average order value as a merchandising team's only target, which rewards shedding smaller buyers.
- Mixing users and sessions as the visitor between reports or between variants.
- Making revenue per visitor primary for a tiny local change, then reading noise as a result.
- Ignoring outliers, so one large order decides which variant wins.
- Changing the revenue convention mid-test, such as switching between gross and net of discounts.
- Reporting revenue per visitor without its two factors, so nobody can tell why it moved.
