Average Order Value (AOV)

What is average order value?

Average order value describes order size. Retailers use it to understand basket composition, merchandising, discounting, shipping thresholds, bundles, and checkout behavior. AOV can rise because customers add more items, choose higher-priced products, buy a longer subscription, or because smaller orders stop converting. Those explanations have very different commercial implications.

AOV is not average revenue per customer, revenue per visitor, gross margin, or lifetime value. A customer who places four $50 orders contributes four orders to AOV; a visitor who never purchases does not appear in the metric at all. That narrow denominator is both its strength and limitation: it isolates spending conditional on order, but it can hide a drop in purchase frequency or conversion.

Use AOV as a diagnostic and decision metric when its definition is explicit. State the currency, observation period, order status, revenue components, and whether the number is gross or net of returns. Without those rules, an AOV trend can be a finance-policy change rather than a customer behavior change.

AOV formula and revenue choices

The standard formula is:

AOV = qualifying order revenue / number of qualifying orders

“Qualifying” must apply equally to both sides of the fraction. Most product teams use completed or paid orders and exclude cancelled orders. The revenue numerator may be gross merchandise value, item revenue after discounts, or net revenue after returns and credits. There is no single correct convention, but mixing conventions makes comparisons invalid.

For recurring products, distinguish first-order AOV from average invoice value. For marketplaces, decide whether AOV refers to buyer payment, merchant merchandise value, or platform revenue. Currency conversion should use a consistent method, especially when experiment allocation varies by country.

AOV in A/B testing

AOV is useful for experiments involving product recommendations, cart design, minimum order thresholds, bundles, promotions, delivery pricing, payment options, and package selection. It is usually a secondary metric or guardrail rather than the sole decision criterion. A design that encourages larger baskets can be beneficial; a design that prevents low-value purchases may also raise AOV while reducing total revenue.

Read AOV alongside conversion rate, orders per visitor, revenue per visitor, contribution margin, return rate, and customer satisfaction. A practical identity is:

revenue per visitor = conversion rate × orders per purchaser × AOV

That decomposition makes trade-offs visible. If treatment AOV rises 8% but conversion falls 12%, the business outcome may be negative. Conversely, a lower AOV can accompany better revenue per visitor if more people complete an order. Set the primary metric and decision rule before launch, using the principles in primary versus guardrail metrics.

For an A/B test, estimate AOV at the visitor or assigned-user level when possible rather than treating orders as independent observations. One purchaser can create multiple correlated orders. Revenue distributions are often skewed, so confidence intervals, resampling, robust methods, or a pre-specified transformation may be more informative than a simple order-level average. See how log transforms affect revenue metrics for the associated trade-offs.

Worked AOV calculation

An online retailer tests a “complete the look” module. In a seven-day, mature-return snapshot, control receives 20,000 visitors, produces 1,000 completed orders, and records $82,000 in item revenue after discounts. Treatment receives 20,000 visitors, produces 1,080 completed orders, and records $85,320.

control AOV = $82,000 / 1,000 = $82.00
treatment AOV = $85,320 / 1,080 = $79.00

Treatment AOV is $3 lower, or 3.7% lower. But revenue per visitor is $4.10 in control ($82,000 / 20,000) and $4.27 in treatment ($85,320 / 20,000), an increase of $0.17 per visitor. The module produced more orders and more total item revenue despite smaller baskets. The decision should additionally consider incremental fulfillment cost, product margin, returns, and whether the estimated revenue-per-visitor interval supports a real improvement.

Data-quality caveats

Order data rarely lives in one pristine table. Ecommerce events can record an order at checkout, payment authorization, capture, fulfillment, and settlement; each gives a different count. Pick the business event appropriate to the decision and join adjustments with a stable order ID. Deduplicate retries, split shipments, partial captures, and payment-provider webhooks.

Returns create a timing problem. A test that ends today may appear to raise net AOV because treatment orders have not yet had time to return. Lock a common maturation window, such as 30 days after purchase, for final analysis. If speed matters, report gross AOV promptly and label it provisional, then monitor return-adjusted AOV as a guardrail.

Check whether variants change exposure to currencies, territories, customer segments, or inventory availability. An imbalance in high-price products, marketing channels, or stock-outs can change AOV independently of the feature. Keep tax, exchange-rate, discount, and order-status rules identical across arms.

Practical guidance

1. Choose gross or net AOV to match the decision, then write the order-status and adjustment rules before analysis.
2. Pair AOV with conversion and revenue per visitor; add margin where product mix or incentives differ.
3. Segment by new versus returning buyers, category, country, and device only when those cuts are pre-planned or clearly labeled exploratory.
4. Inspect the distribution and percentile changes. A mean shift from a handful of enterprise orders is different from a broad basket increase.
5. Allow returns and cancellations to mature before declaring a financial win.

Common mistakes

• Calling a higher AOV a revenue win: the denominator omits non-purchasers.
• Mixing gross and net definitions: returns, discounts, and taxes can reverse a comparison.
• Using order-level significance blindly: repeated orders from a customer are correlated.
• Ignoring margin: an expensive low-margin item can lift AOV while harming contribution.
• Comparing immature cohorts: unequal return windows bias net AOV.

Frequently asked questions

Is AOV the same as average basket size?

Not necessarily. Basket size may mean units per order, while AOV is monetary value per order. Product price and discounting can make them move differently.

Should taxes and shipping be included in AOV?

Usually exclude taxes. Include shipping or fees only if they represent the commercial behavior you want to measure, and use the rule consistently.

Why can AOV fall when revenue rises?

More visitors may purchase, or more purchasers may place smaller orders. Revenue per visitor and total revenue incorporate those extra orders; AOV does not.

How do refunds affect AOV?

They lower net AOV when included. Use a consistent return window and avoid comparing groups with different time since purchase.

What is a good AOV?

Benchmarks vary by category, geography, catalog, and traffic mix. A useful target is profitable, sustainable revenue per visitor, not a generic AOV number.

Summary

Average order value measures the monetary size of qualifying orders, not the value of all visitors or customers. Define revenue and order status precisely, then evaluate AOV with conversion, revenue per visitor, margin, and returns. In experimentation, a lower AOV can be the correct outcome when it produces more profitable purchases overall.

Sources

• Shopify: Average order value
• HM Revenue & Customs: VAT rules on discounts and rebates
• U.S. Census Bureau: Monthly Retail Trade

Related terms

• Metric
• Key Performance Indicator (KPI)
• Primary Metric
• Secondary Metric
• Guardrail Metric