Average Order Value
Date: 2026-08-16
Revenue divided by orders. Useful, universally quoted, and computed as a mean on a heavily skewed distribution — so it describes a customer who doesn’t exist and moves on a single trade order.
What it is
Average order value (AOV) is total revenue divided by order count for a period.
£750,000 revenue ÷ 15,000 orders = £50.00
The mean is the wrong statistic
Order values are right-skewed — bounded at zero, unbounded above — so the mean sits well above the typical order:
£22 £24 £28 £31 £35 £38 £42 £48 £55 £2,400
mean £272.30
median £36.50
Report both. The gap between them is the shape, and one number alone hides it. On a real catalogue the divergence is smaller but always present — Skewed and Heavy-Tailed Distributions, Mean Median and Mode.
Consequences that bite:
- A single trade order moves a monthly AOV enough to look like a trend
- AOV as a test metric is unstable, because the tail lands in one arm or the other by chance — cap it, and pre-register the cap — Winsorisation and Capping
- “Increase AOV by 10%” as a target invites tactics that shift the mean without adding contribution
Revenue AOV isn’t the useful one
What matters commercially is contribution per order, not revenue per order:
AOV £50.00
contribution per order £15.00 (30%)
An AOV increase driven by discount-led bundling can raise revenue and lower contribution. Two orders of £50 at 30% contribution beat one order of £100 at 10%. Track contribution per order alongside — Contribution Margin.
Raising it
Ordered roughly by how well they hold up commercially:
- Free shipping threshold just above current AOV. The most-used lever in ecommerce and it works — provided the threshold clears the fulfilment cost — Shipping Thresholds
- Cross-sell of genuinely complementary items. Adds units without discounting
- Bundles priced above the sum of individual contribution, not below — Bundling
- Volume incentives on consumables, where the customer would have repurchased anyway. Careful: this pulls forward revenue rather than adding it
- Range and merchandising. Promoting higher-value lines shifts mix, and it’s the lever with no discount cost — Merchandising
The ones that raise AOV and reduce profit: threshold-driven padding with low-margin items, “spend £X get £Y off”, and anything that trains customers to wait for a promotion — Discounting Strategy.
Where it interacts
AOV and conversion rate trade off. Pushing customers to spend more usually converts fewer of them. The metric that captures both is revenue per visitor:
revenue per visitor = conversion rate × AOV
= 3% × £50 = £1.50
Optimise revenue per visitor, not AOV — it’s the product, and it can’t be gamed by trading one against the other. The cost is that it’s the noisier metric to test on, needing roughly 2.4× the traffic of conversion rate — Metric Sensitivity, Revenue Metrics.
Segmenting it
Blended AOV hides most of what’s useful:
- New versus returning. Returning customers usually spend more, so a change in acquisition mix moves AOV with no behavioural change
- By channel. Paid social and branded search bring different basket sizes
- By device. Mobile AOV is typically lower, so a mobile share shift moves blended AOV — Simpson’s Paradox
- By category — Basket Composition
A falling blended AOV with every segment flat is a mix change, and it’s the most common reason the number moves without anyone doing anything.