What is AOV?

A store takes A$96,000 across 800 orders in a month, so average order value is 96,000 ÷ 800 = A$120. Order values in that month ranged from A$29 to A$1,400, and the median order was A$78. The average is being pulled up by a small number of large orders, which is normal and is why decisions about free-shipping thresholds and bundles should be made against the distribution rather than against the A$120.

Average order value divides revenue by the number of orders that produced it. There is nothing subtle about the arithmetic, and almost every other number in an ecommerce account depends on it.

It is the revenue half of return on ad spend. It sets the ceiling on what an acquisition can cost. It determines whether a shipping subsidy is affordable. It is one of the four terms in a lifetime value calculation. A store that raises its average order by a fifth improves its return on ad spend by a fifth with no change in the advertising at all, which is a larger effect than most campaign work produces and is available to anyone willing to do merchandising rather than media.

That connectedness is also why the figure has to be computed carefully. Revenue after discounts, with shipping and tax treated the same way every period, divided by orders rather than by customers or by sessions. Small inconsistencies here propagate into every metric downstream.

Order values are right-skewed in essentially every store. Most orders cluster in a band near the low end and a thin tail of large orders pulls the mean above it, sometimes well above it.

The consequence is practical rather than theoretical. Decisions that depend on this metric — where to set a free-shipping threshold, what to bundle, what to offer as an upsell, which customers to treat as high value — are decisions about moving a group of real orders, and the mean frequently sits in a sparse part of the distribution where there are very few orders to move.

A store with a A$120 average and a A$78 median has most of its orders in the seventies and eighties. A shipping threshold set at A$150 asks those customers to almost double their basket, which few will. A threshold set at A$95 asks them for one small addition, which many will, and moves a much larger group.

Plotting the distribution once, rather than tracking the average forever, is the highest-value hour available on this metric.

The levers are merchandising and pricing rather than media, and they divide into four groups.

Thresholds. A free-shipping or free-gift threshold set just above the dense part of the distribution is the most reliable of them. It works because it converts a vague desire to spend less into a concrete reason to spend slightly more.

Bundles and quantity breaks. Selling two or three of a consumable together, or pairing a product with its obvious companion, raises order value without discounting the individual item. Bundles also protect margin better than percentage-off promotions, which train customers to wait.

Upsells and cross-sells. Placement matters more than the offer: a relevant addition on the product page or in the cart converts far better than one after checkout, and an irrelevant one suppresses conversion rate while adding nothing.

Product mix. Which products the advertising promotes determines the order value it produces. A campaign pointed at an entry product and one pointed at a flagship will report very different averages and very different returns, and neither is a fact about the advertising.

Every mechanism that raises order value can lower the number of orders, and the two effects have to be read together.

A higher shipping threshold moves some baskets up and abandons others. An aggressive upsell adds revenue to some orders and adds friction to all of them. A minimum order value protects margin and turns away small customers who might have become large ones.

The figure to watch through these tests is revenue per session or contribution per session, which holds both effects at once. Optimising average order value in isolation is one of the easier ways to make a store smaller and more profitable per order at the same time, and whether that is the right outcome is a business decision rather than a metric one.

Media does not set order value directly, and it influences it in three ways that matter.

The first is which products get promoted. Catalogue campaigns will happily sell whatever is cheapest to sell, which is usually the lowest-priced product in the feed. Constraining the feed, or splitting campaigns by price band, puts that decision back in the merchandiser's hands.

The second is which customers arrive. Broad prospecting audiences buy differently from retargeting audiences, and discount-led creative attracts customers whose first order is small and whose second one may not exist. Reporting order value by campaign and by audience usually shows a spread wide enough to change the budget split.

The third is the offer in the ad. A first-order discount raises conversion rate and lowers order value; a threshold offer — spend this much, get that — raises order value and can lower conversion rate. Both are advertising decisions with merchandising consequences.

Compute it after discounts. A figure based on list prices describes transactions that did not occur and will hide the effect of exactly the promotions it should be measuring.

Decide once whether shipping revenue is in or out, and never change it mid-year. Both are defensible; switching between them turns a flat trend into a step change that means nothing.

Divide by orders, not customers. Revenue per customer is a useful number and a different one, and conflating the two makes a store with strong repeat purchase look as though it has high order values.

Segment before concluding. By channel, by device, by new against returning customer, and by whether a discount was used. The blended average is the least actionable version of this metric and the one most often quoted.

The chain that connects this metric to what can be bid is short and worth stating explicitly.

Order value multiplied by gross margin gives contribution per order. Contribution per order is the most a first purchase can cost to acquire before the transaction loses money. Where customers return, the ceiling is that contribution repeated across the relationship, discounted for the fact that future contribution is less certain and less useful than contribution today.

Raising order value raises every link in that chain proportionally. It is the rare improvement that makes the advertising better without touching the advertising, which is why it belongs in a media review rather than only in a merchandising one.

• Look at the distribution before the average. Order values are skewed in almost every store, and a threshold or a bundle set at the mean will miss the cluster of orders it was meant to move.
• Set the free-shipping threshold just above the densest part of the distribution, not above the average. The point is to move a real group of orders up a small step, not to set a target nobody is near.
• Test bundles and quantity breaks before discounting. Both raise order value without training customers to wait for a sale, which discounts do.
• Segment by acquisition channel. Branded search and retargeting routinely produce higher order values than broad prospecting, and a single blended figure will misprice both.
• Check margin as well as value when order value rises. An increase driven by a low-margin add-on can raise the number and lower the contribution behind it.