Cart Abandonment: The Metric and the Money Behind It

The formula, and the denominator that decides everything

Abandonment rate = 1 − (completed orders ÷ carts created). Ten thousand carts and 2,500 orders give a completion rate of 25% and an abandonment rate of 75%. The numerator is never in dispute — an order is an order. The denominator is where two shops running the same business can report figures ten points apart, because there is no universal definition of a cart being created. Some platforms count the moment an item enters the basket, including baskets a bot filled. Some count only sessions that reached the checkout page. Some count each returning visit to a persistent basket as a new cart, so one shopper who thinks about a purchase across four evenings becomes four carts and three abandonments.

So the first thing to do with the number is not to improve it but to define it. Write down what your analytics counts as a cart, whether bot traffic is filtered, and whether a persistent basket generates a new cart per session. Until that is written down, a month-on-month movement of two points tells you nothing — it might be a checkout change, or it might be a crawler. It is the same discipline that keeps a funnel conversion rate honest: the metric is only as meaningful as the event that starts it.

The headline figure is a benchmark, not a target

Baymard Institute maintains a running aggregation of published cart-abandonment studies and reports an average of roughly seven in ten. That is a real, sourced, useful number and it is also the most misused statistic in e-commerce, because it averages businesses that have nothing in common. A grocery basket that a household edits over three days, a travel booking that requires comparing dates across four sites, a $2,000 sofa and a $12 phone case do not share a shopper, a consideration period or a checkout. The studies Baymard collects individually span a very wide band; the average sits in the middle of a distribution, not at the location of any real shop.

The correct use of a benchmark like that is diagnostic, not aspirational. If your rate is wildly outside the band, something is probably broken or miscounted, and it is worth looking. If it is inside the band, the benchmark has told you everything it can and the only remaining comparison worth making is against yourself last month, on a definition you have not changed. Chasing an industry average is chasing the average of businesses you are not in.

Turning percentage points into money

This is the calculation the metric exists for. Ten thousand carts a month, 2,500 completing, an $80 average order: completed revenue is 2,500 × 80 = $200,000. The 7,500 abandoned carts carry a nominal $600,000, and that headline is where most recovery projects go wrong — it is not money you lost, it is money that was never committed. The honest question is marginal. One percentage point of completion on 10,000 carts is exactly 100 orders, and 100 orders at $80 is $8,000 of revenue. Three points is $24,000. That is a number you can put next to the cost of the engineering work, which the $600,000 never was.

Some abandonment is not a leak at all

A basket is the cheapest bookmark on the internet, and shoppers use it as one. They add a coat to see whether it will fit the budget next month. They fill a cart on three sites to compare the delivered price, then buy on one. They add items to reach a free-shipping threshold and then remove them. None of those sessions was ever going to end in an order today, and no checkout redesign will convert them. They are behaviour, not friction.

The arithmetic of that is worth doing once. Suppose 30% of your 10,000 carts were never purchase intent. The intent denominator is 7,000, and abandonment among people who actually meant to buy is 1 − 2,500 ÷ 7,000 = 64.3%, not 75%. Ten points of your reported abandonment were never yours to recover. You cannot measure that fraction directly, but you can bound it — carts abandoned before the shipping step behave very differently from carts abandoned at the payment step, and separating the two tells you which of your abandonment is browsing and which is friction.

A recovered cart that comes back as a return is worth far less

Recovery programmes are almost always valued in revenue, and revenue is the wrong unit. Start with the $8,000 that 100 recovered orders generate. At a 45% gross margin an $80 order costs you $44 in goods and leaves $36 of contribution. Now apply a return rate — take 20% as an assumption for the worked example, since online return rates vary enormously by category and the National Retail Federation's annual survey shows retail returns running materially higher online than in store. Twenty of the 100 orders come back: the revenue reverses, and you still pay to process them, say $6 each in shipping and restocking. Expected contribution per recovered order is 0.80 × 36 − 0.20 × 6 = 28.80 − 1.20 = $27.60.

So the honest value of recovering 100 carts is $2,760, not the $8,000 headline — about a third. Run that through the whole ladder and a shop moving from 25% to 30% completion gains 500 orders, $40,000 of revenue and $13,800 of contribution. Both numbers are true; only the second one pays for anything. And the effect compounds badly if recovery skews toward hesitant buyers, who are exactly the population most likely to return, so the average return rate on recovered orders will usually be worse than your site-wide rate rather than equal to it.

The discount that has to lift recovery by 30% just to break even

The reflex response to abandonment is a code in the recovery email, and the reflex is expensive because a discount comes off contribution, not off revenue. Your $80 order has $44 of goods in it. A 10% code makes the order $72, so contribution falls from $36 to $72 − $44 = $28. That is a 22.2% cut in what the order earns you, from a 10% cut in the price. After the same 20% return rate and $6 handling, expected contribution per recovered order falls from $27.60 to 0.80 × 28 − 0.20 × 6 = 22.40 − 1.20 = $21.20.

That gives you the test the campaign has to pass. Sending the code to everyone breaks even only if it lifts the recovery rate by 27.60 ÷ 21.20 − 1 = 30.2%. Not thirty percentage points — thirty percent more recovered carts than the same email without a code. That is a demanding bar, and the only way to know whether you clear it is a holdout: send the plain reminder to a random half and the discounted one to the other half, and compare contribution per abandoner rather than recovery rate. The companion piece in this series on sizing an A/B test covers how many abandoners you need before that comparison means anything.

Frequently asked questions

What is a good cart abandonment rate?

There is no single good value, because the rate depends on category, price point, consideration period and how your platform defines a cart. Baymard Institute's aggregation of published studies averages around seven in ten, but that average spans grocery, travel, luxury and marketplaces. Use it to check that you are inside the normal band; after that, the only comparison worth making is against your own previous month on an unchanged definition.

Is the value of my abandoned carts really lost revenue?

No. In the worked example the 7,500 abandoned carts carry a nominal $600,000, but that figure assumes every one of them would otherwise have converted, which is false by construction — if they all converted your abandonment rate would be zero. The number you can act on is marginal: one point of checkout completion is 100 orders and $8,000 of revenue, of which $2,760 survives as contribution after margin and returns.

How much is one percentage point of checkout completion worth?

Multiply your cart count by one percent to get extra orders, then by your average order value for revenue, then by contribution per kept order for money you keep. On 10,000 carts that is 100 orders; at an $80 average order it is $8,000 of revenue; at a 45% margin with a 20% return rate and $6 of handling per return it is 100 × $27.60 = $2,760 of contribution. Compare that last number, not the first, with the cost of the work.

Should my recovery email carry a discount code?

Only if it clears a measured bar. A 10% code on an $80 order with $44 of goods cuts contribution from $36 to $28, and after a 20% return rate with $6 of handling the expected contribution per recovered order falls from $27.60 to $21.20. The code therefore has to lift recovery by 30.2% just to break even. Test it against a plain reminder on a random holdout and judge on contribution per abandoner, not on recovery rate.

What is the difference between cart abandonment and checkout abandonment?

Cart abandonment measures everyone who added an item and did not order, which mixes browsing with friction. Checkout abandonment measures only people who started the checkout flow and did not finish, which is almost entirely friction — they had declared intent. The second is the number worth optimising, because that is where form length, shipping cost disclosure, guest checkout and payment failures live. Report both, and never quote one as the other.