A recovery email cannot repair what the checkout broke, so start with the step the shopper left on
Roughly seven in ten online carts never become an order. Baymard Institute puts the documented average at 70.22%, calculated across 50 studies. A common answer to that loss is a fixed reminder sequence, for example an email after an hour, a second one the next day and a discount in the third. In a Practical Ecommerce column published on 6 August 2026, Armando Roggio argues that this fixed pattern is the weak point of most recovery programmes. An abandoned cart is evidence of a lost sale with no explanation attached, and one sequence for every cart gives the same answer to problems that have little in common. The useful question is therefore not which email to send, but what you already know about why each cart stopped. For one group of carts, you know a great deal.
Failed payments are the carts you can read without guessing
Most abandoned carts leave you to infer the cause. A failed payment does not. A declined card, a bank payment that was started but never approved, a shopper who returned from the banking app to an error page: each of these shows up in your payment status as a failed or cancelled attempt, linked to an order and usually to an email address. That shopper had already chosen the products, entered an address and clicked to pay. The decision to buy was made. Something in the final step stopped it.
Baymard’s survey of US online shoppers gives a sense of how often the final step itself is the problem: 10% say they have abandoned a purchase because their card was declined, and 9% because there were not enough payment methods. Those shoppers are not hesitating. Treating their cart as doubt, with product copy or a sale offer, misreads what happened and spends margin on a problem that price did not cause.
The recovery for these carts is practical. Send a link that reopens the same order at the payment step, with the items, the delivery address and the delivery choice already filled in, rather than a link to the homepage or the product page. Put a different payment method on that page, because the method that failed once can fail again for the same reason. Keep the wording plain: the payment did not go through, the order is being held, this is how to finish it. Send it soon, while the purchase is still on the shopper’s mind. A discount has no place in this message.
What a reminder can fix and what it cannot
Roggio separates three situations. Some shoppers got stuck in the mechanics of buying. Some still had doubts about the product itself or about what happens after the order, such as delivery and returns. Some simply found the deal unattractive by the time they saw it in full. Only the first group can be helped by putting the order back in front of them. The second needs answers, and the third needs a different deal. He also warns that a standard series tends to reward shoppers with a discount they never needed, because it cannot tell a shopper who was coming back from one who was not.
The same survey shows how many carts a reminder cannot rescue cheaply. Set aside the 42% who say they were only browsing, and the leading reason for leaving is extra costs for shipping, tax or fees (40%). Further down, 18% were put off by having to create an account, 13% by the returns policy and 12% by not seeing the total cost early enough. Each of these is a decision made in the checkout and the product page. A sequence that wins those carts back usually does it with a discount, which costs more than showing the shipping charge before the cart would have done. Roggio’s own first step is prevention, beginning with prices and delivery costs the shopper can see without searching.
So fix those causes at the source. Show delivery costs and the total on the product page and in the cart, make guest checkout the default route, and put the returns terms next to the buy button. What is left afterwards is a smaller group of abandoned carts, and one that is easier to sort by cause.
Measure recovery against carts that received nothing
Once the carts are sorted, give each group its own message: a restore message for failed payments and interrupted sessions, an informative one for carts that stopped before the payment step, and a price offer only where price was the likely obstacle and the margin on that product allows it. Then check whether any of it works. Roggio recommends comparing sequences on revenue per recipient and keeping back a control group that gets no message at all. Without that group, every order placed after a reminder counts as recovered, including orders that would have arrived anyway. With it, you see what each message adds. If the restore message clearly beats the control group and the discount message barely does, the discount is mostly paying shoppers who were already on their way back.
Roggio also discusses software that chooses the message per cart from options the merchant has approved and learns from the outcomes. He is clear that such a system cannot see the shopper’s motive either; it estimates from what worked with comparable shoppers. That makes a hard signal such as a failed payment attempt more valuable, not less. Whether a person or a model picks the message, it is one of the few facts you have about why a cart stopped.
Start with last month’s abandoned checkouts. Tag each one by where it stopped: before the delivery step, at the payment step, or after a failed payment attempt. Then look at what your current sequence sent to each group. Wherever a failed payment received the same first email as a browsing shopper, change that message first, and give it a link straight back to the payment step with a second method on offer.
Sources: practicalecommerce.com and baymard.com
