Cart abandonment recovery is one of the highest-return automations in ecommerce, and it is frequently set up once with default settings and never examined. The defaults are usually wrong on timing and almost always wrong on discounting.
Why carts are abandoned
Understanding the reasons determines what the recovery message should say.
Unexpected cost at checkout. Consistently the most common reason. Shipping, tax or fees appearing late.
Just browsing. Using the cart as a shortlist with no immediate intent to buy. A large share of abandonment and largely unrecoverable in the short term.
Comparison shopping. Checking the total elsewhere.
Checkout friction. Forced account creation, too many fields, a payment method not offered.
Interruption. A genuine intent that was interrupted. This is the most recoverable group, and the reason the first message should arrive quickly.
The three-message sequence
Message one, one hour after abandonment. A simple reminder with the items shown, a clear link back to the filled cart, and nothing else. No discount, no urgency. For the interrupted buyer this is all that is needed, and it recovers the largest share of the total.
One hour is the timing most often set wrong. Many default configurations wait 24 hours, by which time the interruption group has either bought elsewhere or lost the thought.
Message two, 24 hours later. Address the likely objection. Shipping cost and delivery timing stated plainly, returns policy, a reassurance about security or quality. Reviews of the specific items in the cart work well here. Still no discount.
Message three, 48 to 72 hours after that. The last message. This is where a discount belongs if you use one, or a genuine stock or price notice if one applies. Make it the final contact and say so.
Stop after three. A fourth message produces complaints and unsubscribes that cost more than the marginal recovery.
When to discount, and when not to
Discounting in abandonment emails is the most commonly over-used tactic in ecommerce, and it has a specific failure mode: customers learn the pattern. Once a segment of your audience knows that abandoning a cart produces a 10% code, abandonment becomes a deliberate behaviour and you are discounting sales you would have made at full price.
Three rules that limit the damage. Never discount in the first message. Discount only in the final message, and not every time. And prefer a non-price incentive where one works — free shipping, a free return, an extended warranty — because these cost less and are less easily gamed.
Segment where you can: a first-time visitor abandoning a cart is a different proposition from a repeat customer who buys monthly and does not need an incentive.
Capturing the address before the abandonment
None of this works without an email address, and most abandoners are not logged in.
Three approaches. Ask for the email early in checkout, as the first field, before the address and payment details — this alone makes a large share of abandoners recoverable. Offer an obvious saved-cart or email-my-basket option. And for identified returning visitors, match on an existing cookie where consent allows.
Exit-intent popups also capture addresses and are intrusive; if used, keep them to one per visitor and make them genuinely dismissible.
Browser and SMS channels
Email is the primary channel and not the only one.
Browser notifications reach people who never gave an address, with much lower consent rates and no inbox competition. Useful as a supplement where the opt-in exists.
SMS has very high open rates and a correspondingly low tolerance for overuse. Where you have explicit consent, a single SMS in place of the second email can perform well. More than one is intrusive and in many jurisdictions tightly regulated.
Retargeting advertisements cover people you cannot message directly. See retargeting strategy for the frequency and exclusion rules, particularly excluding people who completed the purchase.
Prevention is worth more than recovery
A recovery sequence recovers a minority of abandoned carts. Fixing the cause prevents the abandonment entirely, and the causes are known.
Show the full cost including shipping on the product page or early in the basket. Offer guest checkout. Reduce the field count. Offer digital wallets. Make delivery timing visible before checkout.
These changes reduce abandonment at the source and improve the experience for everyone, including the majority who would have completed anyway. Our notes on ecommerce conversion cover the checkout work in more detail.
Measuring it honestly
Recovery rate reported by most platforms counts anyone who received an email and later purchased, which includes people who were always going to return.
The honest measure is a holdout: withhold the sequence from a small random share of abandoners and compare purchase rates. The incremental figure is usually lower than the attributed one, and knowing the real number tells you how much the discount in message three is actually costing you.
Writing the messages
Recovery emails are mostly written as marketing and perform better written as service.
Show the items. Images, names, prices, exactly as they were in the basket. The person may not remember what they were looking at, particularly 48 hours later.
One link, going to the restored cart. Not the homepage, not the product page — the filled basket, ready to complete. Every additional step loses people.
Subject lines that describe rather than tease. “Your basket is saved” outperforms curiosity-gap phrasing, because the recipient recognises what it is and why it arrived.
Keep it short. A recovery email is a reminder, not a sales page.
Include support contact. A proportion of abandonment is an unanswered question, and offering a way to ask converts some of them.
What the abandonment data tells you
Beyond recovery, abandonment behaviour is one of the better diagnostic signals a store produces, and almost nobody reads it as one.
Look at where in the checkout people leave. A spike at the shipping step means delivery cost or timing is the problem. A spike at payment means a method is missing or the form is failing. A spike at account creation is self-explanatory.
Look at which products are abandoned disproportionately. High abandonment on one item usually means the product page is not answering something — size, compatibility, delivery time.
And look at abandonment by device and by new versus returning visitor. These splits turn a general complaint about abandonment into two or three specific fixes, which is worth considerably more than tuning the recovery sequence.
Setting the sequence up correctly
Most platforms ship a default sequence, and the defaults are rarely right. Four settings worth checking before anything else.
The delay on the first message — change it from 24 hours to one. The number of messages — three, not one and not six. Whether a discount is included in the first message — it should not be. And the exclusion rules, particularly whether someone who completes a purchase is removed immediately, which surprisingly often they are not.
Then send yourself through the whole sequence by abandoning a real cart. Reading the messages as a customer finds the broken links, the wrong product images and the discount code that does not work far faster than any checklist.
Frequently asked questions
What is a typical cart recovery rate?
A well-configured sequence commonly recovers a low double-digit percentage of abandoned carts that it can reach. The share of all abandoned carts recovered is lower, because most abandoners cannot be contacted.
How soon should the first email go out?
Within an hour. The interrupted-buyer group is the most recoverable and the most time-sensitive.
Should every abandonment trigger the sequence?
Exclude people who completed a purchase, those who abandoned a very low-value cart where the email is not worth the deliverability cost, and anyone who has received the sequence recently.
Do discounts in abandonment emails train bad behaviour?
Yes, if used consistently. Reserve them for the final message, use them selectively, and prefer non-price incentives where they work.
