Abandoned Cart Recovery: Sequences, Consent and the Number That Is Usually Wrong
Cart recovery is sold as free money: the customer already chose the product, so all that remains is a reminder. The framing is appealing and it quietly skips two questions that decide whether the programme is worth building. Whether you are allowed to send the message at all, and whether the revenue you attribute to it would have arrived without you.
Both have answers. Neither appears in the vendor material, which is unfortunate, because one of them is a legal exposure and the other is the difference between a real result and a reported one.
Most abandoned carts were never orders
Averaged abandonment sits around 70%, and the instinct is to read that as 70% of revenue waiting to be reclaimed. It is not. A large share of abandonment is people using the cart as a shortlist, checking a delivery cost, or comparing a total against another shop. Those sessions were never going to convert in that visit and no email changes it.
The recoverable portion is the group who intended to buy and hit an obstacle: an unexpected shipping cost, a forced account creation, a payment method they do not have, a moment of doubt about returns. Which is why the first and largest gain is almost never in the recovery sequence. It is in removing the obstacle, and the obstacles are well documented in the Baymard checkout data.
A useful discipline before building any recovery flow: read last month’s abandonment reasons, if you collect them, and fix the top one first. A recovery email is a patch applied after the failure. It is cheaper to not fail.
The legal question, which comes first in the EU
An abandoned cart email is marketing. Its purpose is to persuade someone to make a purchase they have not made, which is different from servicing a purchase they did make. That classification matters, because marketing email to individuals in the EU is governed by the ePrivacy rules on top of the GDPR.
The mechanism most vendors point to is the soft opt-in: you may email existing customers about similar products without separate consent, provided they were given a chance to object. The catch is in the wording. Soft opt-in applies where contact details were obtained in the context of a sale. An abandoned cart is, by definition, a sale that did not happen, so whether the exemption covers it is genuinely contested and interpreted differently by different national regulators.
The defensible positions, in order of safety:
- Explicit consent captured at the checkout step, with a clearly worded, unticked opt-in that covers marketing. Unambiguous, and it survives any interpretation.
- Legitimate interest with a documented balancing test, a short, easy objection route and a strict frequency cap. Defensible in several jurisdictions, and it requires you to have actually written the assessment down rather than assert it.
- Soft opt-in relied on alone. Common in practice, weakest in law, and the position that would be hardest to defend if a regulator asked.
Whichever you choose, the record of consent has to be retrievable per person and per moment, which is a data question rather than a marketing one and belongs with everything else in what you may collect and keep.
The sequence that industry practice has settled on
| Timing | Purpose | What it should contain |
|---|---|---|
| Within 1 hour | Catch the interrupted session | The cart contents, a single link back, nothing else |
| ~24 hours | Answer the doubt that stopped them | Delivery, returns, payment options, stock status |
| ~72 hours | Last useful contact | Scarcity if genuine, or an incentive if you use one |
Published benchmarks put open rates for these messages in the 40–50% range and recovery rates typically between 5% and 15%, with strong programmes reported in the high teens or low twenties. A three-message series is commonly reported as producing 60–70% more recovered revenue than a single email.
Two cautions about those numbers. They come overwhelmingly from email platform vendors summarising their own customers, which selects for businesses that run structured programmes. And open rate has been unreliable since privacy-protecting mail clients began pre-fetching images, so a 45% open rate is a soft signal rather than a measurement. Judge the flow on orders placed, not on opens.
What each message should actually say
The first message is a reminder, and it works because it is boring. Cart contents, one button, no marketing copy. The most common mistake is to make it a campaign; the person was two minutes from paying and needs a route back, not a value proposition.
The second is where the real work happens. Whatever stopped them is usually an unanswered question, so this message answers the standard four: when it arrives, what shipping costs, what happens if they want to return it, and which payment methods you accept. In markets where a specific method dominates, saying so explicitly is worth more than any persuasion, which is the same reasoning behind matching the channels and habits of the market you are selling into.
The third message is the one to consider omitting. It is also where discounts usually appear, and a discount at 72 hours has a cost beyond the margin: a customer base that learns to abandon carts deliberately. If you use an incentive, prefer free shipping over a percentage, cap it, and never send it to someone who has already bought at full price this quarter.
Attribution, and the number that is usually wrong
This is the part that separates a real programme from a reported one. When a recovery email goes out and orders follow, the flow is credited with the revenue. Some of those people were coming back anyway. They had the tab open, or they intended to buy on payday, and the email arrived in the same window.
The correction is a holdout: withhold the sequence from a small random percentage of abandoners, five or ten percent, and compare completion rates between the two groups over the same period. The difference is the incremental effect and it is the only number worth reporting upward. In practice it is consistently lower than the attributed figure, sometimes by a wide margin, and knowing by how much changes what you are willing to spend on the programme.
Keep the holdout running permanently rather than as a one-off test. Seasonality, traffic mix and creative changes all move the gap, and a measurement taken once in November describes November. This is the same class of discipline as choosing an event set that shows where money actually comes from.
What to do before the email, and after it
- Capture the email earlier, but honestly. An email field early in checkout enables recovery. Making it feel like a trap does not; a saved-cart offer is a fair exchange and a hidden subscription is not.
- Save the cart itself. A cart that survives a week and a device change recovers people silently, with no message and no legal question attached.
- Handle the on-site moment. An exit prompt that answers the delivery question is often more effective than any email, because it arrives while intent is still live.
- Decide about other channels deliberately. Messenger and SMS have higher open rates and considerably lower tolerance for repetition, and in several markets they carry stricter consent requirements than email.
- Close the loop with support. If people abandon and then ask the same question by phone, that question belongs on the product page or in the checkout, not in a follow-up.
A minimum viable programme
For a small store, the whole thing is a week of work and looks like this. Fix the top abandonment reason first. Add a clearly worded consent checkbox at the email step. Build two messages, at one hour and twenty-four hours, with the second answering delivery, returns and payment. Hold back five percent as a control. Report recovered orders against that control rather than against attribution, and review the sequence once a quarter against the current top abandonment reason, because it moves.
Everything beyond that — branching by cart value, product-specific creative, a fourth message, cross-channel orchestration — is optimisation of a mechanism that should be proven to work first, and is the point at which automation stops paying for itself if the basics are not in place.
The short version
Most abandoned carts were never orders, so the first and largest gain is in removing the obstacle rather than in messaging around it. In the EU, an abandoned cart email is marketing, and the soft opt-in exemption is contested because it applies to details obtained in the context of a sale, which by definition did not happen: explicit consent at checkout is the only position that survives every interpretation. Practice has settled on three messages at roughly one hour, twenty-four hours and seventy-two hours, with reported open rates of 40–50% and recovery rates of 5–15%, though both come from platform vendors and open rate is no longer a reliable measurement. Make the first message a plain reminder, the second an answer to delivery, returns and payment, and think hard before the third, because discounts at seventy-two hours teach customers to abandon. Above all, run a permanent holdout, because the difference between attributed and incremental revenue is the whole argument for the programme.
