Cart abandonment: the causes you can actually fix

Baymard Institute, which maintains an ongoing aggregation of published abandonment studies, puts the average documented rate at roughly 70%. That figure gets quoted as though it described a catastrophe requiring immediate intervention, and the more useful reading is different.

A large share of abandonment was never a lost sale. People use carts as shortlists, add items to check the delivery charge to their pincode, compare a total against another site, or save something for a payday two weeks away. None of those is a checkout defect, and none of them can be recovered by improving anything, because nothing failed.

So the aggregate rate is a poor target. Driving it to zero is not achievable and pursuing it directly leads to interventions aimed at people who were never going to buy today. The productive question is narrower: among customers who actually started checking out, where do they stop, and is the reason something you control? That question has answers you can act on, and the blended rate obscures every one of them.

Instrument the sequence rather than the endpoints. The steps worth recording separately: product viewed, item added, cart viewed, checkout started, contact details entered, address completed, payment method selected, payment attempted, payment succeeded.

With that in place, the drop between any two adjacent steps becomes a specific question with a specific answer. A large fall between cart viewed and checkout started points at something visible in the cart — most often the delivery charge appearing there for the first time. A fall during address entry points at the form. A fall between payment attempted and payment succeeded is not a persuasion problem at all; it is a technical failure, and treating it as disinterest means never fixing it.

This distinction is what separates the three genuinely different phenomena that the single word 'abandonment' bundles together: browsing that was never a purchase, friction inside a checkout you built, and payments that broke. They have nothing in common and no shared remedy, and a business that only tracks one overall rate cannot tell which one it has.

The mechanism is worth understanding precisely, because it explains why this specific cost hurts more than its size would suggest. A customer who has spent several minutes choosing has settled on a price. When the total changes at the final step, the reaction is not arithmetic — it is the feeling of having been shown one number and charged another, and the response is to leave rather than to recalculate.

This is why a delivery charge revealed only after the address is entered performs so much worse than the same charge shown earlier. The amount is identical; the experience is not.

Three remedies work. Show the delivery charge, or a calculator for it, on the product page. Set a free-delivery threshold and state it prominently, which converts the charge into a reason to add another item. Or build it into the price and advertise delivery as included, which is simplest and suits businesses with predictable parcel sizes.

What does not work is the common compromise of mentioning that charges apply without saying what they are. The customer still meets an unknown number at the last step, which is the thing causing the damage.

Requiring an account before purchase asks the customer to complete an unrelated task in order to do the thing they came for. Offer guest checkout, and offer account creation immediately after the order — at that point they have already entered the details, so the account costs them one click and arrives when they can see the benefit.

On the form itself, the discipline is to ask for exactly what you need to fulfil and invoice the order and nothing else. Every additional field is a chance to stop. Date of birth, company name, how they heard about you and a second phone number are all fields somebody wanted for a report rather than for delivery.

The mechanics matter as much as the length. Use the pincode to populate city and state rather than asking. Set the correct input types so a phone number brings up a numeric keypad. Support browser and phone autofill, which requires standard field naming. Never ask for the same information twice. And validate as the customer types, with the error message beside the field rather than at the top of the page after submission — a checkout that rejects an entry without saying which one is a checkout people leave.

A meaningful part of the drop at the final step is not a decision. Bank systems have downtime, one-time passwords arrive late or not at all, additional authentication loops, and the handoff to a UPI app on a phone can fail and return the customer to a broken page.

The seller's view of this is misleading, because a failed payment often looks identical to an abandonment in a report. The customer's view is worse: many of them tried, some tried twice, and they are now unsure whether they have been charged.

Four things help. Offer more than one instrument, including UPI, so a customer whose card fails has an alternative without starting again. Preserve the order across a failed attempt, so a retry does not mean re-entering the address. Show an unambiguous message about what happened and whether money was taken, since the fear of a double charge stops people retrying. And monitor the success rate by instrument over time, because that is how you discover that one method has been failing for a fortnight — a fact invisible in an overall abandonment figure.

A reminder to somebody who was genuinely interrupted can recover the order, and this is a real mechanism worth using — with two constraints.

The first is consent and contact rules. You need a lawful basis to message somebody, which means the address or number was provided and the contact is within what they agreed to. This is a reason to capture contact details early in checkout, and not a reason to message people who never gave them.

The second is that recovery cannot fix structure. If the customer left because the delivery charge appeared unexpectedly, the reminder presents the same charge again with no new information. Recovery messaging succeeds against distraction and fails against every cause in the sections above, which is why the order of work is checkout first, messaging second. A business that adds recovery to a broken checkout has bought a way to be reminded of the problem repeatedly.

Keep it to one or two messages, and resist leading with a discount. A discount that reliably follows abandonment is a discount customers learn to trigger, and you will have taught your most price-sensitive buyers to leave the cart on purpose.