Jul 23,2026 
About 70% of online carts are abandoned before the sale, and the figure most often cited, 70.22%, is an average across 50 separate studies. That headline is where most diagnosis goes wrong. A large part of it is unavoidable, the ordinary browsing and price-comparing that was never going to end in a purchase. The rest is friction you built and can remove. You cannot fix a cause you have not isolated, so working out what abandons your own checkout starts with separating the share you can change from the share you cannot.
Cart Versus Checkout Abandonment
Cart abandonment and checkout abandonment are two different behaviors that need different fixes. Cart abandonment happens when a shopper adds an item and then never starts checkout. Checkout abandonment happens when someone begins checkout, sometimes enters a shipping address or a card, and leaves before the confirmation screen. They sit at different points in the funnel and rarely share a cause.
Read the split this way. Cart abandonment usually signals hesitation, and checkout abandonment usually signals friction. A shopper who leaves before checkout is likely comparing prices or saving an item for later, not ready to buy today. A shopper who quits partway through checkout has usually hit something that annoyed or blocked them. Most stores route both into a single abandoned-cart email and wonder why it underperforms. Knowing which one you have is the first real step, because the two point at opposite fixes.
Unfixable Cart Abandonment Share
A large share of cart abandonment is unfixable, the browsing and price research no design change will recover. Around 43% of US shoppers say they have left a cart because they were only browsing or not ready to buy. That is window shopping and price research, the kind of visit that was never going to convert. Treat those people as a leak to plug and you will redesign a flow that was already working.
The number that deserves your attention is the addressable share underneath the headline, the shoppers who meant to buy and stopped anyway. Measured with the browsers stripped out, Baymard’s ranked reasons for that group are led by cost and process friction:
- Unexpected extra costs such as shipping, tax, and fees: 39%
- Delivery that felt too slow: 21%
- Distrust of entering card details: 19%
- A site that forced account creation: 19%
- A checkout that ran too long or felt too complicated: 18%
Each of those leaves a trace you can find in your own data, which is what makes them worth chasing while you leave the browsers alone.
Checkout Abandonment Diagnostic Stack
A checkout abandonment diagnostic stack moves in a short sequence from where shoppers drop to why they dropped. Knowing the common causes is not the same as knowing yours. Run it in order, because each step narrows what the next one has to explain.
Segmenting the Funnel by Stage and Device
Start by building a real checkout funnel, one tracked event for every stage from the product page down to the confirmation screen. Then break the drop-off down by stage and by segment, keeping desktop and mobile apart. The stage with the steepest fall is your first suspect, and a drop that concentrates on the shipping or payment step points toward cost or trust. A blended rate that looks fine can also hide a device gap, so separate desktop from mobile early. Mobile abandons far more than desktop, near 80% against roughly 66%, so a store whose overall number looks healthy can still be losing most of its phone traffic.
Form Analytics for Field-Level Friction
Once you know the stage, form analytics tells you which field inside it makes people stop. Field-level tracking shows where users hesitate, retype, or quit filling the form, and a drop-off that clusters on one field, a phone number, an address line, a card entry, is a usability signal rather than chance. That is the difference between guessing your form is too long and knowing everyone abandons on the same line. The fix follows the finding. A field that everyone struggles with gets removed, reordered, or made easier to complete.
Session Replay for the Silent Why
Analytics tells you where, and replay tells you why. When a drop is real but the reason is hard to place, session replay lets you watch the moment it happened. Replay surfaces the things a funnel count cannot see, the rage clicks on a button that will not respond, the dead clicks on an image people expect to zoom, the long pause before someone gives up. A cluster of rage clicks on the place-order button means something is broken. A dead click on a product photo means shoppers wanted a closer look you never built. You are reading behavior rather than raw exit counts.
A Short Exit-Intent Survey
Numbers and recordings still leave one gap, the reason that lives only in the shopper’s head. An exit-intent survey asks the departing shopper directly, triggered when the cursor moves to leave or the page sits untouched for a moment. Keep it to one or two questions. A single prompt, what stopped you from finishing today, with a few options and one open box, turns a silent exit into a stated cause. It will not explain everyone, since the most annoyed shoppers rarely stop to answer, but it catches reasons your instruments never could.
Predictive Checkout Testing With Evelance
Evelance finds checkout hesitation before real shoppers arrive, which covers the two things the diagnostic stack assumes and you may not have, a checkout that already exists and live traffic to read. Give Evelance the checkout flow, describe the shopper, for instance a first-time mobile buyer placing a $60 order who expected free shipping, and predictive personas return a scored read in about the time a coffee break takes.
The scores name the exact hesitation you are chasing. Objection Level shows what makes a shopper resist, and Risk Evaluation shows if they trust the page enough to enter a card. A feature called Deep Behavioral Attribution goes further and explains why a persona reacted the way it did, the surprise-fee flinch or the pause at the coupon box a funnel count can never see. Run the same read on two versions of a checkout and you can compare where each one loses people before you commit any traffic.
Evelance predicts where personas will hesitate. It does not measure your real drop-off, which stays the job of live analytics and session replay, so the two work best in sequence, a predictive read to catch friction early, then real behavioral data to confirm it on real shoppers. On the question every team asks about AI feedback, the platform reports 89.78% thematic agreement between its persona predictions and real-user feedback in its own case study, enough to trust the direction while you verify the destination on live traffic.
Common Cart Abandonment Causes
The common cart abandonment causes are surprise costs, forced account creation, an overlong checkout, an empty coupon field, and the mobile penalty. With the method in place, each one stops being trivia and becomes something you can confirm, because it leaves a recognizable fingerprint in the data you have gathered.
Surprise Costs at the Final Step
The most common fixable cause is a cost that appears too late. Shoppers judge a price in sequence, so a shipping charge or a fee that lands at the final step comes across as unfair after everything before it looked settled. The trace is a drop that concentrates on the step where the total updates. If your steepest fall sits right where shipping is added, you are almost certainly losing people to a surprise rather than to the product. Showing the full cost early, or a free-shipping threshold that makes the number visible from the start, moves that drop.
The Forced Account Gate
A checkout that demands an account before it will take an order turns a first-time buyer into someone doing paperwork for a store they have bought nothing from yet. The trace is a drop right at the account step, heaviest among new visitors. One merchant who required registration described a flow that bounced shoppers to a login screen and then a signup before returning them to the cart. He lost the majority of them along the way. Setting accounts to optional and offering guest checkout is one of the most reliable single fixes in the field.
A Checkout With Too Many Fields
Length is measured in fields, not pages. Baymard’s testing puts an efficient checkout at 7 to 8 form fields, while the average store shows close to 14.9, nearly double what the task needs. The trace here is diffuse rather than sharp, a steady bleed across the whole flow instead of one cliff, because no single field is the culprit and the sheer count is the burden. Counting your own fields and cutting the ones you do not need is unglamorous work that pays back, since removing even one field has been enough to raise completion.
The Coupon Code Field
An empty promo code box is a subtler cause that instruments miss unless you look for it. The moment a shopper sees the box, it implies someone else is paying less, so they open a new tab to hunt for a code and never come back. Standard funnel counts show only that they left the payment step. The tell shows up in replay, a shopper focusing the coupon field, typing nothing, then going cold. Collapsing the field behind a small have-a-code link keeps it available without advertising a discount to everyone who does not have one.
The Mobile Penalty
Mobile deserves a line, because the same checkout that works on a desktop can fail on a phone. Typing a 16-digit card number on a touchscreen has a high error rate, and every mistyped digit adds time and another chance to abandon. The trace is the device gap you found earlier, the 13 to 16 point spread between phone and desktop abandonment. Autofill and address lookup help here, and one-tap wallets that collapse the form to a few taps help most of all, more than any wording change you could make.