The Baymard Institute puts the average documented online shopping cart abandonment rate at 70.22%, from a meta-analysis of 50 published studies on the subject.
Read that figure carefully
It is an average of other people's studies rather than a single measurement, and the individual studies it draws on range widely. Baymard is explicit that this is a meta-analysis. What makes it useful is not precision but stability: the figure has hovered around 70% for well over a decade across many different studies, which is a strong hint that it reflects something real about how people shop rather than a quirk of one dataset.
Your own rate is the one that matters, and it is sitting in your store's analytics right now. Look at it before you act on anyone's benchmark, including this one.
The practical read: if roughly seven in ten people who start a purchase do not finish, the cheapest available growth is not more visitors. It is the people already in your checkout.
Surprise costs at the final step, forced account creation, too many fields, missing payment methods. The most common causes of abandonment in published research are mundane and fixable.
Product pages
Not enough photographs, no size or specification detail, no delivery date, no returns information. People do not abandon because they changed their mind; they abandon because they could not answer a question.
Post-purchase
The order confirmation is the most-opened message you will ever send and most stores waste it. Second orders are the cheapest revenue in ecommerce.
Repeat purchase
No reason to come back, no reminder when it is time to reorder, no idea who your best customers are.
Acquisition
Last, deliberately. Sending more people through a leaking funnel multiplies the leak.
Immediately. This is the highest return in the list
Reviews
A reviews app that collects after delivery
Once you have around fifty orders to ask
Analytics
Your platform's own reports plus one analytics tool
Day one, free
Search on your own store
Better on-site search
When your catalogue passes a few hundred items
Support
A shared inbox, live chat only if someone is there
When email volume becomes unmanageable
Loyalty or subscription
A retention app
Only for genuinely repeat-purchase categories
Ads
Platform ad accounts
After the funnel above stops leaking
Everything else in a typical app store is a bet. Popups, upsell widgets, urgency timers, spin-to-win: each adds friction and load time in exchange for a promised lift that is rarely measured properly once installed.
The app that costs more than it charges
Every app you install runs code on your storefront. Ten apps mean ten scripts loading before your product page is usable, and slow pages lose sales quietly, on mobile, in a way that never shows up as an app's fault.
Before installing anything, record your product page load time. After installing, record it again. If it got meaningfully worse, the app needs to earn back both its subscription and the sales the delay costs. Most do not, and nobody ever checks, which is why small stores accumulate apps that are net negative.
Intent already proven. Remove the obstacle rather than adding a discount by reflex, or you train people to abandon deliberately.
Post-purchase
Confirmation, delivery updates, then a review request timed after the product actually arrives. High open rates, almost no cost.
Replenishment or return
Timed to your product's real consumption cycle. For consumables this is the single most profitable automation available to a small store.
On discounting the abandoned cart
A discount in the first abandonment email works, and it teaches your customers that waiting produces a voucher. If you use one, put it in the last message rather than the first, and lead with the practical objections instead: delivery cost, delivery time, returns policy, payment options. Fixing the reason people left is more durable than paying them to come back.
The 70.22% figure is a meta-analysis average across separately published studies with differing methodologies, not a single measurement. Individual studies vary considerably. It is cited here as a stable long-run indication rather than a precise benchmark.
