Definition:
An abandoned cart, also called an abandoned basket, is an electronic cart to which one or more products have been added but for which no attributable purchase is recorded within the period defined for analysis.
Abandonment is an operational classification, not proof that the person has permanently rejected the purchase. A cart may later be recovered, continued on another device or used only to save and compare products.
Table of contents
When a cart is considered abandoned
Identifying an abandoned cart requires the event that starts the cart, the event representing a purchase, the identifier being used and the permitted time between them to be defined.
A cart may begin when the first product is added to the electronic cart. However, the absence of a purchase in the same session does not necessarily represent permanent abandonment: the cart may persist, be modified or be completed later.
Cart abandonment and checkout abandonment are not equivalent. The former includes carts that never reach the purchase process. The latter occurs after checkout has started and can be analysed across stages such as identification, shipping, payment details and confirmation.
A payment failure is not automatically an abandonment either. The operation may be retried or completed using another method. Similarly, a wishlist stores products for possible future consultation and does not by itself constitute an abandoned cart.
How cart abandonment is measured
When each cart is classified as completed or not completed within a consistent window, the rate can be calculated as follows:
(initiated carts without a purchase / initiated carts) × 100
It may also be expressed as 1 - purchase rate, provided that both rates use exactly the same population and rules.
The calculation needs a consistent unit of analysis. Counting carts, people, sessions or events does not produce equivalent results. One person may start several sessions, modify the same cart on different devices or generate duplicate events without representing several purchase intentions.
Measurement requires identifiers, persistence, cross-domain traffic, consent, missing events and order deduplication to be checked. In Google Analytics 4, events such as add_to_cart, begin_checkout and purchase can represent different stages. The GA4 ecommerce documentation describes these events, but their implementation does not automatically determine when a cart should be considered abandoned.
The Baymard Institute compilation consulted in 2026 places the documented average at 70.22%, calculated from 50 studies. This is an aggregated reference rather than a target for every retailer: the product, market, device, traffic source and measurement definition all affect the result.
Causes that may explain an abandoned cart
The event sequence shows that a purchase was not completed within the defined conditions, but does not identify the reason by itself. Hypotheses that may be investigated include:
- Exploration and context: The person compares products, calculates an order, saves items, becomes distracted or is not yet ready to buy.
- Offering and conditions: Price, additional costs, delivery times, returns, availability or the absence of a suitable option change the decision.
- Checkout and usability: Mandatory account creation, lengthy forms, ambiguous instructions or poor accessibility make progression difficult.
- Technical operation and payment: Errors, slow performance, duplicate events, validation failures or the absence of a compatible payment method interrupt the process.
- Trust and fulfilment: Insufficient information about security, seller identity, delivery, privacy or after-sales support may prevent the purchase.
The same drop may combine several causes. Different segments may also leave the same stage for different reasons, so an explanation needs to remain a hypothesis until evidence supports it.
Abandonment management
Investigation begins by validating the instrumentation and conversion funnel. Events, identifiers, denominators, time windows and actual orders need to be checked before attributing the drop to design or the offering.
Progression can then be compared by stage, device, browser, channel, market, product and customer type. Technical logs, support enquiries, surveys, usability tests and session recordings provide complementary evidence, but no individual source explains every motivation.
Possible improvements include displaying the total cost and conditions earlier, preserving cart state correctly, removing unnecessary fields, explaining errors, providing suitable payment methods and checking performance and accessibility. CRO can be used to prioritise and evaluate these hypotheses without assuming that any reduction in steps will improve results.
Reminders and remarketing require a valid contact route, appropriate information, an applicable basis and controls for frequency and opt-out. It should not be assumed that every cart requires a discount: discounts may reduce margin, condition future behaviour or encourage deliberate abandonment.
Evaluation needs to consider errors, margin, cancellations, returns, order quality, complaints and satisfaction alongside the purchase rate. Reducing the abandonment percentage is not an improvement if it produces incorrect purchases or worse commercial outcomes.