Skip to content
GetProfit

Shoppers Abandon Their Baskets: What to Check First

Trace the journey from basket to purchase: locate the drop-off, check checkout on a phone and choose one change whose result you can assess.

A shopper chose a product, added it to the basket and left without placing an order. For the shop owner, that is frustrating: money has already gone into attracting them, they seemed interested, but there is no sale. It is tempting to offer a discount code, redesign the basket or commission new reminder emails straight away.

I would first work out exactly where the purchase stops. The shopper may never have started checkout. Or they may have entered an address and encountered a payment error. These are different situations, and applying the same solution to both would be a guess.

First, let’s build a journey we can compare

Take a hypothetical tableware shop. Suppose 200 mobile visitors added a product to their basket during one full week. Of that same group, 120 opened the first checkout screen after adding the product, and 60 went on to place an order. In our example, begin_checkout is sent when that screen opens. We count unique users, steps in sequence and only actions within that week. These figures are fictional, not a benchmark for other shops.

In our ecommerce conversion guide, we look at the whole journey to a sale. Here, let’s narrow it to three steps so we can choose where to investigate first.

Where the journey stops
One group. Three steps in sequence.
Added to basket200
Started checkout120
80 of 200 did not reach this step — 40%
Purchased60
60 of 120 did not complete — 50%
Illustrative funnel for one full week. Unique mobile users; subsequent steps involve the same group and take place within the week. Bar lengths are relative to the initial 200. The chart does not show why people left.

Of the 200 people, 80 did not reach checkout — 40%. After starting checkout, 60 of the 120 did not complete a purchase within our window — 50%. The first drop is larger in people; the second is larger as a share of the previous step. This helps us choose an area to investigate, but does not explain why people left. Some may have returned and bought the following week.

In a real report, I would first check that these steps are being measured correctly at all. GA4 provides the events add_to_cart, begin_checkout and purchase for them. They need to be configured. Three numbers from an events report do not, on their own, form this kind of funnel: one person can add products or start checkout several times.

We need consistent units, the same group and a defined sequence. Separately, let’s reconcile the purchases sent to analytics with the shop’s orders. We should not expect the people in analytics to match the customer database exactly, though: analytics consent and tracking limitations also affect what we can observe.

Now let’s go through the purchase on a phone

I would open the site as a new shopper, without a saved sign-in or a pre-filled address. I would choose the same tableware, reach the basket and look at what I am asked to do next. To test the payment itself, we use an agreed test method so we do not accidentally place a real order.

Suppose that, in our hypothetical shop, pressing “Checkout” opens an account sign-in screen. There are two large fields, an invitation to register, and a small “Continue without registering” link below. It is easy to read that as a secondary option.

This gives us a specific hypothesis to test: some new shoppers may not notice guest checkout. Baymard’s research explains why making that route prominent matters. But it does not prove that registration caused all 60 people who started checkout in our illustrative funnel to leave without completing a purchase.

What the shopper sees
Guest checkout exists. But is it easy to see?
Suppose it looks like this now
Sign in
Create an account
Continue without registering
A hypothesis to test
Checkout as a guest
Already have an account? Sign in
Let’s make the next step clear to a new shopper.
An illustrative screen, not a real shop interface. We check whether shoppers complete the order, not just click the new button.

We now have a small change to try: make the existing guest route clear and easy to see. Before redesigning the whole basket, we can test this particular change. At the same time, we can check whether the link opens and whether shoppers can continue on the phones and browsers they use.

If people leave later, we check a different stage

Imagine that reaching checkout works well, but a noticeable drop begins after the address step. I would then look at when the full delivery cost appears and whether the required delivery option is available. If the problem occurs at payment, can we reproduce the error, and is there a matching record with the payment provider?

We need to check these ideas against the actual checkout flow. A missing intermediate analytics event does not necessarily mean someone got stuck on a screen: an express payment method may follow a different route, or the event may simply not be configured.

From the stage to the check
The chart tells us where to look
At the start of checkout
Does the button work? Is the guest option clear? Let’s check on a phone while signed out.
After entering an address
What delivery terms does the shopper see? Is the delivery option available for their address?
At payment
Can we reproduce the error? Does the provider confirm it? Are events being sent correctly?
These are areas to investigate, not a diagnosis. The actual sequence of screens and express payment routes may differ.

We do not assign a single cause based on the shape of the chart. First, we find the stage, then check what the shopper sees and whether the action works. We fix a reproducible technical error. If there is no error, we are left with a usability hypothesis that still needs testing.

Let’s choose one change and a way to assess it

Back to guest checkout. Before changing it, let’s record whom we will compare: for example, new mobile visitors who reach the basket. We care about completed purchases as well as clicks on the more prominent link. It is also useful to check whether errors and support enquiries increase.

If there is enough traffic, randomly allocating comparable visitors between two versions will give us a more reliable estimate than simply comparing weeks. If traffic is low, we can observe the result after the change, but we should not declare the change the cause of any growth: advertising, stock availability, prices or days of the week may have changed at the same time. One or two purchases cannot settle that question.

For those who still leave, a basket reminder may sometimes be appropriate. We have a separate article on abandoned basket emails. A reminder helps people return, but sending them back to the same broken form makes little sense.

The first useful result here is modest: we know where people drop out, have checked that stage on a phone, and have chosen one change with a clear way to assess it. That is enough to get started without treating every abandoned basket as a reason for a complete redesign.