From visitor to customer: Conversions and checkout success

Decide what success means on this page.

A website conversion is a useful action you have chosen to measure. It might be a completed purchase, a confirmed appointment or an enquiry received. Those are different outcomes. A visitor clicking an email link or opening a booking calendar has shown intent, but has not necessarily sent a message or booked a session.

Start by naming the real outcome and the steps before it. For a store, that might be product view, add to cart, checkout start, payment attempt, confirmed payment and confirmed order. This gives the design and engineering work a shared target. It also prevents a rise in button clicks from being presented as a rise in sales.

Make the offer and the next step clear.

Before asking a visitor to act, answer the questions that make the action sensible: what is being offered, who it suits, what it costs and what happens next. A strong call to action cannot rescue an unclear service or a product page missing the details someone needs to choose.

Use one clear primary action for the page’s main task. A service page might lead to a consultation; a product page should make choosing the product and adding it to the cart straightforward. Secondary information can still be available without competing for the same attention.

Our practical review starts with the first screen and follows the journey on a phone. Can the visitor understand the offer? Can they find the total cost or the conditions that affect it? Do button labels describe what will happen? Does the next screen fulfil the promise of the previous one?

Find the avoidable friction.

Baymard’s checkout research identifies issues including unexpected costs, account-creation requirements, complicated checkouts, trust concerns and errors. It also distinguishes people who are browsing without being ready to buy. Its reported research populations and benchmarks are not measurements of your South African store, so use them to guide investigation rather than predict a conversion lift.

Our recommended inspection is concrete: show delivery and other charges before the final commitment, offer guest checkout where the business model permits it, explain required fields and preserve useful information after an error. Ask for the details needed to complete the order, rather than collecting extra information simply because the form allows it.

Further reading: Baymard Institute — Cart abandonment and checkout research (2026) ↗

Design for phones, mistakes and interrupted journeys.

A checkout should work with the keyboard open, long names, a slow connection and a customer who changes their mind about an address. Check readable text, usable controls, appropriate input types and error messages placed where someone can act on them. Test the actual payment-provider handoff as well as your own pages.

A payment failure needs a specific next step where one is known: check details, try a different method or contact the provider. Avoid promising that a payment failed merely because the browser timed out. The interface should represent what the system knows, including when a result is still being confirmed.

When someone returns from a payment provider, do not strand them on an empty screen. Explain whether the order is confirmed, still processing or needs attention. Provide a reference and a sensible route to help without exposing sensitive payment information.

A success screen must reflect a real payment state.

For integrations using asynchronous payment events, the browser’s return page is only part of the flow. Stripe’s documentation describes verifying incoming events and handling event delivery reliably, including duplicate events. The exact event model differs by provider; follow the provider you actually use.

Our engineering principle is to connect the order state to a verified provider result. Process repeated notifications safely, keep a stable order reference and make it possible to reconcile a payment when the browser closes or the connection drops. Showing ‘paid’ because a URL contains a success parameter creates a gap between the interface and the money.

Test cancellation, a declined attempt, a slow response, a duplicate notification and a successful payment with no browser return. Keep the order confirmation and customer communication consistent with the verified state. A dependable checkout needs both an understandable interface and a reliable record of what happened.

Further reading: Stripe — Receive, verify and process payment events ↗

Separate abandonment, completion and payment success.

Choose a reporting period, a consistent unit and a clear denominator. Cart abandonment describes carts that did not become completed purchases. Checkout completion describes checkout starts that reached a completed order. Payment success describes successful payment attempts out of all eligible attempts. Retries, cancelled orders and multiple sessions can change the figures, so document how you count them.

For a simple illustration, suppose 100 distinct checkouts start, 60 become completed orders and those customers make 80 payment attempts, of which 60 succeed. Under those definitions checkout completion is 60%, while payment-attempt success is 75%. Neither number establishes the cart-abandonment rate because the number of carts created is missing.

Review the stage that is losing people before choosing the change. A weak product-to-cart journey needs a different investigation from repeated payment declines. Search Console can help you understand Google search traffic, but it does not measure these onsite stages. Checkout measurement requires separate instrumentation and order/payment evidence.

Improve the journey one bottleneck at a time.

Choose a specific problem, record the current evidence and make a focused change. Check that reporting still uses the same definitions afterwards. When traffic allows, a properly designed controlled experiment can help distinguish the effect of the change; with small volumes, describe observations cautiously and avoid claiming causation from a handful of orders.

A useful checkout review ends with prioritised work: the customer questions to answer, the friction to remove, the failure states to test and the signals to inspect. Bringing more visitors to a broken journey can increase frustration. Fixing that journey gives relevant traffic a better chance to become a successful customer interaction.