Conversion and measurementincreasing online sales

Increasing online sales without more traffic: where orders are lost

Product page, basket, payment by TWINT or invoice, mobile speed: where a shop loses its sales, how to measure it and how to read an A/B test.

Mehdi B.

Co-founder · SEO & digital strategy

Increasing online sales without more traffic: where orders are lost
In brief

To increase online sales without buying more traffic, you first need to find the step where visitors give up: incomplete product page, total cost discovered too late, form too long, missing payment method, slow site on mobile. Our position: measure the funnel step by step, fix the documented obstacles, then test a single change at a time with a sample size set in advance.

An online shop that receives visitors but sells little does not necessarily need a bigger advertising budget. To increase online sales with existing traffic, you first need to know at which exact moment buyers give up: on the product page, in the basket, at payment, at the choice of delivery. Our view at Smart Impact is simple. We measure these steps before touching the site, we fix the obstacles that research already documents, and only then do we test, using a method that stops you from mistaking a coincidence for a win.

This article sticks to the fundamentals, without artificial intelligence. The AI side of optimisation (personalisation, automated messages) is covered in our guide to augmented conversion. Here, we look at what must work before any tool: a clear product page, a readable total price, a familiar payment method, a page that loads quickly on a phone.

Why increasing online sales starts with the traffic you already have

A small calculation is enough to see what is at stake. Take a Valais wine shop based in Sion, a typical scenario, which receives 10’000 visits a month and converts 1% of them. It records 100 orders. If some of the visitors who abandon today went through to the end, each extra order would come without a franc more spent on advertising, since these people are already on the site.

That said, this reasoning has a limit that must be stated up front. Nobody can announce in advance by how much a conversion rate will move. Everything depends on the product, the price, the season and the quality of the traffic. What you can do, however, is spot the known obstacles, remove them one by one and honestly measure what changes.

Where sales are lost: what the Baymard Institute documents

The Baymard Institute, an independent research centre specialising in e-commerce usability, maintains an average of published studies on cart abandonment. In its latest update of October 2026, this average reaches 70.22%, calculated across 50 studies whose results range from 55% to 84%. In other words, more than two carts out of three never become orders, and that is the norm for the sector.

Some of these abandonments are unavoidable. According to the same survey, 42% of the American shoppers questioned had abandoned a cart simply because they were browsing with no intention to buy. Among the others, the reasons break down in a telling way:

  • extra costs that were too high (delivery, taxes, fees) for 40%;
  • delivery that was too slow for 20%;
  • lack of trust in entering their bank card details for 19%;
  • being required to create an account for 18%;
  • a checkout process that was too long or complicated for 17%, as many as site errors and crashes;
  • an unsatisfactory returns policy for 13%;
  • a total cost that could not be calculated in advance for 12%;
  • too few payment methods for 9%.

These figures should be read with caution, because the panel is American and does not exactly reflect a shopper in Neuchâtel or Fribourg. Their value lies elsewhere: almost every reason matches an element of the site that you can check, page by page.

The product page, the first decision page

The Baymard Institute assesses the quality of product pages on more than 155 benchmark sites. According to its summary updated in March 2026, 52% of desktop sites and 62% of mobile sites receive a rating of “mediocre or worse”. The gaps identified are very concrete. For example, 67% of sites do not show an estimate of the total cost near the buy button, and 44% do not give access to the returns policy from the product page, while 60% of the users questioned look for precisely that information there.

Photos play an equally direct role. According to the same research, 42% of users try to judge the size of a product from the images alone. For a cosmetics brand from Lausanne, another typical scenario, the question is therefore very practical: can the visitor see the bottle in a hand, know its volume and know what they will pay in total without leaving the page?

The basket and the checkout form

Next comes the checkout funnel itself. The Baymard Institute counts an average of 5.1 steps between the basket and confirmation for a new customer, a figure almost unchanged since 2012. Its tests show, however, that the number of fields to fill in matters more than the number of steps. In 2024, an average checkout form had 11.3 fields, whereas most sites need only eight.

The typical fixes are modest: a single field for the name rather than two, the second address line hidden by default, the billing address filled in automatically from the delivery address, the promo code field collapsed. That leaves the question of the customer account. Again according to the Baymard Institute, 84% of sites offer to create an account during checkout rather than at confirmation, whereas the institute recommends deferring it to the confirmation page, once the purchase is complete.

studio portrait of a smiling young woman with dark hair using an app on her mobile phone to make a purchase or communicate

Payment: TWINT, invoice and card in Switzerland

In Switzerland, the choice of payment methods is a subject in itself. According to a TWINT press release of February 2026, the app has more than 6 million active users, it processed 901 million transactions in 2025 and it is offered by 86% of online shops. A shop in French-speaking Switzerland that does not offer it therefore belongs to a small minority.

Invoice payment also holds a place of its own. The Swiss Payment Monitor of the University of St. Gallen and ZHAW, whose 14th edition is based on a survey conducted at the end of 2025 among about 1’670 people in the three main language regions, put a fictitious situation to its respondents: paying 1’000 euros to an unknown foreign merchant. In this edition, published in February 2026, 48.5% chose purchase on invoice and 31.8% the credit card, because they judged these methods the safest. The authors draw a useful conclusion for any little-known shop: when trust is lacking, security wins over convenience.

PostFinance, which draws on the same Swiss Payment Monitor, states for its part that mobile payment, mainly TWINT, accounts for 37.7% of transactions in Swiss online shops, ahead of credit cards (26.3%) and payments by invoice or transfer (18.0%). The bank also points out that invoice payment reassures the buyer but leaves the merchant with a risk of non-payment, which can be handed to a specialist provider. For a modest-sized shop, this trade-off is made with your accountant, not only with your developer.

Delivery and returns

Delivery time and cost explain a significant share of abandonments in the Baymard Institute survey. The amount is not the only issue: 12% of the shoppers concerned also cited being unable to calculate the total cost in advance, which points to the moment these fees appear in the journey. A Fribourg shop selling bike accessories, again in a typical scenario, becomes clearer if the product page states the shipping time, the postage price and the returns conditions from the outset.

Trust and customer reviews

Reviews are the best-studied trust element. The Spiegel Research Center at Northwestern University showed in its 2017 report that the probability of purchase increases markedly as soon as a product displays its first reviews, then that the effect diminishes quickly beyond the first five. The same report observes that the probability of purchase generally peaks for ratings between 4.0 and 4.7 stars, before dropping as it approaches 5.0. Our reading: there is no point filtering out mixed reviews, because a perfect rating is less reassuring than a slightly imperfect, well-argued one.

Trust also comes from simple things: a visible Swiss address, a phone number, readable terms and conditions. Scarcity, for its part, should only be used if it is true. Displaying “limited stock” on an item that is available in quantity risks costing the shop precisely the trust it is trying to win.

Mobile speed

Google defines three metrics, the Core Web Vitals, to judge the loading experience of a page. The first, LCP (the time taken to display the largest visible element), should stay under 2.5 seconds. The second, INP (the delay in reacting to a click or an input), should stay under 200 milliseconds. The third, CLS (the layout shifts during loading), should stay under 0.1. According to Google's documentation, these thresholds are read at the 75th percentile of visits, separately on mobile and on desktop.

Google also publishes several company case studies, including Vodafone and redBus, which linked an improvement in these metrics to an increase in their sales. These are results reported by the companies themselves, on their own sites. They show that speed matters, without saying what it will bring elsewhere.

Measuring the funnel before changing the site

All these avenues remain theoretical as long as you do not know where your shop loses its visitors. Google Analytics 4 provides a series of e-commerce events that follow the purchase journey in order: viewing a product (view_item), adding to the basket (add_to_cart), starting checkout (begin_checkout), choosing delivery (add_shipping_info), entering payment (add_payment_info) and purchase (purchase). Once these events are correctly installed, the funnel exploration in GA4 displays up to ten steps and shows how many users leave the journey at each of them.

The most instructive setting is to break this funnel down by device category. A bigger drop on mobile than elsewhere at the delivery step points to a form poorly suited to phones. A uniform drop at payment points rather to a missing payment method or a cost discovered too late.

For speed, Google's documentation distinguishes two sources. Field data, from real visitors, appears in PageSpeed Insights and in the Core Web Vitals report of Search Console. Lab measurements, such as Lighthouse, are useful for diagnosis, but they do not replace field data and cannot measure INP, for lack of real clicks. It is therefore the field report that tells you whether your customers are really waiting.

The figures do not tell the whole story. The questions received by phone or by email, order returns and customer service remarks often point to the same flaw as the funnel, with the customers' own words as a bonus.

What to test first on a shop in French-speaking Switzerland

Our recommendation is to start with what does not need to be tested. A bug in one browser, a needless mandatory field, the absence of TWINT, shipping costs invisible until the end of the journey: these fixes are housekeeping, not experimentation. You make them, then you observe the funnel over a comparable period. This housekeeping is where we recommend starting any e-commerce project, before any experimentation at all.

Next come the changes whose effect is uncertain, the ones that deserve a real test. Examples are a new order of information on the product page, a free-delivery threshold, a guest checkout given more prominence or an invoice option given more prominence for new customers. For each, it is worth writing, before starting, a sentence of the type “if we show the delivery price on the product page, the share of baskets reaching the payment step should increase, because buyers will no longer discover this cost at the end”. This hypothesis fixes what you measure and how you will judge the result.

Finally, you need to choose the metric before the test. The click rate on a button is not revenue: a more visible button can attract clicks without producing a single extra order. The main metric should stay close to the sale (orders, average basket, share of visitors who buy), with an eye on returns and cancellations.

Reading an A/B test without fooling yourself

The first trap has a name among English-speaking statisticians, “peeking”: looking at the results along the way and stopping the test as soon as it seems conclusive. The statistician Evan Miller illustrated it with a simulation. With a change that produces no real effect and a check after every visitor, his worst-case scenario shows a “significant” result in 26.1% of cases, more than five times the 5% threshold one believes to be applying. His recommendation comes down to one rule: fix the sample size before starting and do not stop before reaching it.

This size is often larger than expected. Evan Miller proposes an approximate formula, n = 16σ²/δ², where δ is the smallest difference you want to detect. Take the Sion shop from our scenario, which converts 1% of its visitors and wants to detect a move to 1.2%. The calculation gives about 39’600 visitors per version, or nearly 80’000 in total. At 10’000 visits a month, such a test would take about eight months, seasons included.

For many SMEs in French-speaking Switzerland, the conclusion is therefore clear: an A/B test in the strict sense is only possible on the busiest pages and for large differences. Elsewhere, you proceed by successive fixes, comparing the funnel before and after, over equivalent periods. These comparisons remain fragile, because a holiday weekend, a Google Ads campaign or a stock-out can shift the figures. They are recorded in a change log, with the date of each modification and the outside events of the period, so that the results can be reread without error.

One last reflex protects against many mistakes. A result that looks too good for a minor change deserves a second reading, or even a second test, before it is generalised to the whole site.

If you want to know at which step your shop loses its buyers, we can talk about it during a first conversation.

Sources

e-commerceconversion
Mehdi B.
About the authorMehdi B.

Co-founder of Smart Impact. Hooked on the web from the start, he launched his first project in 2006: an online music magazine still running today. With close to 20 years of SEO experience, a Swiss federal diploma in marketing and a solid geek streak, he turns clients' ideas, sometimes hazy ones, into concrete digital projects with his team.

LinkedIn
Further reading