A landing page is judged in fractions of a second. Google Research finds that the aesthetic judgment of a page forms within the first 50 ms, and a second run showed 17 ms was enough (Google Research). What follows – load time, form length, trust signals – decides whether that first impression turns into a purchase. This article collects the levers that can be measured and the target values that are documented for them.
Why E-Commerce Landing Pages Underperform
Online retail in Germany is no longer a side channel: 86% of people aged 16 to 74 have bought online at least once, 70% within the last three months (Federal Statistical Office of Germany). Gross merchandise revenue reached EUR 83.1 billion in 2025 (bevh). Anyone running a landing page in this market is not competing for attention as such, but for the few seconds a visitor grants the page.
The first hurdle is technical. Before a visitor sees anything, the server has to respond: web.dev names 0.8 s as a good value for Time to First Byte. Then there is the order of evaluation – Google uses the mobile version of a page, crawled with the smartphone agent, for indexing and ranking (Google). The mobile view is therefore not the secondary version but the decisive one.
| Factor | Current State | Optimized Target |
|---|---|---|
| Largest Contentful Paint | above the threshold | below 2.5 s (web.dev) |
| Time to First Byte | above the threshold | below 0.8 s (web.dev) |
| Interaction to Next Paint | above the threshold | below 200 ms (web.dev) |
| Cumulative Layout Shift | above the threshold | below 0.1 (web.dev) |
| Form elements in the flow | 23.48 on average | 12–14 elements (Baymard Institute) |
Above the Fold: The Critical First Seconds
Google Research showed participants pages for a fraction of a second and then asked for their judgment. Visual complexity and prototypicality took effect within the first 50 ms, and in a second run 17 ms was enough (Google Research). That does not mean a page sells in that time – it means the first second must not be spent on assembling the layout.
How well that works is measurable. The Web Almanac by HTTP Archive finds a good Largest Contentful Paint for 74% of desktop pages and 62% on phones; 18% on desktop and 25% on mobile need improvement, 7% compared with 13% perform poorly (Web Almanac). On phones, roughly a third of all pages therefore miss the target value.
One message, not five
The visible area carries exactly one message. Anything that needs explaining belongs below it, not beside it.
A call to action without searching
The next step is visible without scrolling and recognizable as a button. A second call further down repeats it.
Show trust early
Reviews, delivery details and payment methods belong in the visible area, not in the footer.
Load Time as Conversion Killer: Every Millisecond Counts
The relationship between load time and conversion is documented in independent analyses: the longer a page takes to load, the lower the conversion rate – and the drop is steepest in the first few seconds. Which thresholds Google applies here is covered in the article on Core Web Vitals.
Deloitte arrives at the same picture: an improvement of 0.1 s raised the conversion rate in retail by 8.4% and the average order value by 9.2% (Deloitte). Amazon measured back in 2006 that 100 ms of added load time cost 1% in sales (Conductor). What counts is not the average but the 75th percentile of page views (web.dev) – the slow visits decide, not the fast ones.
- Measure Time to First Byte: web.dev names 0.8 s as a good value.
- Identify the LCP element, usually the hero image, and check its weight: the Web Almanac measures 911 KB of image weight on home pages (HTTP Archive).
- Use field data instead of lab data and read it at the 75th percentile (web.dev).
- Evaluate phone and desktop separately, never as an average.
- Measure again after every change instead of estimating the effect.
Google names three thresholds for a good user experience: Largest Contentful Paint below 2.5 s, Interaction to Next Paint below 200 ms and Cumulative Layout Shift below 0.1 (web.dev). Interaction to Next Paint has replaced the former First Input Delay – optimizing against the old metric means measuring in the wrong place.
Less Is More: Page Complexity and Conversion
Complexity costs conversion, and it can be counted. The Baymard Institute finds an average of 23.48 form elements in a checkout, while a short flow gets by with 12–14 elements (Baymard Institute). The same applies to a landing page on a smaller scale: every field, every choice and every box is a decision the visitor has to make.
Baymard names 8 form fields as a benchmark for a complete checkout against an average of 11.3 fields (Baymard Institute) and considers a reduction of the displayed form elements by 20–60% achievable (Baymard Institute). Arithmetically this is the largest untapped share: the conversion potential of better checkout design averages 35.26% (Baymard Institute).
Review mandatory fields
Every field needs a reason rooted in the business process. Whatever is only needed after the purchase is only asked for then.
Pre-fill choices
Shipping method, country and salutation can be pre-filled sensibly. A correctable default is faster than an empty list.
Show errors immediately
Validation directly at the field avoids the jump back after submitting – the most common abandonment point in a form.
Offer alternative routes
Guest checkout and a second payment method catch the visitors who do not want to take the intended path.
Page Weight: Keeping Images and Scripts Under Control
Images are the largest single item in page weight. The Web Almanac by HTTP Archive measures 1,054 KB of images and 613 KB of JavaScript for the median page (HTTP Archive). Both travel on the same connection as the content the visitor is waiting for.
Broken down by page type the picture is even clearer: home pages carry a median of 911 KB in image weight, secondary pages 354 KB (HTTP Archive). A landing page behaves like a home page – it wants to show rather than explain – and therefore regularly ends up at the upper end.
That the work on these values pays off is documented: Vodafone improved Largest Contentful Paint by 31% and recorded 8% more sales (web.dev). The lever is almost always in the same place – the largest image in the visible area.
Trust Signals: Reviews and Verification Badges
Trust is not a soft factor but a measurable quantity. The Spiegel Research Center studied with an online retailer what displayed reviews achieve: the purchase likelihood of a product with five reviews was 270% higher than that of a product without any review (Spiegel Research Center).
The effect is not evenly distributed. For lower-priced items the conversion rate rose by 190%, for higher-priced ones by 380% (Spiegel Research Center). The more expensive and the more explanation a product needs, the more weight other buyers carry.
- Reviews with count and date, not just a row of stars.
- Delivery time and shipping cost before the click on the call to action.
- Payment methods as a row of icons, not as body text in the imprint.
- Return conditions in one sentence, linked to the full version.
- A contact route with stated availability instead of an anonymous form.
A badge identifying the verified buyer behind a review improved the odds of purchase by a further 15% (Spiegel Research Center). What matters is traceability: a review that can be placed in context is more readily believed.
Mobile First: Closing the Conversion Gap
In Germany, 50.91% of measured page views came from mobile devices in the month recorded (Statcounter Global Stats). Google uses the mobile version of a page for indexing and ranking (Google) – a landing page built on desktop and adapted for the phone afterwards is therefore built in the wrong order.
This extends to operation. web.dev names around 48 device independent pixels as the recommended minimum size for a touch target on a site with a properly set mobile viewport (web.dev). Links placed closely below one another in body text regularly fall short of that value.
- Touch targets of around 48 device independent pixels (web.dev).
- Form fields with a matching input type so the right keyboard appears.
- No horizontal scrolling, not even for tables and price overviews.
- The call to action stays reachable without scrolling.
- Largest Contentful Paint below 2.5 s on mobile networks too (web.dev).
- Font size large enough that nobody has to zoom.
- Sticky bars narrow enough not to cover the content.
- Verify every change on a real device, not only in the simulator.
Data-Driven Optimization: Building a Testing Framework
A test needs its decision rule before it starts. CXL names 95% as the usual confidence level (CXL). Setting that threshold only after looking at the interim results is no longer measuring – it is picking the result that fits. What a clean test setup looks like in a shop is shown in the article on A/B testing in the online shop.
Then there is statistical power: 80% is the standard for A/B testing tools (CXL). Together the two determine how many visitors a test needs. Without that calculation upfront, a test is either stopped too early or left running forever.
How large the lever of a single headline can be is shown by an example from Harvard Business Review: a test on how advertisement headlines were displayed at a search engine operator increased revenue by 12%, which amounted to more than USD 100 million a year in the United States alone (Harvard Business Review).
Before building a variant, look at the field count: 8 form fields are enough for a complete checkout, while the average is 11.3 (Baymard Institute). Removing fields is the test with the least build effort.
From Landing Page to Purchase: The Conversion Funnel
The path does not end on the landing page. The Baymard Institute aggregates studies on cart abandonment and arrives at an average of 70.22% across 50 surveys (Baymard Institute). What is won on the landing page can be lost again in the checkout. Where those abandonments arise in detail is covered in the article on checkout optimization.
A large part of that is operation, not missing purchase intent. Baymard puts the average gain from better checkout design at 35.26% (Baymard Institute) – the share reachable by tidying up the flow, without a single additional visitor.
That usability is not a marginal topic is shown by the European survey: 63% of e-shoppers reported no problem at all when buying online in 2025 (Eurostat). For 12% the website was difficult to use or worked unsatisfactorily (Eurostat) – a share that can be addressed with engineering, not with advertising.
The figures in this article come from the Web Almanac by HTTP Archive, the checkout research of the Baymard Institute, the thresholds and case reports on web.dev, the review study by the Spiegel Research Center as well as surveys by Eurostat, the Federal Statistical Office of Germany and bevh. Every figure is stored in the article file with a deep link and a verbatim quotation.
Mobile Landing Page Optimization: Beyond Responsive Design
For June 2025 the Web Almanac measures a good overall Core Web Vitals score for 48% of mobile pages and 56% on desktop (HTTP Archive). The gap has narrowed, but it remains – and it affects exactly the channel most visitors arrive through.
The causes are known. For Time to First Byte, 44% of mobile pages reach a good value, 40% need improvement and 17% perform poorly (HTTP Archive). On top of that comes the median image weight of 911 KB on home pages (HTTP Archive), which weighs differently over a mobile connection than in the office.
That mobile work pays off is documented: after moving to a Progressive Web App, Flipkart recorded 70% more conversions and Lancome 17% (web.dev). Together with touch targets of around 48 device independent pixels (web.dev) this separates a page that works on a phone from one that is merely displayed there.
Image weight first
The largest image in the visible area determines Largest Contentful Paint. 911 KB on a home page is the usual starting value (HTTP Archive).
Response time before looks
As long as the server takes more than 0.8 s (web.dev), no frontend optimization helps. The route runs through cache and hosting.
Stay operable
Touch targets of around 48 device independent pixels (web.dev) and visible states are a prerequisite, not an add-on.
Landing Page Testing: Systematically Reaching the Optimal Page
The effect of technical work can be estimated without running your own test. The Chromium Blog reports that pages meeting all Core Web Vitals thresholds show a 24% lower probability of being abandoned during load (Chromium Blog).
A documented case from retail: after working on the Core Web Vitals, revenue per visitor rose by 53.37% and the conversion rate by 33.13% (web.dev). Such values are not a promise; they show the order of magnitude technical work can move in.
Small steps pay in as well: for pages 0.1 s faster, Deloitte measures 9.1% more progressions from the product page to the cart (Google/Deloitte via web.dev). A test therefore does not have to rebuild the whole page to show a measurable effect.
The order matters: measure first, then formulate a hypothesis, then change exactly one thing. Swapping two elements at once leaves you knowing that something changed, but not what caused it.
- Collect the baseline over at least one full weekly cycle.
- Write down the hypothesis: what changes, what effect is expected, and why.
- Set a confidence level of 95% and a power of 80% before the start (CXL).
- Calculate the runtime from baseline and expected change instead of estimating it.
- Touch nothing during the runtime – no interim readings, no adjustments.
- Document the result even when the variant loses.
FAQ: Landing Page Optimization for E-Commerce
The target values are a Largest Contentful Paint below 2.5 s, an Interaction to Next Paint below 200 ms and a Cumulative Layout Shift below 0.1 (web.dev).
As few as the business process strictly requires. The Baymard Institute names 8 form fields as a benchmark for a complete checkout against an average of 11.3 fields and considers a reduction of the displayed form elements by 20–60% achievable (Baymard Institute).
Displayed reviews carry the most weight: the purchase likelihood of a product with five reviews was 270% higher than that of a product without any review, and a verified-buyer badge added a further 15% (Spiegel Research Center).
Google uses the mobile version for indexing and ranking (Google), and in Germany 50.91% of measured page views came from mobile devices (Statcounter Global Stats). Operation matters too: web.dev names around 48 device independent pixels as the recommended minimum size for a touch target.
As long as the sample size calculated upfront requires. A confidence level of 95% and a power of 80% are customary (CXL); together with the baseline they determine the number of visitors needed. Stopping at the first pleasing interim reading turns the measurement into a guess.
The documented effects do not depend on shop size. For pages 0.1 s faster, Deloitte measures 8.4% more conversions and a 9.2% higher order value (Deloitte); pages with good Core Web Vitals show a 24% lower probability of abandonment during load (Chromium Blog). Both work regardless of the assortment.