Research
- Checkouts ask for more than they need: The average checkout had 11.3 form fields in 2024 (11.8 in 2021, 12.7 in 2019), while most sites need only 8. The number of fields matters more than the number of steps (5.1 on average). 17% of users report having abandoned over checkout complexity. What it doesn’t show: Sample size for the average is not given in the article. Baymard sells this research. The 17% is self-reported. Source
- Where product pages fall short: Over 155+ US and European shops: 52% of desktop and 62% of mobile sites rate mediocre or worse. Common failures: size hidden in a dropdown instead of buttons (57%), no estimate of total cost near the buy section (67%), return policy not shown or linked (44%), no in-scale picture (37%). What it doesn’t show: Baymard’s own benchmark, of larger commerce sites, and it sells this research; individual percentages have no stated sample. Numbers updated March 2026. Source
- Most mobile shops left the phone keyboard working against people: Baymard’s 2015 benchmark of 48 leading mobile shops: 79% didn’t turn autocorrect off where its dictionary is weak (street names, e-mail addresses); 54% didn’t show the optimised keyboard for e-mail, phone or numbers; 27% had auto-capitalisation wrong; 25% didn’t bring up the keyboard consistently. What it doesn’t show: Old (2015), and only shops. Sites have probably moved on since; the behaviours themselves still exist. No figure for what it costs in sales. Source
- Home page versus product page: the section outline: An infographic lists the sections of each page. Home: value proposition, primary call to action above the fold, shop by category, featured products, how it works, personalised discovery, promotions with a reason to act, FAQ, final call to action. Product: strong hero, call to action impossible to miss, social proof near the decision, benefits not just features, reviews, evidence for claims, comparison, cross-sell, FAQ after the evidence. What it doesn’t show: A practitioner’s outline, shown without data. The slots of the shop home and product page types follow it, so their order is advice, not a measurement. Read from the frames of the reel; the original infographic was not found. Source
- Core Web Vitals thresholds: Good means: Largest Contentful Paint at most 2.5 s, Interaction to Next Paint at most 200 ms, Cumulative Layout Shift at most 0.1, at the 75th percentile of page loads, for phone and desktop apart. Specimens use the first and third as lab readings. What it doesn’t show: Topogeny measures one load on this machine with no throttling, so it is a rough reading and not field data, and it cannot read INP without interaction. Source
- The five levels of atomic design: Atoms are the foundational building blocks; molecules are simple groups of elements working as a unit; organisms are complex components made of those; templates place components into a layout; pages are templates with real content. Topogeny’s atoms, molecules, sections and page types follow this. What it doesn’t show: A way of organising work, not evidence that any layout works better. Source
- How people hold phones, and a correction to the thumb charts: 1,333 people observed in public, 780 of them touching the screen: 49% one-handed, 36% cradled, 15% two-handed. In a 2017 comment the author says his reach drawings were based on informal sampling and calls them “a lie”; he reports that people touch the centre of the screen more, more quickly and more accurately than corners and edges. What it doesn’t show: Observed in 2012–13 on much smaller phones. The later correction is reported in a comment, not given as a new published study. Source
- A handwritten page beat the formula: Neville Medhora says his most successful product page ever was written by hand on an iPad and “broke every rule”: it made a lot of money and many e-mail sign-ups. His conclusion: section formulas are guides to look at, not laws. What it doesn’t show: One person’s account of his own page, with no numbers shown and an audience that already knew him. It shows that a formula is not required, not that handwriting works. Source
- Accordions earn their keep on phones, with care: On phones, accordions save space and work as a small table of contents. They hurt when opening moves content so people think a new page opened and press Back, when long answers make the next item hard to reach, and when the close control is not understood. What it doesn’t show: Observations and guidance from 2015; no numbers given. Source
- A rotating box hid the biggest offer on the page: In a usability test of the Siemens Appliances home page, someone asked to look for deals on washing machines could not find a £100 offer, though it was the largest item on the page: the box auto-rotated every 5 seconds and the offer was in view about 20% of the time. The article also names problems for people with motor difficulties, low literacy or another first language. What it doesn’t show: One observed case from 2013, in an article making a general argument. Auto-rotation is the target here, not every carousel. Source
- Unformatted text gets read in an F: People scan a wall of text in an F shape: two sweeps across the top, then down the left. It happens when text has little formatting, the reader wants to be efficient and is not committed to every word. Put the main points first, use headings that start with the informative words, bold key phrases, use lists. What it doesn’t show: About scanning text on a page; good formatting reduces it, and other patterns exist. Studies of about 47 and 45 people; article from 2017. Source
- Hidden navigation is used less and finds less: Remote tests with 179 people on 6 live sites. On phones, navigation was used in 57% of tasks when hidden against 86% when partly visible (combo); discoverability was more than 20% lower when hidden, on phones and desktop; on desktop, tasks took at least 39% longer with hidden navigation, and on phones 15% longer than combo. What it doesn’t show: Six sites, 2016. Visible-only navigation was not tested on phones. Later generations may know the menu icon better, but the cost of hiding comes from the hiding. Source
- Most attention stays near the top: In an eyetracking study of 120 people (more than 130,000 fixations, on a 1920 × 1080 screen), 57% of page-viewing time fell above the fold and 74% in the first two screenfuls. Topogeny lets earlier sections count for more in a page’s score because of this. What it doesn’t show: Desktop screens, a mix of tasks and pages, published 2018. It says where attention goes, not what converts. The exact position weighting is judgement. Source
- Sticky headers help if they stay small: Sticky headers make navigation, search and utilities easier to reach and use. They should be small, opaque and steady; tap targets at least 1 cm × 1 cm; and ask first whether people need them often enough to pay for the space. Headers that reappear on scrolling up are a good compromise. What it doesn’t show: Guidance based on research across projects, 2021; no effect size given. Source
- Touch targets of at least 1 × 1 cm: NN/g recommends targets of at least 1 × 1 cm rendered, and gives no pixel value because densities vary. It cites MIT Touch Lab: fingertips are about 1.6–2 cm wide, and thumbs make contact over about 2.5 cm. Small targets are slower to hit and cause slips. About 2 mm between targets is the often-recommended minimum, though spacing doesn’t fix small targets. What it doesn’t show: Guidance from research, without figures on how much it improves task success. Physical size can’t be set exactly from CSS. Source
- Few click a carousel, and mostly the first slide: Six months of traffic in 2013 on five University of Notre Dame sites. On ND.edu, with 3.76 million home page visits, 1.07% clicked a feature and 89.1% of those clicks went to position 1. The other manual carousels had 1.27%–2.99% with 62%–85% on position 1. The one auto-forwarding site, a news site, had 9.41% with 54.6% on position 1. What it doesn’t show: One university, 2013, clicks only and no control. The auto-forwarding result is a different kind of site, so it does not show that rotating is better. Clicks also miss whoever just read the slide. Source
- No sideways scrolling at 320 px: WCAG 2.2 success criterion 1.4.10 (AA): content can be shown without loss and without scrolling in two dimensions at a width equal to 320 CSS px, other than parts that need two dimensions for their meaning, such as maps and data tables. Topogeny tests every wireframe against this and measures it on specimens. What it doesn’t show: Passing it says nothing about whether the layout is good at that width, only that it fits. Source
- Pointer targets of at least 24 × 24 px: WCAG 2.2 success criterion 2.5.8 (AA): targets are at least 24 × 24 CSS px, with exceptions for spacing, an equivalent control, inline links, browser-controlled targets and where the size is essential. Measured on every specimen. What it doesn’t show: A minimum, not a recommended size. The measurement here counts small targets without working through every exception. Source
- Nearly every home page has detectable accessibility failures: WebAIM’s 2026 scan of 1,000,000 home pages found detected WCAG failures on 95.9%, 56.1 errors per page on average. Most common: low-contrast text 83.9%, missing alternative text 53.1%, missing form labels 51%, empty links 46.3%, empty buttons 30.6%, missing document language 13.5%. What it doesn’t show: Home pages only, and automated detection only. WebAIM itself says that absence of detected errors does not indicate that a page is accessible or conformant. Source
- Restaurant visitors go for the menu: In a UK poll: 70% of adults view restaurant websites and 41% check menus online to estimate the cost. 50% name general menu problems as their biggest complaint: 31% are most irritated by there being no menu at all, and 17% by one that is hard to reach, such as a downloadable PDF. What it doesn’t show: Commissioned by Wix, which sells website builders; sample size not given in the release; UK, 2021; what people say, not what they do. Source
Real pages