It is easy to see what a successful website looks like and hard to see how it works. So the visible layer gets copied — the spacing, the type, the way the hero is composed — and the copy performs nothing like the original, because the visible layer was never what made it work. The genuinely shared traits sit underneath: in what loads first, what happens when something fails, what is deliberately absent, and what is never allowed to change. What follows are the habits that show up again and again across categories that otherwise have nothing in common, and that remain rare on ordinary business sites.
← All resources
— Guide
The uncommon things the most popular websites have in common.
Look at the biggest site in any category — retail, government, reference, enterprise software — and the things they share are not the things anyone copies. Nobody imitates the habits that actually hold those sites up, because those habits are invisible in a screenshot.
The same habits, across categories with nothing else in common
| Category | What the leaders in it do | What the typical site does instead |
|---|---|---|
| Reference & news — Wikipedia, BBC News | Serve a complete, readable page before any JavaScript runs, then test the browser and upgrade it only if it passes | Render nothing at all until a bundle has downloaded, parsed and executed |
| Government — GOV.UK | Progressive enhancement is written policy, not preference: all government services must follow it | Treated as a nice-to-have, and the first thing dropped when a deadline tightens |
| E-commerce | Treat a tenth of a second as a revenue line — 0.1s of mobile speed is worth 8.4% in retail conversions | Measure speed once, in a lab tool, on the day the site launches |
| Enterprise & SaaS | Old URLs keep resolving for a decade or more; redirects are permanent infrastructure | A redesign renames everything, and every link anyone ever shared becomes a 404 |
| Every category above | The most valuable space on the page is never a rotating carousel | A carousel in the hero, because it settled an argument about whose message went first |
| Every category above | Change arrives continuously, in pieces small enough that nobody calls it a relaunch | A full redesign every three years, shipped in one night, with a fresh set of unknown bugs |
These are patterns, not laws — plenty of large sites break one of them. What is striking is how rarely any of them breaks two, and how consistently the exceptions are the sites people complain about.
The things everyone copies are the things that matter least
Ask why a well-known site works and the answers come back visual: the layout is clean, the photography is good, the typography is confident. Those things are real, and they are also the easiest part to reproduce — a competent designer can match the look of almost any site on the internet in a week. If appearance were the mechanism, the copies would work as well as the originals. They routinely do not.
What separates them is a set of decisions that never appear in a screenshot, because they are about failure and time rather than composition: what a visitor sees when something does not load, how long the page takes to become useful on a bad connection, whether a link shared five years ago still resolves. These are unglamorous, they are hard to sell, and they are almost perfectly correlated with the sites that stay popular.
Habit one: the page works before the JavaScript does
BBC News published the technique it uses for this in 2012 and gave it a name that stuck — “cutting the mustard.” Every page ships a core experience built from HTML and CSS that is complete and readable on its own. A short inline script then tests the browser for a handful of capabilities, and only browsers that pass are given the enhanced version. The consequence is that there is no such thing as a blank page: if the enhancement fails, breaks, or is simply slow to arrive, the reader is already reading.
GOV.UK takes the same position and makes it binding. Its service manual states that all government services must be built with progressive enhancement, on the reasoning that HTML is the most resilient layer and that a service which needs JavaScript to show its content has made every visitor dependent on a network request that will sometimes fail. Set that against the current baseline: the HTTP Archive measured the median page shipping roughly 570 KB of JavaScript on mobile in 2024, up 14% in a single year. The most-read sites on the internet are moving in the opposite direction to everyone else.
Habit two: the most valuable space on the page is never a carousel
The rotating hero carousel is the most reliably copied element in web design and one of the least defensible. Erik Runyon published click data from the University of Notre Dame that has never really been contradicted: the homepage slider drew a click-through rate of barely 1%, and roughly 84% of those clicks landed on the first slide. On other sites in the same dataset the top slot took 48–62% of all carousel clicks. Whatever is on slide three is, in practice, not published.
The reason carousels persist has nothing to do with visitors. They exist because they resolve an internal argument — marketing, sales and the product team each get a slide, and nobody has to be told no. That is a real organisational benefit and it is paid for by the audience. The large sites that dropped carousels did not do it for taste; they did it because they were measuring, and the measurement said the space was being wasted.
Habit three: a tenth of a second is treated as money
Deloitte and Google studied 37 European and American brand sites across more than 30 million user sessions and isolated what a 0.1-second improvement in mobile speed was worth. In retail it raised conversions 8.4% and consumer spend 9.2%. In travel, conversions rose 10.1%. On lead-generation pages, bounce rate improved 8.3%. A tenth of a second is below the threshold at which a person consciously notices anything, and it moved every one of those numbers.
The habit is not “be fast” — everyone agrees with that and it changes nothing. The habit is treating speed as a number someone owns, measured continuously on real visitors rather than once in a lab tool on a developer’s laptop. Lab scores measure a simulated device on a simulated connection; field data measures the people who actually turned up, on the phones they actually own. When the two disagree, the field data is the one attached to revenue. This is also why the sites that care about this run Monthly Maintenance-style ongoing measurement rather than a one-off audit — a score taken at launch describes a site that no longer exists.
Habit four: URLs outlive the redesigns
Tim Berners-Lee wrote “Cool URIs don’t change” in 1998, and the largest sites on the web have quietly obeyed it ever since. A Wikipedia article URL from 2004 still resolves. BBC News stories from the early 2000s are still where they were. Government guidance pages survive changes of department, of government, and of the software serving them. None of this is an accident — it is the result of treating redirects as permanent infrastructure that is maintained, rather than as a migration chore that gets closed off once the traffic dies down.
The typical business site does the opposite, and the damage is invisible because it is distributed. Every 404 is a link someone published, a search result that no longer converts, an email from two years ago that now leads nowhere. The value was in the accumulated references, and a rename discards all of it at once. If you are planning a rebuild, the redirect map is not the boring part of the project — it is the part that decides whether you keep what the old site earned. We cover the mechanics of this in when to redesign your website.
Habit five: nothing is ever relaunched
Ask when Amazon, Wikipedia or GOV.UK last redesigned and there is no clean answer, because none of them did — not in the sense of a date when the old site was switched off and a new one appeared. They changed continuously, in increments small enough that no individual change was an event. Each one could be measured on its own, and reversed on its own when the measurement came back wrong.
The big-bang relaunch is popular because it fits how work gets approved: a budget, a deadline, a reveal. It is also the riskiest way to ship anything. Hundreds of changes go live simultaneously, so when a metric falls nobody can say which change did it, and the only available response is to argue about it. Sites that evolve continuously always know what caused what. This is the single hardest habit to adopt on a normal budget, and the one that pays back longest.
What any of this is worth on a normal budget
None of these habits require a large team. Serving usable HTML before JavaScript is a choice made at the start of a build and costs nothing afterwards; retrofitting it later is expensive, which is the whole argument for deciding early. Not building a carousel is cheaper than building one. Keeping a redirect map is an afternoon during a migration and worthless if left until after. Measuring real users is a free browser API and somewhere to put the numbers.
What they do require is someone deciding they matter before the visual design is signed off, because three of the five become costly the moment the structure is fixed. That is the reasoning behind how our own packages are scoped — a Business Website is specified around what it has to survive rather than how many pages it has, and if you would rather have the structure argued out before anything is built, a Strategy Session is the cheapest point at which these decisions are still free. The full range and what each includes is on Prices.
Sources
Every figure and quotation above is from the original publication rather than a summary of it. Checked August 2026.
- BBC Responsive News — Cutting the mustard ↗
The BBC team’s own description of the technique: a core HTML experience for every browser, with an inline capability test deciding which browsers receive the enhanced version.
- GOV.UK Service Manual — Building a robust frontend using progressive enhancement ↗
The binding standard for UK government services: build with HTML first, and all services must follow progressive enhancement even where part of the service needs JavaScript.
- GDS Technology Blog — Why we use progressive enhancement to build GOV.UK ↗
The reasoning behind the policy, framed as resilience as much as inclusiveness — HTML as the layer least able to fail.
- Erik Runyon — Carousel Interaction Stats ↗
Click data from University of Notre Dame sites: a homepage carousel click-through rate of barely 1%, with about 84% of those clicks on the first slide, and top-slot shares of 48–62% elsewhere in the dataset.
- Deloitte & Google — Milliseconds Make Millions ↗
37 European and American brand sites, 30 million-plus user sessions: a 0.1-second mobile speed improvement raised retail conversions 8.4% and consumer spend 9.2%, travel conversions 10.1%, and improved lead-generation bounce rate 8.3%.
- HTTP Archive — Web Almanac 2024, JavaScript ↗
The median page shipped roughly 570 KB of JavaScript on mobile and 620 KB on desktop, with the median payload rising 14% over the year.
- W3C — Cool URIs don’t change ↗
Tim Berners-Lee’s 1998 argument that URLs are a commitment rather than an implementation detail, and that changing them destroys value that was never yours alone.
— FAQ
Frequently asked questions
Want a site built on the habits rather than the appearance?
Tell us what it has to do and who it has to survive, and we will write the scope down before anyone designs a page.