What a slow website actually costs
The 7% figure comes from a 2008 survey of 160 executives. Here is what has actually been measured, and where your site sits against sixteen million others.
Web
Brief, platform, landing pages, forms, speed, measurement and legal duties: what decides whether a visit becomes a conversation, and what merely decides whether the site looks good.
A website is one of the few purchases judged almost entirely on appearance, and almost never on what it produces. The brief says « modern », the quotes come back at wildly different prices, and nobody can explain the gap.
These articles are about the parts that can be specified, verified and measured. They are deliberately unromantic.
The common thread: a site is a machine with an input and an output. Everything here is about making the output visible, so the next decision is an arbitration rather than a matter of taste. Our full approach is on the B2B web design page.
32 articles in this category. See all resources
The 7% figure comes from a 2008 survey of 160 executives. Here is what has actually been measured, and where your site sits against sixteen million others.
The brief that specifies features before anyone has looked at the problem is the one that produces a site nobody uses. Two governments publish the alternative.
Most usability advice is taste. Six requirements added in 2024 are not: they have numbers, they apply to your forms, and almost no B2B site meets them.
One controlled study nearly doubled first-attempt form success. Most of the advice built on top of it has no sample size at all.
Attention is front-loaded, not fold-bounded. The measured figures are more useful than the folklore, and one of them is twenty years old.
Google says it does not read your folder structure to understand your site. It reads your links. That single sentence rewrites most architecture advice.
Two governments publish theirs, including the parts nobody copies: the rule for admitting a new component, and the status that means stop using this.
Choosing a structure is easy. Knowing whether it worked is the hard part, and the published experimentation data says most teams are reading noise.
You do not own your domain. ICANN says so in those words. What follows from that governs transfers, expiry, disputes and the day you change it.
Ad blockers and consent refusal cut the visitor count before you divide. What is actually in that denominator, and what it does to the ratio.
No federal cookie law and no legal notices page. One privacy policy is genuinely mandatory, with no size threshold, and it applies to a two-person firm.
US law does not ask for consent before your form. It asks for notice at the moment of collection, and forbids collecting at all without it.
Eight questions Google publishes, the warning signs it lists, and why a ranking guarantee is a legal exposure and not merely a bad promise.
Most maintenance contracts cover updates and backups. The failures that actually take a B2B site down are certificates, redirects and the person who left.
Server location is not a definitive ranking signal, in Google's own words. Time to first byte is not a Core Web Vital. Two hosting claims, both qualified.
Google publishes the retention period, the chain limit and which moves its change of address tool covers. Most migration checklists get the last one wrong.
Google publishes exactly three minimum technical requirements. It also says most sites should not read its crawl budget guide, and gives the page count.
The 100ms claim traces to a 2006 blog post containing no number at all. The real experiments were run by search engines and found far smaller effects.
Thousands of accessibility cases a year, under a regime that sets no technical standard for private business. Europe has the standard and almost no suits.
The SEO tools say 58%. The two studies with causal designs say 40% and 5%. Google says stable, and publishes no figures at all.
Every event you add costs cardinality, review time and a reason to argue. The useful document is the one that says no, and records why.
Four mechanisms silently remove or alter data before you see it. Three publish exact thresholds. The one that hides the most publishes no number at all.
The 21x figure is not from Harvard. It comes from a 2007 study co-published by a company selling software that dials leads faster.
The figure exists, the sample is disclosed, and the arithmetic is fine. It compares companies, not strategies, and the source itself says correlate.
The 120% figure is real and traceable. It rests on 10 submissions before and 26 after, a year apart. The largest dataset points the other way.
The famous button tests publish no per-variant sample size. The accessibility thresholds are exact, normative, and checkable without a single visitor.
It is now the inverse of engagement rate, so it carries no information that metric does not. And on a landing page it was always the wrong instrument.
The rule is not that paid traffic needs a dedicated page. It is that a page you cannot maintain is worse than a site page you already do.
No single platform is best for B2B, only the right one for your goal. A result-first look at Webflow, WordPress, Astro and Framer, with sourced data.
Redesign your B2B website when the numbers say so, not when it looks old. Seven measurable signs, the cost of waiting, and how to redesign without losing SEO.
The sections a high-converting B2B landing page needs, annotated one by one, with US benchmarks for conversion rate, form fields and lead response.
A custom build runs $10,000 to $50,000, a lead-generation site past $100,000. What puts a project at the top of the range, and the cost of ownership.
Would you rather talk it through? A conversation often beats a long article. Let's discuss your situation.