When visitors do not understand your website, you pay for traffic you never use.

Most companies have a budget for bringing people to the website. Search campaigns, content for SEO, posts on social media, often an agency to run it all. Fewer companies give the same attention to what happens after the click.

That imbalance is supported by data. According to the Nielsen Norman Group, the average page visit lasts a little under a minute and people often leave during the first ten to twenty seconds. In 2011 Jakob Nielsen concluded from this that if you want a few minutes of attention, you have to communicate your value proposition clearly within ten seconds. This is not his own measurement but his reading of 2010 Microsoft Research data covering more than two billion recorded dwell times across 205,873 pages. Treat ten seconds not as a law, then, but as a rough estimate of the patience available to you.

A website is not a standalone graphics product that gets redrawn once every three years. It is the last and most expensive metre of the acquisition funnel. If it cannot communicate value quickly, marketing is pushing traffic into a leaky system.

A paid click that fails to convert is friction with a price tag

At JUR0 we assess waste with the Cost of Friction model: time × frequency × number of people × cost. On a website, time is the number of seconds a visitor spends working out what you sell, frequency is the number of sessions per month, the number of people is your visitors, and cost is your cost per click. That product is the real price of an unclear website, a budget line nobody ever invoices you for. Unlike the internal friction covered in Good design is not decoration, here you are not paying for the time of your own people but for the time of other people, which you have bought.

That this is an acquisition metric rather than a technical one is shown by a study commissioned by Google and carried out by the agency 55 and Deloitte across 37 brand websites, with data from more than 30 million sessions at the end of 2019. An improvement of 0.1 seconds lifted the conversion rate in retail by 8.4%. On lead generation sites, the number of people who clicked all the way through to the form submission page rose by 21.6%. According to the methodology note, no website was redesigned during the measurement period. A decimal place changed and revenue moved.

The visitor's first five questions

Someone who lands on your page is not wondering whether the typography is tasteful. They are asking five questions, mostly without realising it: Where am I? What do you offer? Is this for me? Why should I trust you? What do I do next?

If any of them goes unanswered on the first screen, there will be no second screen. According to a measurement cited by the Nielsen Norman Group, during an average visit a user reads at most 28% of the words on a page, and realistically around 20%. Every additional hundred words buys only 4.4 more seconds on the page. Adding paragraphs therefore does not solve the problem, it only stretches it.

Message hierarchy: most websites start with themselves, not the customer

Message hierarchy is the order in which a page delivers information, what someone should read first and what last. Most corporate websites have it backwards. They open with company history, values, the team and a photo of the building. The customer learns who you are before learning whether you can help.

Kara Pernice of the Nielsen Norman Group points out that people scan pages in an F shaped pattern: a longer horizontal movement at the top, a shorter one below it, then vertically down the left side. That is bad news for business, because people skip important content simply because it sits on the right or further down, not because it is uninteresting. Important information therefore belongs in the opening paragraphs, and headings have to carry meaning in their first few words. A heading that says “About us” carries no information. A heading that names the customer's problem does.

That this is not a matter of taste was shown by a measurement study by Morkes and Nielsen in 1997 with 51 experienced users. Against the control version, the scannable text scored 47% better on usability, the concise version 58%, objective language instead of promotional style 27%, and all three changes combined 124%. The study is old and should be read with that in mind, but the order of the causes has not changed.

A value proposition that survives the competitor test

A value proposition is one sentence saying what you do, who you do it for and why it is better than the alternative. The Nielsen Norman Group recommends treating the homepage as an elevator pitch that says quickly and clearly what the organisation does, typically through a descriptive tagline in the opening section.

A test I recommend running before every redesign: put your headline sentence next to the sentences of three competitors and remove the logos. If even your own salesperson cannot tell them apart, you do not have a value proposition, you have a category description. “Comprehensive bespoke solutions” is not value, it is the name of an industry. A company that can say exactly where it is better can also charge more for it, which is the subject of Why some companies can charge twice as much.

Information architecture as a scent that has to get stronger

Information architecture is the way content is divided, named and connected: what sits in the menu, what it is called and where it leads. Nielsen explains it through information foraging theory: on the web a person behaves like a forager and decides on the basis of the information scent, the signals telling them whether the path leads to their goal. The key point is that the scent has to keep getting stronger, otherwise people give up. Links such as “Click here” or “Learn more” break the scent, because they say nothing about what is on the other side. Nielsen also warns that the better search engines get at surfacing quality sites, the less time users spend on any one of them. The alternative is one back button away, so bought traffic has less patience than someone who already knows you.

Trust signals and which of them actually work

Trust signals are the elements that reduce perceived risk. Aurora Harley of the Nielsen Norman Group summarises the four credibility factors Jakob Nielsen defined in 1999: quality of execution, upfront disclosure (prices, terms and contact details with no hidden fees), comprehensive and current content, and connection to the rest of the web.

More interesting is what research participants said about testimonials. They did not believe claims made on a company website, because a company only publishes positive reviews there, and they relied on external, impartial sources. That changes the priorities: verifiable figures, project parameters and links beyond your own domain do more than another block of quotes. Trust also comes from behaving consistently across channels, which is the subject of A brand is not a logo.

According to research by Gitte Lindgaard and colleagues from 2006, cited by the Nielsen Norman Group, the judgement about a page's aesthetics is made within 50 milliseconds and rarely changes afterwards. That is not an argument for decoration but for order: the Nielsen Norman Group recommends no more than four accent colours, two typefaces and one visually dominant call to action.

Every field in a form has a price

Friction is any obstacle between an intention and carrying it out. In a form it can be counted field by field. The Baymard Institute, which is built on more than 200,000 hours of research, reports that the average checkout for a new user has 5.1 steps and, in 2024, 11.3 form fields, while eight fields would be enough for most sites. Roughly 29% of the fields are therefore unnecessary.

Baymard also aggregates 50 third party studies from 2006 to 2025 into an average cart abandonment rate of 70.22%. An ordering process that was too long or too complicated was given as a reason by 17% of US online shoppers. From its own testing it estimates that the average large e-commerce site can gain a 35.26% increase in conversion through better checkout design.

The same logic applies to a corporate lead form: every extra field is a question someone may not want to answer while they do not yet trust you. If you ask for a phone number, a company registration number and headcount from someone who only wants a price list, you are buying data at the cost of enquiries. It is the cheapest fix on the list, because deleting three fields does not require a new website.

Mobile is the primary scenario, not a derivative

In 2017 Google commissioned a measurement of 900,000 mobile ad landing pages from 126 countries. The average full load took 22 seconds, while 53% of visitors abandon a page that takes longer than three seconds to load. A model built on the bounce and conversion data showed that as load time goes from one second to seven, the probability of leaving rises by 113%. When the number of elements on a page grows from 400 to 6,000, the probability of conversion falls by 95%.

That last figure matters most to designers. It is not only about servers, it is about the number of things we put on the page. Every extra banner, animation, consent pop-up and chat widget competes for the same budget of attention and loading. Mobile therefore cannot be designed as a shrunken desktop. It is where you have to decide what gets cut.

Conversion rate is a system, not a button colour

Inside companies, conversion rate optimisation has often narrowed to details: button colour, microcopy wording, section order. Tests like that can deliver small improvements, but they will not fix a page that answers the wrong question. Conversion is the result of a chain: ad relevance, the match between ad and page, clarity of value, trust, ease of action and speed. If one link in the chain is weak, optimising the others only moves the loss elsewhere. In my experience it therefore pays to find the weakest link first and only then start testing.

How to diagnose a website in a week

An in-house team can do this without an agency and without new tools.

1. Five questions. Open the homepage and your most visited subpages on a phone and answer the five questions from the opening.
2. Competitor test. Put your headline sentence next to the sentences of competitors with the logos removed and let people outside marketing try to tell them apart.
3. The path from the ad. Click through your most expensive keywords as a customer and watch whether the landing page answers what the ad promised.
4. Speed and elements. Measure mobile load time and list everything that loads. Remove whatever has no owner and no purpose.
5. Forms. Count the fields and write next to each one the name of the person who works with that piece of data. Delete the fields with no owner.
6. Outsiders. Let people who do not know the company walk through the site out loud and note where they hesitate.
7. Priority. Rank the findings by the Cost of Friction formula, not by whatever annoys you most.

Sometimes the positioning needs fixing before Figma is opened

This is the part design studios happily skip, because it is hard to invoice. If a company cannot say exactly who it sells to and how it is different, no amount of visual work will save it. Design can make information clearer, but it cannot invent it. A new website built on unclear positioning is only a better looking expression of uncertainty.

I will say it plainly, even against our own short term interests. Often the cheapest correct answer is to rewrite the headline sentence, remove three fields from the form, speed up loading and leave the existing website alone. A template with good copy beats an expensive bespoke website with a vague offer. Custom development makes sense when a company hits the limits of a template or has a specific process that cannot be assembled from ready made blocks. If someone offers you a new website before asking about your cost per click and your conversion rate, they are selling an output, not a solution.

Key takeaways

• Work out your Cost of Friction on acquisition: cost per click times the number of visitors who leave without acting, times twelve months. That is the annual price of being unclear.
• Test your headline sentence next to the competition with the logos removed. If it cannot be told apart, the problem is in the positioning and Figma will not solve it.
• Delete every form field for which you cannot name the person who works with that piece of data. The Baymard Institute measured 11.3 fields in the average checkout, while eight are enough for most sites.
• Before you order a redesign, run a week long diagnostic and rank the findings by cost.