When an old website starts costing your company money

Most companies replace their website only when it starts to look old. That is the weakest possible criterion, because the age of a design has nothing to do with economics. For a business with a sales team and a marketing budget, the website is infrastructure: it qualifies demand, builds trust before the first conversation and creates data for the next decision. Infrastructure is not replaced on the strength of its appearance, but on what it costs to run.

The most expensive website is therefore not necessarily the one that cost a lot to build. It is usually the inexpensive one that quietly reduces conversion for years and takes hours every month from the people who operate it. This article is not about how a site should communicate value, which is a separate discipline. It is about the economics of one decision: the point at which a corporate website redesign stops being a question of taste and becomes a line item with a return. By rebuild we mean reconstruction on a new foundation, not redrawing the existing one.

A strong website shortens the path to a decision, and that is a cost line

In a high-value B2B purchase, a prospect does not decide after a single ad. They verify competence, references, the team, the process and delivery risk. If the website does not answer those questions quickly, the sales team starts every conversation with a credibility deficit. That is not a visual problem but a question of sales cycle length: every additional week ties up a salesperson's capacity as well as the money already spent on marketing.

The window is narrower than it seems. Drawing on Microsoft Research data, the Nielsen Norman Group reports that users often leave a page within ten to twenty seconds and that the average visit lasts slightly less than a minute. The Stanford Web Credibility Project, based on three years of research with more than 4,500 people, reports that people judge a site quickly and on the basis of its visual presentation alone. Both studies are older, and the first rests on data from 2010, so treat them as a principle rather than a measurement of your own market.

How the first few seconds of comprehension are assembled is covered in When customers do not understand your website. This article takes the opposite end of the equation: what it costs you to keep postponing the decision.

When patching stops making economic sense

The common warning signs are poor mobile performance, a CMS that limits marketing, difficult localisation, fragmented analytics, forms disconnected from CRM, an inconsistent visual language and content that has grown without an information architecture, meaning without a deliberate arrangement of pages and the links between them.

Each of those problems can be patched, and in most companies that is exactly what happens. Past a certain point, however, the business stops paying for functionality and starts paying to preserve an old decision. The cost moves into media spend, support, development and employee time, where nobody records it as a website cost. That is why such a site looks inexpensive right up to the day somebody adds it up.

With speed, the problem can be named with a number. Google defines three Core Web Vitals metrics and a threshold for “good”: LCP within 2.5 seconds, INP within 200 milliseconds and CLS no higher than 0.1. LCP is the time it takes to render the largest visible element on the page, INP measures responsiveness to interaction, and CLS captures how much the content jumps under the reader's finger. Google Search Central documentation states that good Core Web Vitals, together with other aspects of page experience, align with what its core ranking systems try to reward.

The methodology is the key part. Google evaluates the 75th percentile of page loads and treats mobile separately from desktop. An average taken from analytics is therefore not enough: you can have a flattering average and still miss the threshold for the slowest quarter of visits, the ones on a weaker signal and an older phone.

Fewer than half of all sites meet the threshold on mobile, so speed still makes a difference and is not merely hygiene. The Web Almanac 2025, using HTTP Archive and Chrome UX Report data from July 2025, reports that 48 per cent of sites pass all three Core Web Vitals on mobile and 56 per cent on desktop, with LCP as the bottleneck: on mobile it is passed by 62 per cent of sites.

Cost of Friction: putting a number on an old decision

In front of the person who holds the budget, the decision is only defensible as a number. We use the Cost of Friction framework, the price of friction: time × frequency × number of people × cost. The aim is not accuracy to the euro but an order of magnitude. Small friction multiplied by twelve months is often a larger item than the entire project, and nobody sees it because it has no invoice of its own. The exercise has four steps:

• Time. How many minutes an operation takes today and how many it would take on a healthy system. The difference is friction.
• Frequency. How many times a month it is actually performed, not how many times it should be.
• Number of people. Friction for one person is an annoyance. Friction for five is a process.
• Cost. The internal hourly rate of those people, not the average salary in the company.

What follows is a worked example, illustrative arithmetic rather than a measurement taken at a client. A marketer publishes a landing page: three hours in the old CMS, twenty minutes on prepared templates, a difference of 2.5 hours, twice a month, one person, so 60 hours a year. An assistant retypes contacts from email into the CRM, four minutes per contact, sixty times a month, which is 48 hours a year. Two salespeople spend five minutes a day looking for the current price list, roughly 40 hours. That is almost 150 hours in total, and we have not yet touched conversion.

Substitute your own numbers. If the result is twenty hours a year, you will not justify a rebuild on those grounds and probably do not need one. If it runs into hundreds of hours, you have an argument that will survive a meeting with finance.

The other half of the equation is conversion, and that calls for caution. A study commissioned by Google and produced by the agency 55 and Deloitte rests on more than 30 million sessions across 37 brand websites from the end of 2019. It found a relationship between a 0.1 second improvement in the measured mobile speed metrics and an increase in retail conversions of 8.4 per cent, in customer spend of 9.2 per cent and, in lead generation, of 21.6 per cent in progression to the form submission page. This is a correlation, not a controlled experiment. Do not derive from it how much you will earn, derive the direction and the order of magnitude.

A rebuild is a chance to clean up the whole digital system

A serious rebuild does not start with colours. It starts with an audit of demand, traffic, CRM, data, content, services and the sales process. The result may be fewer pages, but more precise ones; fewer forms, but better qualified ones; and less manual work, because website data moves on into the rest of the stack automatically.

This is where most of the return is created and where most projects lose it. If the new site is built as a copy of the old structure with nothing but a new typeface, the company pays for design and buys no saving: the friction from the calculation above remains, simply in nicer packaging. We set out the same logic away from the website in Good design is not decoration.

Four isolated projects cost more than one architecture

A well-funded company should not buy design, SEO, analytics and automation as four separate projects. The site needs a shared architecture: a design system, a fast front end, conversion measurement, sound indexing, multilingual readiness and an integration layer for CRM or internal tools. Without one, four suppliers produce four local optima that begin to obstruct each other, and you then pay for the integration a second time.

A foundation of that kind is not the ordinary standard. According to Eurostat, in the 2025 reference year 79.01 per cent of EU enterprises with ten or more employees had a website, but only 29.51 per cent had content in at least two languages and 28.51 per cent used CRM software. Multilingual content and CRM integration are therefore a minority discipline, which makes them a relatively cheap opportunity to stand apart.

If you go multilingual, do it properly from the start. Google Search Central documentation recommends a separate URL for each language version instead of switching through cookies or browser settings, together with hreflang annotations, the markup that declares which version belongs to which language. It also warns against automatic redirection between language versions and against adapting content on the basis of IP address analysis, which it describes as unreliable.

The other underestimated point is that the mobile version is not a scaled-down copy but the original. Google states in its own documentation that only content shown on mobile is used for indexing, and recommends the same content and the same structured data on mobile as on desktop. If half the text does not appear on mobile, Google knows half as much about the company.

A rebuild is neither free nor free of risk

The calculation has to include the cost of the change itself, and migration is the part most often underestimated. If URLs change, Google Search Central states that you may see fluctuations in rankings while Google recrawls and reindexes the site, and that for small to medium sites this can take a few weeks or more, and longer still for larger ones.

In practice that means mapping old addresses to new ones, permanent server-side 301 redirects, rewriting internal links to the new URLs, a new sitemap in Search Console and updated hreflang annotations. A supplier whose proposal does not contain this is not offering you a rebuild, but a risk.

When the cheaper option is the right option

Not every problem that presents itself as a reason for a new website actually is one. This is the part you hear least often from a supplier, because there is no money in it.

• If the homepage does not explain what you do and who it is for, the problem is in the content, not the technology. Copywriting and rebuilding one section cost a fraction of a rebuild.
• If one template is slow and the cause is a large uncompressed image and unnecessary fonts, fix those. New technology will not solve it better, only more expensively.
• If marketing cannot publish, it is sometimes enough to add a set of content blocks and train the team rather than replace the CMS.
• If the company is smaller and the offer is stable, an off-the-shelf platform or a good template is a legitimate choice. Custom development only makes sense once the standard blocks something you genuinely do.

The order is always the same: first understand the business and the process, remove what is unnecessary, simplify the workflow, design the experience, and only then select the technology. Anyone who starts with technology buys a new problem with a higher monthly fee. We developed the related argument in Your company does not need more software.

Redesign or rebuild?

A cosmetic redesign makes sense when the technical foundation is healthy and the problem sits in presentation or content. A rebuild is more appropriate when the offer, the information architecture, the technology, the measurement or the integrations are changing. In that situation, preserving everything simply because it already exists tends to be the more expensive choice. To keep this from being a matter of instinct, work through five questions:

1. Is the offer, the segmentation or the pricing changing? Then the information architecture changes with it and cosmetics will not help.
2. Do you miss the Core Web Vitals thresholds at the 75th percentile on mobile, with the cause being the template or the technology rather than a single image?
3. Do you need an integration layer the current solution does not have: CRM, multilingual content, product data?
4. How many hours a year did the Cost of Friction calculation produce, and how does that compare with the quotation?
5. Would today's system stand in the way of your two-year plan?

If you answered yes to one or two of them, act selectively rather than across the board. If it is three or more, patching is the more expensive route. And if you cannot answer at all, that is the most important outcome of the exercise: you have no measurement. Adding analytics and measuring both speed and friction is cheaper than rebuilding a website blind.

A good website is not justified by the number of animations or pages, but by the fact that it improves the quality of enquiries and can be developed further without turning every request into a technical project.

Key takeaways

• The age of a design is not a reason to rebuild. The reason is quantified friction and a measurable loss in conversion.
• Calculate the Cost of Friction for the three most frequent operations around the website. An order of magnitude is enough.
• Measure speed at the 75th percentile and keep mobile separate from desktop, rather than working from an average.
• Budget for the migration: URL mapping, 301 redirects, the sitemap and hreflang, plus fluctuations in the first weeks.
• If copywriting, image optimisation or an off-the-shelf platform solves the problem, do that. A new foundation only makes sense once the current one blocks what the business genuinely needs to do.