Rebuild or refresh: a decision framework

Your website is five years old. It still loads, the phone still rings, and nobody inside the business is happy with it. Someone, possibly an agency, has told you it needs to be rebuilt.

Maybe. Most of the sites we are asked to rebuild do not need it, and a few of the sites nobody is worried about need it urgently. Rebuild or refresh is not a judgement about how a site looks. It is a judgement about four things underneath the surface: what it runs on, how its addresses are arranged, who is able to change it, and whether the business it describes is still the business you run.

Get that judgement wrong in the expensive direction and you pay again for a foundation you already had. Get it wrong in the cheap direction and you spend the budget on paint over a structural fault, then spend it again eighteen months later.

What a refresh is, and what a rebuild is

The two words get used loosely, usually by whoever is writing the quote.

A refresh keeps the structure. Same platform, same page architecture, same URLs, same content model. What changes is the surface and the language: typography, layout, spacing, imagery, page copy, sometimes a handful of new pages. The sitemap you start with is roughly the sitemap you end with, and the pages that already rank keep ranking, because they never moved.

A rebuild replaces the foundation. A new platform or a clean install of the current one, a new page architecture, a new URL structure, content migrated deliberately rather than carried across out of habit. The surface changes too, but that is a consequence rather than the reason.

The gap in price between them follows the structure, not the design. A refresh where the content already exists starts from S$4,000. A rebuild starts higher, because migration, redirect mapping and content that has to be written from nothing are all real work before a single layout decision gets made.

When a refresh is the right call

Often. Probably more often than an agency will tell you. Here is the test, in five parts.

The platform is current and someone has been patching it. Not perfect, not the newest thing on the market. Supported, receiving security updates, and actually receiving them.

The URLs are readable and stable. Your pages live at addresses that describe them, and the ones bringing you traffic have not moved in years.

Your team can edit a page. Somebody who is not a developer can change a phone number, swap a photograph, or add a project, today, without raising a ticket.

The business is the same shape. Same services, same buyers, same proof, same name. The site describes the company you currently run.

What is actually wrong is age and drift. The layout reads as a few years old. The copy describes your services in the order you happened to add them. The projects stop somewhere around the year the site was built.

If all five hold, you have a design and content project, not an engineering one. Spend the money on photography, on rewriting the pages that decide something, and on the handful of facts a buyer needs and currently cannot find. You will get most of the benefit of a rebuild for a fraction of the cost and none of the risk.

One caution, because it is the reason most rebuilds get started in the first place. Looking dated is a real problem and a weak argument on its own. Across 54 Singapore B2B sites we reviewed in August 2026, one batch of 21 was assessed for visual age, and all 21 read as dated. It is close to universal in this market, which means it is rarely the thing separating you from the firm winning the work.

The four conditions that make a rebuild the cheaper answer

The platform is end of life

In that same review, 12 of 54 sites were running an outdated or end-of-life content management system or plugin. Two were running Joomla 1.5, which reached end of life in 2012. Fourteen years without a security patch, on a site that is still the first thing a buyer sees.

The reason this forces a rebuild is not the security exposure, although that is reason enough on its own. It is that there is no upgrade path. You cannot patch forward from a version that far behind. Moving to a supported release means rebuilding templates, re-entering content, and replacing extensions that no longer exist for the current version. That is a rebuild whether you call it one or not, so the only decision left is whether you get a better site out of the work you are already paying for.

Worth separating out: a plugin two releases behind is a maintenance problem, not a rebuild trigger. Maintenance from S$200 per month covers that, and it is the difference between a site that ages gracefully and a site that has to be replaced.

The URL structure is broken or arbitrary

Older systems produce addresses that are really database queries: an index file, a component name, a view, and a numeric id on the end. Nothing in that address tells a buyer or a search engine what the page is. Two pages about completely different things look identical apart from the number. Nobody can link to them with confidence and nobody can read them out over the phone.

Moving URLs is survivable. Google’s own guidance on 301 redirects is clear that a permanent redirect passes ranking signals to the new address, and a careful one-to-one map is standard work on any competent rebuild. The problem is when there is nothing worth mapping, or when the map cannot be built because the old addresses do not identify anything stable.

The worse version is a structure that is already failing. In the review, 7 of 54 sites had a contact page or a navigation link returning a server error or a 404. One fabricator’s site had SERVICES, ABOUT US, PROJECTS and CONTACT all pointing back at the homepage, while the obvious addresses for its contact and about pages returned 404. There is no refresh for that. The navigation is the architecture, and the architecture is gone.

Nothing can be changed without a developer

This is the condition that hides, because it does not present as a technical problem. It presents as staleness.

Of the 23 sites in the review whose footer we examined, 19 carried an out-of-date copyright year. One crane rental business still describes its fleet “as at” a date in 2009, on the single page doing the selling. Nobody in either case decided to publish stale information. They simply could not change it without a quote, a lead time and somebody else’s calendar, so it never got changed.

That cost compounds quietly. Every fact that is slightly wrong stays slightly wrong. Every new project goes unpublished. The site drifts further from the business until the only honest fix is to replace it.

A rebuild here is not buying a new design. It is buying back the ability to edit, which over 5 years is worth more than any layout decision on the page.

The business has changed shape

New services, a new sector, a different buyer, a certification you did not hold when the site was built, a merger, a name change. The site is accurate about a company that no longer exists in that form.

The clearest symptom is a missing project record. In the review, 19 of 54 sites published no portfolio, project list or case studies of any kind. For a firm whose entire case rests on comparable work, that is not a content gap you can patch with a new page. The site was never built to hold that kind of content, so adding it properly means changing the architecture, which is a rebuild.

Hong Hock Global is a version of this. Water infrastructure work is awarded on prequalification rather than persuasion, so the site was built as a prequalification document rather than a brochure, with BizSafe Star, BCA-certified personnel and current contract scope stated plainly and early. That is a structural decision about what the site is for. You cannot arrive at it by restyling a brochure.

How do you tell which one you have?

An afternoon, no agency required.

Try to change one sentence on your homepage yourself. Time it. If you cannot, or if the answer is “ask the developer”, condition three has already failed.

Find out what the site runs on, and which version. Whoever hosts it can tell you in a minute. Then check whether that version is still supported.

Take the five URLs that bring you the most traffic. Confirm each one still exists, then read the address out loud. If it is unreadable, it will stay unreadable unless you rebuild.

Check that both the www and non-www forms of your domain answer, and that the certificate is valid. In the same review, 6 of 54 answered on only one form of their own domain, and 5 threw a full-page browser security warning that no buyer will click past.

Write down the ten facts a buyer needs before they will call you. Certifications, capacity, scope, named contact, recent projects. Count how many appear on the site and how many clicks each takes to reach.

Fail only the last one and you have a refresh with a content project attached, which is the most common honest answer. Fail the first, second or third and you are rebuilding, whatever the invoice calls it.

What moves the number

For a refresh, the variables are page count, how much copy is being rewritten rather than re-laid out, and whether photography has to be commissioned. Content that already exists in a usable form is the single biggest saving available to anybody.

For a rebuild, the migration is the cost. Redirect mapping across a few hundred old addresses. Content that exists nowhere and has to be gathered out of people’s heads, old proposals and PDFs. Integrations that were never documented. The design is rarely the expensive part, which surprises people.

Our pricing bands are published rather than quoted on request, and we have written separately about what a B2B website actually costs in Singapore. Hosting from S$200 per year and maintenance from S$200 per month are what keep you out of this decision entirely in 3 years’ time, and every build carries 30 days of post-launch support.

The bias worth naming

We make more money on a rebuild. That is worth saying out loud, because it is the reason this advice is usually skewed.

A rebuild of a site that only needed a refresh spends your budget re-laying ground you already owned, and it carries risks a refresh does not: lost addresses, lost ranking history, a longer programme, and a period where the new site is worse than the old one because the content did not all make it across.

If you are unsure, run the afternoon test before you talk to anybody about design. Four of the five questions have nothing to do with how the site looks, which is the point.