When to redesign your website, and when not to

The honest answer to when to redesign your website is that a rebuild is worth buying only when the thing that is failing is something a rebuild changes. Most sites that feel wrong are failing on one of four things, content, structure, speed or brand, and a new design reaches two of them.

We should say what we sell before we argue about this. Redefine Web builds and rebuilds websites, our services page describes the web design work as rebuilds that convert, and the web design and development page publishes a one-time price ladder we are paid on. So run the tests below before you call anyone, us included. If you want the wider diagnostic rather than the decision, the website audit hub sits behind this page.

When to redesign your website starts by naming what is actually failing

The request almost always arrives as a feeling. The site looks tired, a competitor launched something sharper, somebody on the board used the word outdated. A feeling is a real signal that something is wrong and a poor guide to what. A rebuild is a specific intervention with a specific reach, and you cannot tell whether it reaches your problem until the problem has a name.

The two most useful answers on this search agree about that from opposite ends of the market. iFactory, writing about higher education sites, puts it this way: “A refresh applied to a structural problem just makes a confusing site prettier. A redesign applied to a surface problem spends two budget cycles rebuilding what was not broken.” Connective Web Design is blunter: “Most businesses need refreshes. They buy redesigns.” Two agencies that sell the work both say the buyer usually buys the wrong size of it.

So the first move is not a mood board and it is not a quote. It is a diagnosis you can run yourself this week, on your own site, with tools that cost nothing. That diagnosis is also the thing that makes an audit worth commissioning at all, which is the argument in our piece on why a website audit is important. This page is narrower than that one. It asks only which fault you have, and whether a rebuild touches it.

The four faults, and why they get confused

Almost everything a business calls a bad website is one of four faults, sometimes more than one at once.

• Content. The pages do not say the thing a buyer arrived to find out.
• Structure. The information exists, but a stranger cannot reach it, or the wrong page is the one search engines show.
• Speed. Pages load or respond slowly enough that people leave before they read anything.
• Brand. The site says true things badly, or says things that used to be true.

They get confused for two reasons. All four produce the same visible symptom, which is a line on a chart going the wrong way. And the one that is easiest to see is brand, because you look at your own homepage every day. You do not look at your own load time, your own click depth, or the questions your best landing pages quietly fail to answer.

A rebuild is bought to fix brand and sold as a fix for all four. That gap is what this page exists to close. The next four sections give each fault a test with an observable answer, meaning a test where two honest people running it separately end up with the same result.

Is it content, and how you would know

Open Search Console, go to Performance, set the range to the last three months and sort queries by impressions. Take the top twenty queries where you collect impressions and almost no clicks. Open the page Google is showing for each one. Read only what sits above the fold. Ask whether the first screen answers that query in a sentence.

The answer is observable rather than aesthetic. Either the query is answered on the first screen or it is not, and you can hand the same twenty to a colleague and get the same tally back. Count how many of the twenty fail.

If most of them fail, you have a content fault, and here is the part that gets skipped in proposals. A rebuild does not fix it. The words move to a new template. The new template renders the same sentences in a better font, on the same URL, answering the same query just as badly. You will have paid a build price for a typography change, and the chart will not move.

The confusing part is that a rebuild feels like it fixes content, because most rebuild projects include a copy pass somewhere in the middle. The copy pass is the thing that worked. You can buy it on its own, for a fraction of the price, without touching a single URL.

Is it structure, and how you would know

Two tests here, both with observable answers.

First, the click test. Hand your site to somebody who does not work for you and does not know your jargon. Give them one task, the thing you most want a visitor to do. Count the clicks and count the seconds, and do not help. Around three clicks to the page that matters and your navigation is carrying the site. Six clicks, a wrong turn, or a reach for the search box, and it is not.

Second, the duplication test. Filter Search Console to a query that matters and look at which page ranks for it. If two of your pages trade places for one query across several weeks, or if the page ranking is not the page you would send a customer to, your structure is arguing with itself. Our guide to conducting a website audit covers the crawl that finds these at scale rather than one query at a time.

Structure is the fault a rebuild genuinely does fix, on one condition. The project has to change the information architecture and the URLs, not just the skin stretched over them. Ask a prospective agency which URLs will change and what the new navigation map looks like. If the answer is that URLs stay put and the navigation is the same with new styling, you are buying a refresh, and your structural fault will survive it intact.

Is it speed, and how you would know

Speed is the one fault you can settle without anybody’s opinion, because Google published the numbers and the method for reading them.

Core Web Vitals sets three thresholds. On loading, web.dev states that “LCP should occur within 2.5 seconds of when the page first starts loading”. On responsiveness, the same page states that “pages should have a INP of 200 milliseconds or less”. On visual stability, the documented target is a CLS of 0.1 or less. Then comes the clause that ends most arguments, because it names how to read all three: “a good threshold to measure is the 75th percentile of page loads, segmented across mobile and desktop devices”.

Your site is not fast because it is fast on your laptop on office wifi. It is fast if the slowest quarter of real visits still clears those thresholds, on mobile as well as desktop. Run your five most important pages and write the three numbers down for each.

Speed is also the fault a rebuild most often makes worse. A new build tends to ship a hero video, a slider, three font weights, a consent banner and an animation library that the old site never carried. Nothing about being new makes a page fast. If you want to know what the various scoring tools are actually measuring before you trust a number from one, our website checker comparison goes through them.

Is it brand, and how you would know

Brand is the fault everybody starts with, and the only one where looking at the site is a valid test. It still has a version with an observable answer, and that version is not do I like it.

Run three checks. The truth check first. Read every service page, every price, staff name, location, phone number and claim on the site, and mark anything that is no longer true. Then the recognition check. Put your homepage beside five competitors at the same screen size and ask three people who does what. If they cannot tell your business apart from the others, the problem is positioning rather than pixels. Then the era check. Does the site work at 390 pixels wide, or does it merely shrink to fit?

Brand is genuinely fixed by design work, so this is the fault where a rebuild earns its price. Notice the shape of it though. A site full of things that are no longer true is a maintenance failure wearing a brand costume, and the fix for that is an afternoon of edits by somebody with a login. Our website maintenance work exists for exactly that gap, and it is not priced like a build because it is not one.

Reading the four answers together

Write down four answers, pass or fail. Then read them as a set rather than one at a time, because the count is the part nobody can talk you out of.

The reason to count rather than debate is that an agency can talk you out of a feeling, and we would be as capable of that as anyone. Nobody can talk you out of three failing tests or into four passing ones. That is the whole value of writing the four answers down before the first call rather than after it.

What a rebuild does not fix

This is the list that proposals leave out, so it is worth being plain about it.

• Content that does not answer the question. New templates, same sentences.
• A weak offer. A clearer page sells a bad offer more clearly.
• Traffic you never had. If nothing ranks today, a new design gives search engines no new reason to rank it.
• A hosting problem. If the server responds slowly, a new front end waits on the same server.
• A sales process. Leads landing in an inbox nobody reads arrive faster and go cold the same way.
• Speed, on its own. More often a new build spends some.

Every item on that list is real work, and a rebuild project frequently includes several of them as side effects of the schedule. That is precisely why rebuilds appear to fix them. Buy the side effect directly and you keep your URLs, your rankings and most of your budget.

When a rebuild loses rankings instead of gaining them

This is the cost that almost never appears on a quote.

The moment a rebuild changes URLs, you are running a site migration, whatever the invoice calls it. Google’s documentation on site moves with URL changes puts the expectation in one sentence: “Expect temporary fluctuation in site ranking during the move.” On duration, the same page says: “As a general rule, a small to medium-sized website can take a few weeks for most pages to move, and larger sites take longer.”

So the floor is weeks of unsettled rankings even when the migration is done properly, with every old address redirected to its closest match. Done badly, the loss stops being temporary. The failures are ordinary rather than exotic. Pages get quietly dropped because nobody could see what traffic they earned. Redirects get pointed at the homepage instead of the matching page. Internal links get rebuilt around the new navigation, so the pages that used to receive them no longer do.

Now put that beside your diagnosis. If your only failing test is content, a rebuild buys you weeks of volatility and a migration risk in exchange for solving nothing you measured. That is not a close call, it is a clear no. Afterward, our piece on knowing whether your SEO is working covers how to read the dip and tell an expected recovery from a real loss.

What targeted repair looks like when it is the right call

Targeted repair is not a smaller version of a rebuild. It is a different job, and it is the honest answer more often than an agency’s incentives make comfortable to say out loud.

• Content fault. Rewrite the twenty pages the Search Console test flagged, on their existing URLs, answering the query in the first screen. Nothing else changes.
• Structure fault that does not need a new platform. Redraw the navigation, merge the pages competing for one query, redirect the losers into the winner, repair the internal links. Same design.
• Speed fault. Compress and correctly size images, defer what is not needed to paint the first screen, cut font weights, remove plugins nobody uses, then measure again at the 75th percentile.
• Brand fault that is really a truth fault. Correct the things that are no longer true and republish.

Every one of those keeps your URLs, which means it keeps your rankings. Every one can be measured within weeks rather than quarters. And every one can be stopped halfway without leaving you holding half a website, which is not true of a rebuild. Our SEO audit checklist is the version of this work you can run without hiring anybody.

When a rebuild is genuinely the honest answer

There are cases where repair is the expensive option. They are foundation cases rather than appearance cases, and they look like this.

• The platform cannot do what the business now does. You sell what you did not used to sell, to people you did not used to sell to, and the site has no page type for it.
• Nobody can edit it. Every change is a developer ticket, so nothing gets changed, so the content fault becomes permanent.
• The structure cannot be changed in place. Templates hard-code a hierarchy the business abandoned.
• Three or four of the tests fail together. At that point you are quoting four repairs that all touch each other.
• It is not secure and nobody maintains the code.

Notice what is missing from that list. Age is not on it. A site built years ago that passes all four tests does not need replacing because of its birthday. A competitor’s launch is not on it either. Neither is being bored of it, which is a real feeling and still not a business case.

The question to ask anyone quoting you a rebuild

One question separates a scope from a sales deck. Ask which measurement will read differently ninety days after launch, and how the two of you will read it.

A good answer names a specific number and where it comes from. Mobile LCP at the 75th percentile across your top five pages. Click depth to the contact page. The share of your top twenty queries whose landing page answers the query above the fold. Form completions per thousand sessions. Every one of those was measurable before the project and will be measurable after it, by you, without access to the agency’s dashboard.

A weak answer is adjectives. Modern. Conversion focused. Future proof. On brand. None of those can be shown to be false, which is the reason they get used. Then ask the follow-up, which is what happens if that measurement does not move. You are not fishing for a guarantee, because nobody honest gives one. You are finding out whether the person has thought about being wrong. Our breakdown of web design packages covers what actually differs between tiers, so you can tell scope from decoration inside a quote.

What we would do first, including when we would say no

We would not quote a build first. We would run the four tests, because we have watched businesses buy a rebuild for a content fault and come back a year later with the same chart and less budget to fix it with.

If one test fails, our answer is repair, and that holds even where the repair is not a thing we would be selling you. If content is the only failure, that is a writing job on your existing URLs. If speed is the only failure, that is a performance pass on the site you already have. Neither is a rebuild and neither should be priced as one.

We sell the other answer as well, and we put what it costs on the page rather than behind a call. Our web design and development page lists one-time builds at $2,500 for a starter site, $4,500 for growth, $7,500 for scale and from $12,000 for enterprise, at ten to sixteen weeks to launch. Those numbers are published so you can hold them against the cost of the repair you might need instead, and decide that the repair is better value.

The test we apply to ourselves is the one above. If we cannot name the measurement that should read differently in ninety days, we should not be taking your money for a rebuild, and you should not be handing it over.

Run the four tests this week. They cost an afternoon and a Search Console login, and they turn a feeling into four answers you can act on. If most of them pass, spend the money on the one that failed. If most of them fail, you have a foundation problem and a rebuild is the cheaper road. Either way you will know which conversation you are having. If you would rather have the four answers written up for you, our free website audit returns a written report with three ranked fixes inside twenty four hours.