Most website redesigns start with a brief that was written before anyone looked at the market. Somebody decides the site looks dated, a budget is agreed, a designer is chosen, and the research, if it happens at all — is a competitor’s homepage opened in another tab. The new site launches, it looks better, and eighteen months later the same conversation starts again. I know this from personal experience.

This guide is the research I would do before spending a single penny on a redesign, in the order I would do it, and what each piece decides for your plan. It is written for a UK small business, not a SaaS growth team with a heatmap subscription. Most of what ranks for this question is USA focused and assumes you have a conversion-rate team at your beck and call. In reality you probably have an afternoon and a tight budget, so the point is to spend both on the right things.

First, decide whether you need a redesign at all

The question people actually type into ChatGPT is “does my small business website need a redesign, and how can I tell?” The answers it gets are lists of signs: the site looks old, it is slow on mobile, the bounce rate is high, it does not match the brand. All of those can be true of a site that is doing its job, and none of them tell you whether a redesign would change what the business earns.

Three questions do. If the answer to all of them is no, fix the site you have and keep the money.

  1. Is the site losing work it should be winning? Not “could it be prettier” but: are people who actually need what you sell finding a competitor instead, or arriving and leaving without making contact? You can see the second in your analytics - you are using analytics right?. You can only see the first by researching the market, which is the rest of this guide.
  2. Does your target market search for things the site does not answer? If buyers search for “website redesign price” and your site has no pricing page, that is a structural gap no amount of visual polish will close. Structure is a redesign question. Colour is not.
  3. Is the platform costing you ownership or options? If you cannot export your content, cannot move hosts (where the website actually lives), or are paying for features you no longer use, then a redesign may really be a migration. That changes the brief, the risk and the budget. Rented Ground covers this in full.

Notice that none of these is about how the site looks. A dated design is a reason to update the design. It is not, by itself, a reason to rebuild the site, and conflating the two is how a £2,000 refresh becomes a £12,000 rebuild that solves the wrong problem.

The five things to research, in order

The order matters because each step narrows the next. Demand tells you which competitors matter. Competitors tell you which page shapes win. AI citations tell you what the machines already trust. Your own data tells you what to protect. Only then do you know what the site has to do, which is the brief.

1. What your market actually searches for

Start with the language. Not the language you use to describe the business, but the words buyers type when they have the problem you solve. These are usually pretty different, and the gap between them is the single most common reason a good-looking site earns nothing from search.

When I ran this research on KOODOS itself, the positioning I had written assumed people would search for an ongoing “SEO retainer”. In the UK, that phrase gets around 40 searches a month. “Ongoing SEO services” gets 10. The phrase buyers actually use for the same thing is “AI SEO agency”, at roughly 590 a month with almost no competition.1 Same service, different words, a fifteen-fold difference in demand. I rewrote the page.

What to collect for each service you sell:

  • The head term — the two- or three-word phrase with the most searches that a buyer, not a browser, would use. “Website redesign UK”, not “web design”.
  • Its difficulty — how strong the pages already ranking are. A term with 1,600 searches and difficulty 10 is worth more to a small business than one with 3,000 searches and difficulty 60 that you will never reach.
  • The modifiers — price, cost, near me, for small business, the county you work in. These tell you which sections and which extra pages the site needs.
  • The dead ends — terms with no volume, or volume owned entirely by software tools rather than service providers. Knowing what not to target saves more than knowing what to target.

The tools for this are DataForSEO, Ahrefs, Semrush or Google’s own Keyword Planner. Any of them will do; the discipline is what matters. Choose one target term per page you intend to build. If you cannot name the term, you cannot judge whether the page worked.

2. Who ranks, and where they are thin

Now you know the terms, look at who owns them. Not the competitors you think of — the ones on page one for the phrases from step one. Those are often not the same firms. In my own research, the businesses that came up in AI recommendations barely overlapped with the ones ranking in Google for the same question.2

For each competitor, pull the keywords their site ranks for and look at three things:

  • Which pages do the work. Usually a small number of pages earn most of a site’s search traffic. When I analysed twelve UK agencies, one cost guide on a single site accounted for 136 ranking keywords on its own. That tells you the shape of page that wins the pricing question: a detailed guide with real numbers, not a thin FAQ.
  • Which page shapes win which terms. Service pages win the commercial terms (“seo audit services”, “website redesign services”). Guides win the question terms (cost, how to, checklist). Comparison pages win platform-choice terms. If your redesign plan has the wrong shape for a term, it will not rank for it regardless of quality.
  • Where nobody is doing well. Terms with real demand where the top results are generic, foreign, or out of date. These are the openings. A small firm cannot outrank a national brand on its strongest term, but it can own a specific question that the brand answers badly.

This step produces the page list. It is the difference between a redesign brief that says “services page, about page, contact page” and one that says “a Build page targeting website redesign UK with a pricing block, a migration section and a comparison against template providers, because that is what the page-one competitors lack”.

3. Who the AI assistants already cite

A growing share of buyers now ask ChatGPT, Perplexity or Google’s AI Mode instead of scanning ten blue links. Those systems cite a handful of sources per answer, and the sources are not the same as Google’s top ten. Before you redesign, you should know who they are citing for your market’s questions, because that is the bar the new site has to clear.

The method is simple enough to do by hand. Write down the ten or fifteen questions a buyer would ask before choosing a business like yours. Put each one to three or four AI assistants. Record every business named and every page cited. Count.

When I did this for KOODOS — fifteen questions, four surfaces, sixty answers — the result was zero. Not cited once, not named once.3

What was more useful was the shape of the field. No domain dominated: the most-cited source appeared six times across sixty answers. Small UK agency blogs were cited alongside Webflow, Semrush and HubSpot. The bar is not “be a famous brand”. The bar is “have a specific, well-framed page that answers the question directly”.

And one of the questions nobody owned was this one. “What research should I do before redesigning my company website?” returned a field of American SaaS guides and no UK small-business answer at all. This page exists because the research said it should.

4. What your current site already earns

This is the step that gets skipped, and it is the one that costs the most when it is. Before you change a single URL, you need to know which pages on the existing site earn traffic, links or enquiries, because a redesign that changes structure is a migration, and migrations are where rankings die.

Open Google Search Console and your analytics for the last twelve months and list:

  • Every page that received search clicks, with the queries that brought them. These pages must survive, at the same URL or with a redirect.
  • Every page with links from other websites. Those links are equity you have already earned. Losing them is losing money you already spent.
  • Every page that preceded a contact, call or purchase. These are the pages that work, whatever they look like.
  • Everything else. This is your candidate list for retiring, merging or rewriting. Most sites have more of this than they expect.

The output is a URL map: old address, new address, what happens to it. If the redesign is being done by an agency and they have not asked for this, ask why. The Build service includes the migration work as standard because leaving it out is how a redesign ends up ranking worse than the site it replaced.

5. What the site has to do — the brief, written last

Only now do you write the brief, and it should be short. Not a mood board, not a list of pages copied from a competitor: a statement of what the site must do, derived from the four steps above.

  • The page map: each page, the term it targets, the shape it takes, and what it links to.
  • The content priorities: which pages are written first, because they earn first.
  • The protected list: the URLs and links that must survive the move, and how.
  • The platform decision: chosen for what the site has to do and what you need to own, not for what the designer likes using.
  • The measure: what “worked” means: rankings for the target terms, AI citations for the question set, enquiries from the pages built to earn them.

A brief like this is a page or two. It is also the first thing that makes a quote comparable. Three agencies quoting against “redesign our website” will give you three different sites. Three quoting against this brief will give you three prices for the same site, which is the only comparison that means anything.

What you should have at the end

One document, written for a decision rather than a dashboard or an emotional perspective about site design. I call mine a Demand Map. Whatever you call yours, it should contain the search demand for each service, the competitor gaps, the AI-citation baseline, the protected URL list, the page map, and a prioritised list of what to fix and what to build. If it takes more than an hour to read, it is a data dump, not research.

If you want to see what one looks like, there is a full sample Demand Map on this site, written for a development consultancy and anonymised. Read the summary and the action list first. That is the part a redesign should start from.

How long & how much I hear you say?

Done properly, with the tools above, this is a couple of days of focused work for someone who does it regularly, or a couple of weeks of evenings for someone doing it for the first time. GroundWork is priced at £250 and completes in three or four working days including the walkthrough.4

The most expensive SEO audit on page one of Google for the equivalent term was £1,499, and it was a technical review of an existing site rather than a map of the market.

Compare either figure with the cost of a redesign that targets the wrong terms, drops the pages that were earning, and has to be done again. The research is the cheap part. The redesign is the expensive part. Do them in that order.

When not to redesign

Sometimes the research says stop. The site already ranks for the terms that matter, the pages that earn are fine, and the thing that looks dated is the photography. In that case the honest answer is a content update, better images and a technical tidy-up, at a fraction of the cost. A research-led agency should be willing to tell you that, and half the point of doing the research before choosing the agency is that you will know it yourself.

The launch is a moment. The work is the system. Research is what makes the moment worth having.

Where to go next

  • GroundWork — the research above as a service: £250, three or four working days, one Demand Map.
  • Build — website redesign and SEO scoped against the Demand Map, with the migration work included.
  • Rented Ground — what you actually own when you own a website, and the four questions to ask before choosing a platform.
  • What should a website redesign cost in the UK? — the first supporting guide in this series: what £149, £399 and £995 buy, and how to compare quotes for the same site.

The rest of the series — the SEO migration checklist, what an audit fee should buy, bespoke versus template, and how to research competitor keywords before a rebuild — is listed alongside this guide and will be linked here as each one publishes.

  1. DataForSEO Labs, United Kingdom, English, desktop, September 2026. Volumes are monthly averages; difficulty is DataForSEO keyword difficulty on a 0–100 scale. ↩
  2. KOODOS AI-visibility baseline, 17 September 2026: the SEO-audit question produced 22 different agency names across four AI surfaces; the research-led web design question produced 15 with none repeated. ↩
  3. KOODOS AI-visibility baseline, 17 September 2026: 15 questions × Perplexity, ChatGPT, Gemini and Google AI Mode, 60 completed calls, 0 citations or mentions of getkoodos.com. ↩
  4. KOODOS GroundWork pricing and timing, September 2026. ↩