In a study of 892 domain migrations published by Search Engine Journal in January 2025, it took an average of 523 days for the new domain's estimated organic traffic to match the old one, and 17% had not recovered after 1,000 days. A same-domain rebuild is gentler than a domain move, but the lesson holds: the bigger the change, the bigger the bill. Most sites asking "rebuild or redesign?" need something smaller.
This guide gives you a four-hour check, one hour each in Search Console, GA4, PageSpeed Insights and your CMS, and a table that turns what you find into one of four answers: maintain, refresh, redesign or rebuild. The numbers come from Google's own documentation, the HTTP Archive, WebAIM, WordPress.org, Databox, Orbit Media and Nielsen Norman Group.
Four options, and what each one changes
The words get used loosely, so fix the definitions before you open a dashboard. The useful line is not how different the site looks afterwards. It is what changes underneath: URLs, templates, platform.
| Option | What changes | What stays | URL risk | Typical trigger |
|---|---|---|---|---|
| Maintenance | Content, plugins, platform versions, small fixes | Design, templates, page map, platform | None | Routine upkeep, security updates |
| Refresh | Visual layer, key page copy, a few templates and components | Page map, URLs, platform | Low | Dated look, leaks on a few templates |
| Redesign | Page map, navigation, templates, messaging | Usually the platform and domain | Medium | New positioning, products or audience |
| Rebuild | Platform or codebase, often with a redesign on top | Domain, brand, the content worth keeping | High | Editing bottleneck, unsupported stack, platform limits |
Source: Octo Agency definitions; risk levels follow Google Search Central guidance on site moves, 2026
A rebuild without a redesign, same look on a new platform, is rare. Most rebuilds become redesigns too, because nobody wants to pay to rebuild a page map they dislike.
Why the smallest change is the default
Nielsen Norman Group's Hoa Loranger put the rule plainly in 2015: before throwing out the old site, make sure you have evidence that you need to. Her article lists the honest reasons for a major overhaul: incremental fixes have stopped paying, the technology blocks critical changes, the architecture has become a tangle after years of patching, conversion is low site-wide, or benchmarking shows competitors serve users clearly better. Boredom with your own site is not on the list.
The survey data shows how common the bigger move is anyway. Databox's 2024 survey of more than 145 mostly small companies found 49% had fully redesigned within the past two years. Orbit Media found top marketing brands replaced their sites every two years and one month on average, while its own 78 client sites lasted six years and four months. Sites that are kept current last longer. Our post Website Redesign Statistics 2026 covers those figures; this guide is about the decision.
Hour one: Search Console, to see what you could lose
Open the Performance report for search results. It shows clicks, impressions, CTR and average position, defaults to three months, and lets you change the date range and compare periods (Google Search Console Help). Set the longest range available and compare the latest 12 months with the 12 before.
Count where the clicks land. Sort pages by clicks and note what share the top 10 and top 50 URLs take. If a handful of pages carry most of the organic traffic, those URLs are the asset any redesign or rebuild must protect. That list becomes the first page of your redirect map.
Look for decline by page, not by site. A site-wide drop points to something structural. A drop on a few pages points to aged content, which is a maintenance job.
Check impressions without clicks. Pages with many impressions and a weak CTR often need better titles and descriptions, not a new site.
Record organic clicks, the trend against the prior year, and the top 10 pages' share.
Hour two: GA4, to see where visitors stop
GA4 counts a session as engaged if it lasts longer than 10 seconds, has a key event or has two or more page views; engagement rate is the share of sessions that qualify, and bounce rate is the rest (Google Analytics Help). Open the landing page report with sessions, engagement rate and key events side by side.
Group pages by template. Homepage, service or product pages, case studies, blog posts, landing pages. A redesign changes templates, so you want to know which templates fail, not which pages.
Find the leak. A template with traffic, a low engagement rate and no key events is leaking. If one or two templates leak and the rest work, refresh those templates. If every template leaks, the problem is likely the message or the page map, which is redesign territory.
Mark each template healthy or leaking. If key events are not configured, set them up first; no redesign decision should be made without them.
Hour three: speed and accessibility, to separate fixes from foundations
Speed. Run your five most visited templates through PageSpeed Insights and read the field data, not the lab score. Field data covers real Chrome users over the previous 28 days. The thresholds are LCP within 2.5 seconds, INP of 200 milliseconds or less and CLS of 0.1 or less, measured at the 75th percentile of page loads (web.dev, Google).
One failing template is a fix: an oversized hero image, a heavy script. Every template failing the same way points to the theme, page builder or hosting. Even then, check before blaming the platform. The HTTP Archive's 2025 Web Almanac found WordPress passed Core Web Vitals on 45% of mobile sites and concluded that implementation decisions often matter more than platform defaults. A slow WordPress site can often be made fast without leaving WordPress.
Accessibility. Run the same templates through an automated checker such as WAVE. WebAIM's 2026 scan of the top million home pages found an average of 56.1 detected errors per page and WCAG failures on 95.9% of them, but 96% of all errors fell into six types.
| Failure type | Home pages affected | Usually fixed by |
|---|---|---|
| Low contrast text | 83.9% | Refresh: colour tokens |
| Missing image alternative text | 53.1% | Maintenance: content entry |
| Missing form input labels | 51% | Refresh: form component |
| Empty links | 46.3% | Refresh: component markup |
| Empty buttons | 30.6% | Refresh: component markup |
| Missing document language | 13.5% | Maintenance: one setting |
Source: WebAIM Million, March 2026, 1,000,000 home pages; fix mapping is Octo Agency's
None of the six needs a new platform. Errors outside these types on every page, such as keyboard traps, may mean the templates need rebuilding.
Hour four: the CMS and the business, to test the foundation
Platform support. Check the versions you run. On WordPress, the PHP version matters: php.net lists 8.2 to 8.5 as supported on 21 October 2026, with security support for 8.2 ending on 31 December 2026. When we pulled WordPress.org's version statistics the same day, 36.4% of reporting sites ran PHP 8.1 or older and another 23.8% ran 8.2. An unsupported version is a maintenance job, unless the theme or plugins break on a current version. That condition is a genuine rebuild signal.
Editing cost. List the last 20 changes the marketing team asked for. Count how many needed a developer, and how long a new landing page takes from brief to live. Webflow's 2026 State of the Website survey of 1,000 marketing and technology leaders, a vendor-sourced figure from a company that sells a CMS, found only 34% were completely satisfied with their website and 97% of technical leaders said technical debt affected their ability to manage it. If most of your 20 changes needed a developer, the editing model is the problem, and no refresh fixes an editing model.
Business fit. Put your current offer and audiences next to the page map. If the navigation has no place for what you sell now, that is a redesign signal no metric will show.
Turning the afternoon into a decision
Score each finding against the table and go with the option that most signals point to. When two options tie, choose the smaller one.
| Finding | Maintain | Refresh | Redesign | Rebuild |
|---|---|---|---|---|
| Traffic declining on a few pages | Yes | |||
| Traffic declining site-wide with no technical cause | Yes | |||
| One or two templates leaking | Yes | |||
| Every template leaking, message unclear | Yes | |||
| Core Web Vitals failing on one template | Yes | Yes | ||
| Core Web Vitals failing everywhere, theme or builder bound | Yes | |||
| Accessibility errors in the six common types | Yes | Yes | ||
| Unsupported platform version that upgrades cleanly | Yes | |||
| Unsupported version blocked by theme or plugins | Yes | |||
| Most content changes need a developer | Yes | |||
| Offer or audience changed, page map no longer fits | Yes | |||
| Looks dated, numbers healthy | Yes |
Source: Octo Agency decision table, built on NN/g, Google Search Central, HTTP Archive and WebAIM findings cited above
A dated look with healthy numbers is a refresh: new type, colour and imagery, no URL changes. A changed business on a platform the team cannot edit is a redesign and rebuild together, the one case where doing both at once costs less.
What a rebuild costs you in search, and how to limit it
If the table says rebuild, price the search risk in before you commit. Google's site-move guidance says a small to medium site can take a few weeks for most pages to move, that rankings may fluctuate in the meantime, and that redirects should stay for at least a year. The Databox survey found SEO concerns were the top reason companies held back from updating their site, at 40.15%.
The SEJ figures from the opening, 523 days on average and 17% never recovering, come from domain migrations measured with third-party traffic estimates. Read them as a ceiling for what can go wrong, not a forecast for a same-domain rebuild with a complete redirect map. The protection is the work from hour one: every URL with clicks, links or key events, mapped to its new home before launch. Our post Website Migration Without Losing Rankings covers the redirect map and the 30-day watch in detail.
What this means for your next quarter
Most afternoons end with maintenance or a refresh: no URL risk, weeks rather than months. When the data points to a redesign or rebuild, you already hold the brief: the URLs to protect, the templates that leak and the editing problems to solve.
We should state our bias. We build marketing sites in Webflow and applications in Next.js, so when the table says rebuild, Webflow is what we would usually propose for a marketing site. We run marketing sites on a four-week process, and our USLege legal-tech redesign is on this site as a case study. We would start any redesign brief from the same four hours of data. We quote per project, and our marketing websites page explains how we scope a redesign or rebuild.
