B2B Website Redesign: Rebuilding Without Killing Your Pipeline

A company decides the site looks old. They bring in a studio, get something cleaner, launch it, feel good about it for a week. Then the demo requests don’t move. Sometimes they drop. Six weeks later somebody opens Search Console, half the old rankings are gone, and nobody in the room can say why. I’ve been called in to fix that mess more times than I’d like.

A B2B website redesign is a rebuild of a site that already has traffic, rankings and a pipeline attached to it, run as one plan: strategy, information architecture, content, SEO migration, interface, development, measurement, in roughly that order. New colors, new type and a new homepage are the last ten percent of the job and the part everyone has opinions about. The first ninety decides whether demo requests go up or rankings go away.

The technical layer any new site needs (semantic core, Core Web Vitals, schema) is in SEO Starts Before the First Line of Code, which assumes a clean sheet. This guide assumes you have equity you can’t afford to set on fire. Most of what ranks for this query is an agency describing its process; what follows is the process with the failure points marked, from autopsies on redesigns that died the way I just described. Some of the screenshots come from Algorithm.com, the analysis tool I built when I got tired of arguing about SEO from memory.

Website refresh, redesign, rebuild or rebrand: scope the project first

Four words get used for four different projects. The one you pick decides the budget, the timeline and how much of the migration section below applies to you, so it comes first.

The last row is the one people underestimate. A rebrand moves the entity along with the URLs. Google and the AI engines hold a picture of who the company is, assembled from your site, your profiles and third-party mentions: name, domain, people, what it’s known for. Change the name and the domain in the same quarter and that picture gets rebuilt from scratch, which is slow even when every redirect is perfect. A rebrand that changes what the company is needs a year of entity work alongside the migration: Organization schema with sameAs pointing at the renamed LinkedIn, Crunchbase and Wikidata records, the old name kept in alternateName, press coverage that uses both names in one sentence, and the Change of Address tool in Search Console, which only exists for domain moves and does nothing for URL changes inside a domain. Google’s own guidance is to keep the redirects in place for at least a year; I’ve yet to see a reason to remove them at all. A rebrand that changes the logo is a refresh; call it one and keep the name.

Everything below is written for the two middle rows. A refresh needs the baseline export and the title-parity check and little else. A rebuild needs all of it, twice as carefully.

When a B2B website needs a redesign: two failure modes

The standard triggers are legitimate: conversion has sagged and the funnel math no longer closes, the brand has moved on and the site still tells the story from three years ago, the CMS fights every change, positioning has shifted to a segment the site never speaks to, or the company has outgrown a site built for ten people and needs governance, localization, a resource center. Every framework in the top ten lists some version of these five, and they’re right as far as they go.

The trigger tells you the site is in pain, and stops there. Every weak B2B site I’ve audited has one of two problems underneath, and I sort by that before anything else. One: the right people never see it. Your buyers ask narrow questions, in Google and increasingly in ChatGPT, Perplexity and the AI Overviews: “[category] for a 200-person fintech,” “how do we do [the job] without hiring for it.” If no page on your site is the answer to a question like that, the buyer never meets you, and no visual refresh or ad budget changes it. Two: they see it and leave. Traffic’s fine, demos aren’t. This is the one everybody reaches for a redesign to fix (new hero, shorter form, cleaner homepage), and it’s the one a redesign fixes least reliably, because the problem is rarely visual.

There’s a version of the second mode that isn’t a redesign problem at all. Traffic is healthy, the pages load, the design is fine, and the demo rate is still flat. Nine times out of ten that’s positioning: the site is clear about what the product does and silent about who it’s for and why it beats the alternative the buyer already has. A redesign repackages that silence in a nicer grid. If you can’t write the one-sentence answer to “why us, for this buyer, instead of the incumbent” before the project starts, fix the sentence first. It costs nothing and it’s the brief for everything below.

The pain recurs after a redesign because of how the project gets staffed. A studio takes the visuals. An SEO agency gets “looped in.” A freelancer writes the words. Three competent people, zero shared plan, and the money leaks out of the gaps between them. I inherit both wrecks this produces: the beautiful invisible one, minimal and nicely done and a ghost to Google, URLs shuffled with no redirect map, ranking pages binned for not fitting the new grid; and the opposite, keyword sludge wall to wall, no hierarchy, no air, no clue who it’s for. Both come from running the two jobs as a relay, build it then optimize it, when every SEO decision that changes anything is a build decision.

B2B website redesign process, step by step

Baseline export and SEO audit before design starts

The first deliverable of a redesign is an export. Wireframes come after. Search Console keeps sixteen months of data and then drops it, so the first thing I do is pull the full window, queries, pages, countries, devices, before the project changes anything. Then the ranking URLs with their positions, backlinks per URL (a per-domain count hides which pages hold them), conversions by landing page from analytics and from the CRM, and a full crawl of the current site so the internal-link graph exists somewhere other than the live server. Two settings decide whether that export is even possible. The Search Console UI caps a table at 1,000 rows, so on any site with more than a few hundred URLs the baseline comes through the Search Analytics API or the Looker Studio connector, not a CSV download. And GA4 ships with event-level retention set to two months; if nobody switched it to fourteen before the project, the pre-launch conversion data you’ll want for explorations is already gone by the time the site is live. Check that setting on day one. Six months in, when someone asks whether the old pricing page was earning, that export is the only place the answer lives.

The baseline feeds the audit that follows: which pages carry the pipeline, which ones rank for intents the site doesn’t serve, where good traffic stops turning into deals. That’s its own discipline, and I’ve written up how I run it in a B2B SEO audit; run it yourself, or have us run it. The redesign inherits its findings.

Stakeholder alignment and one primary metric

The second deliverable is a one-page agreement, and it is harder to get than the export. Marketing wants more leads, sales wants better ones, product wants the new module on the homepage, the CEO wants it to look like the competitor that just raised. All four will sign off on wireframes, and if their definitions of success differ, the site gets redesigned again in eighteen months.

What I ask for: one primary metric, two supporting ones, and a named segment the site is allowed to ignore. For most B2B sites the primary metric is qualified demos or SQLs per month from the site, counted in the CRM. Analytics counts submits; the CRM counts the ones sales accepted. Supporting metrics are usually organic leads by landing page and the MQL→SQL rate. The ignored segment is whichever one makes the homepage impossible: students, job seekers, the ten-seat customer a company now selling to enterprise no longer wants. Writing that segment down is what lets the homepage speak to one buyer.

The metric also fixes the review criteria. A design gets judged on whether it moves a security lead to the security page in one click, not on whether the CEO likes the hero. That changes the meetings more than any framework does.

Buyer profiling and the B2B buying committee

Most guides say “understand your audience” and move on, which helps no one. Drop the demographics for a second. What’s the problem this buyer is trying to solve, and what’s the voice in their head when they hit the page? A VP of Engineering checking out an infrastructure tool isn’t thinking “I’d love a flexible, enterprise-grade solution.” She’s thinking will this fall over in production and will my CTO have my head for it. Answer the fear, not the feature.

Then read where the buyer sits, because it rewrites the design. An enterprise security lead reads white space and restraint as competence, opens the security page before anything else, and sees a “chosen by 1,000+ companies” strip as a consumer product wearing a suit. An SMB owner reads the same restraint as a site nobody finished, wants the price on the page and the logos above the fold, and bounces off a form that asks for company size. Good design here is whatever this buyer reads as trust, and the two of them read it backwards.

Then the part that separates B2B from B2C: the committee. A consumer decision is one person, one sitting; a B2B decision is a long cycle that finance, IT and leadership will all pick apart before anyone signs. Forrester’s 2025 buyers’ survey puts the typical decision at thirteen internal stakeholders and nine outside influencers, and the same survey has 94% of them using generative AI somewhere in the process. Gartner’s numbers, older and still the ones procurement quotes back at you: buying groups of six to ten, each member arriving with four or five pieces of information they gathered on their own, and 17% of the total buying time spent with any supplier at all, which for a four-vendor shortlist leaves a rep with about 5% of the buyer’s attention. The website gets the rest. These buyers judge credibility before capability, starting on the first screen, and one homepage can’t carry three roles: the engineer, the RevOps lead and the CFO each arrive with a different problem and a different bar for trust.

Positioning and messaging before UI

Design buys the first few seconds. Copy has to carry the buyer from “interested” to “did something,” and the copy can’t be written until the positioning is. Before anyone opens Figma I want six lines on one page: the category, who it’s for, the problem, the promise, what makes it different from the incumbent, and the proof. If those six lines don’t exist, the designers fill the hero with “powerful, flexible, enterprise-grade,” which every competitor’s homepage also says and which is therefore worth nothing. The test for the six lines is mechanical: paste them under the nearest competitor’s logo. If they still read true, you have a category description, and positioning hasn’t started.

Lead with the problem and your edge; the feature list has its own page. Name the problem the buyer has, then plant a point of view they’ll remember: “see production incidents before your customers do” earns a different kind of attention than “comprehensive infrastructure monitoring.” The messaging page then feeds every template. Hero, solution page, comparison page, the headline above the demo form: one source, so the site stops contradicting itself between pages written by different people in different months.

Competitor analysis: table-stakes blocks and open slots

Do the standard teardown: which blocks the top sites have (FAQ, comparison table, pricing, the usual trust elements) and make sure you’ve got them too. A page missing the blocks buyers expect feels unfinished; they notice, and so does Google. But matching the top ten is a floor. If your new site is a tidier remix of what already ranks, you’ve built a copy, and copies hand the index nothing it didn’t already have.

I put a number on the floor. Take the five pages ranking for the query your solution page wants and score them on the blocks they carry: published pricing or a range, a comparison against named alternatives, an integrations list, a security or compliance page, named people with credentials, an implementation timeline, a guarantee with terms. Anything four of the five carry is table stakes; without it the buyer reads your page as incomplete before reading a word. Anything none of them carry and a buyer would use, a cost calculator, a migration-effort estimate, live onboarding dates, is open ground. Owning one of those slots beats matching ten of the others.

Information architecture and navigation: one page per intent

Most B2B sites are filed under the company’s own furniture, Features, Solutions, Integrations, Resources, and buyers don’t search or think in that vocabulary. They arrive with an industry, a problem, or a job to be done, and the site asks them to translate. A data platform refiled around buyer context, “business intelligence for operations teams,” “embedded analytics for product teams,” catches searches a single Analytics page will never see, and it gives each visitor a page written for the problem they came with.

The architecture that does this runs down a ladder: Industry → Solution → Use Case → Product → Proof → Conversion. Not every site needs every rung, and a 40-page company shouldn’t build 400 pages to fill them. But every commercial intent the audit found needs a home somewhere on that ladder, and each home carries one intent. The mechanics of mapping intents to URLs, semantic core, clustering, the URL scheme, are in the build guide linked above. The redesign question is narrower: does the existing site have the rungs, and does each page stay on its own rung?

Primary navigation follows the same rule. Solutions by industry and by role go in the menu; the company’s own categories go second. A homepage’s job is to route, three doors for the three committee roles, with one primary CTA and the proof each role needs one click away. The legal-services buyer, the property developer and the industrial manufacturer live in different worlds; a homepage talking to all three talks to none. One framework, tuned per industry.

Staying on one rung is where I find the most damage, and it’s measurable. A solution page exists to move a buyer who already knows the category toward a demo. Score its passages by the intent they serve and a surprising number of “commercial” pages turn out mostly educational: a definition of the category, a history of the problem, three paragraphs of why this matters before the product appears. The ranking system reads that page as informational and it competes with guides instead of vendors. On one page I scored recently, 54% of the passages were informational and the page turned informational in the first paragraph. The buyer who arrived ready to evaluate was being taught; the buyer who arrived to learn was two clicks from a demo button they had no use for.

The fix during a redesign is cheap: the educational passages move to a guide that links down to the solution page, the solution page keeps what a buyer needs to choose, and internal linking makes the relationship explicit both ways. The check that the fix took is in Search Console, not in the copy deck: filter the Performance report to the solution page and look at the queries it earns clicks for. If the top ten are “what is [category]” and “how does [category] work”, the page is still a guide, whatever the template says.

UX, conversion paths, forms and proof placement

Every design article opens with the same first-impression studies, so I’ll skip the recital. What they leave out is that the split-second verdict gets rendered against this buyer’s expectations, and there’s no universal standard to fall back on, which is why the profiling work above is the design brief. “Make it modern” is how you end up with a costly site that doesn’t sell.

What I build toward starts with functional minimalism, and I mean the word functionally. Strip everything that doesn’t help the next step, and white space becomes what makes the one important thing legible. Hierarchy then walks the buyer from “what is this” to “why it’s for me” to “what now” in an order that feels almost inevitable. Proof sits where the doubt spikes: technical credibility up top for an engineer, ROI and peer logos mid-page for a business buyer, a low-commitment next step at the end, and none of it parked in a testimonial carousel at the bottom of the page. Density gets matched to the buyer from the profiling stage. Don’t force minimalism on people who read it as unfinished. Don’t bury people in clutter who read clutter as a con.

Mobile is where B2B sites still lose the field buyer: the plant manager, the site engineer, the distributor rep placing an order from a truck. Templates get signed off on a 27-inch monitor and the demo form gets tested on a phone the week before launch, if at all. Walk the demo path on a mid-range Android over 4G at the wireframe stage; the full list of what to check on a phone before sign-off is in our mobile SEO checklist. Core Web Vitals per template is a launch gate later, at the p75 thresholds Google publishes (LCP 2.5 s, INP 200 ms, CLS 0.1), and it’s easier to pass when nobody designed a 4 MB hero for desktop first. INP is the metric a redesign is most likely to break, because it measures what the new site adds: a mega-menu on a heavy framework, a chat widget and a consent banner all initializing in the same second, each one competing for the main thread on the phone the field buyer carries. The check is the INP breakdown in PageSpeed Insights per template, on mobile, before launch, when the widgets can still be deferred.

The most common self-inflicted wound is the form, and I’ve stopped quoting conversion statistics about field counts because the ones that circulate are a decade old and unsourced. What I test instead is three questions per field. Does sales need this answer before the first call, or could they ask it on the call? Could it be inferred instead of asked, company size from the email domain, industry from the company? Does the answer change anything: routing, priority, the follow-up sequence? A field that fails all three is friction with no payoff, and most B2B forms carry four or five of them. What survives is usually name, work email, and one field that changes routing. Everything else is enriched after the submit from the email domain (HubSpot’s Breeze Intelligence, Clearbit before it, ZoomInfo, or a Clay workflow), and a free-mail domain routes to nurture instead of to an SDR’s calendar. Lead qualification happens after the submit, in the CRM. One primary CTA per page is my default, and I test against it rather than assume it.

Website copy and AI-generated content

With the messaging locked, the trap on the copy stage is one that didn’t exist five years ago: letting a model write it unsupervised. Google’s position is clearer than the industry gossip suggests. AI-generated text carries no penalty for being AI-generated. The penalties are for scaled content abuse, the March 2024 spam policy against producing many pages primarily to manipulate rankings regardless of how they were made, and separately for a site whose content is unhelpful overall, which since the same update is scored site-wide. Google’s own after-action number for that update was a 45% reduction in low-quality, unoriginal content in results, and the sites that absorbed it were mostly not spam farms; they were ordinary domains with a large templated section. A hundred templated solution pages spun from one prompt trips both. Buyers have their own detector and it’s harsher: B2B readers have developed an allergy to copy that reads like a polished template, and on a page whose job is trust, that allergy costs demos directly. I use models inside a human process: strong brief, cases, data, then a human edit that scrapes the machine patterns off. Faith in a magic prompt gets you digital landfill.

Design system, CMS ownership and accessibility

Three decisions here look like engineering and are marketing decisions. First, the design system: components, not pages. If the solution template is a set of blocks the CMS can rearrange, marketing adds the eleventh industry page in an afternoon. If every page is a composition, each new landing page is a design project, and the architecture above stops growing the week the agency leaves.

Second, who owns the stack. The test is whether a marketer can publish a new solution page, change a CTA and drop a case study into the proof block without filing a ticket. Webflow collections, WordPress block themes with the tokens in theme.json, a headless CMS with a working preview: any of them pass. A hand-coded template passes Lighthouse and fails this one. The detail to write into the contract is the content model: a case study should be a structured item with industry, product, metric and quote as fields, so the proof block on each solution page can pull the three relevant cases by industry. A case study saved as a rich-text page can’t be pulled by anything. Platform choice itself is simpler than the vendor comparisons make it: stay if the platform can render what you need and the team can edit it; move only if the CMS is the bottleneck, because moving turns the project into a rebuild and prices like one.

Third, accessibility, which in B2B has moved from the ethics column to the procurement column. Enterprise and public-sector buyers now ask for a WCAG 2.2 AA conformance statement in the RFP; the European Accessibility Act has applied to new digital services since June 2025, and the US ADA Title II rule puts state and local government sites on WCAG 2.1 AA from April 2026, which reaches every vendor whose product embeds in them. Of the nine criteria 2.2 added, three hit B2B sites directly: 2.5.8 Target Size, 24×24 CSS pixels minimum, which most icon-only nav buttons fail; 3.3.7 Redundant Entry, which forbids a multi-step demo form from asking for the company name twice; and 2.4.11 Focus Not Obscured, which a sticky header and a cookie banner break together. The design system is the cheapest place to get there: contrast, focus states, form labels and keyboard order fixed once at component level. A page-by-page audit after launch costs more and finds the same six issues.

Development: JavaScript rendering, staging and AI crawlers

The build mechanics live in the technical guide; what belongs here is what I check inside the build. Rendering first: if the new front end ships content through client-side JavaScript, the HTML Googlebot receives gets compared against what the browser shows before any template is signed off, because the difference between those two views is the most common reason a redesigned site indexes worse than the one it replaced. The comparison is one click in URL Inspection: “View crawled page” shows the HTML Google fetched, and anything that appears only after a click, a tab change or a scroll event isn’t in it. Google’s renderer will wait for most fetches; it won’t click. And Google is the generous one. The Vercel and MERJ crawl study from December 2024 found that OpenAI’s, Anthropic’s and Perplexity’s crawlers fetch JavaScript files and execute none of them, so a solution page whose definition is injected client-side is blank to every assistant, whatever robots.txt allows.

Then staging. The block that works is HTTP authentication returning 401 plus an X-Robots-Tag: noindex header; a CMS-level password page returns 200 and gets indexed as a login screen, and pairing noindex with a robots.txt Disallow is self-defeating, because a crawler that’s disallowed never reads the noindex. Every redirect from the migration map gets tested on staging, and removing the block goes onto the launch checklist as its own line. A staging noindex shipped to production is the launch-day bug I get called about most, with canonicals still pointing at the staging domain a close second.

The same build is what the AI assistants read, and a redesign is where their access gets switched off by accident. Review robots.txt by agent name, because the agents aren’t interchangeable. OpenAI runs three: GPTBot for training, OAI-SearchBot for ChatGPT search, ChatGPT-User for the fetch a user’s question triggers live. A rule written in 2023 to keep GPTBot out doesn’t touch search, and a blanket block written last year keeps you out of the answers. Google-Extended controls whether your content trains Gemini; it has no effect on AI Overviews or AI Mode, which run on Googlebot. Anthropic splits the same way (ClaudeBot, Claude-SearchBot, Claude-User), as does Perplexity (PerplexityBot, Perplexity-User), and Copilot reads the Bing index, so Bingbot decides that one. And since July 2025 Cloudflare blocks AI crawlers by default on new zones, so a redesign that moves the site behind Cloudflare can turn the assistants off without anyone making the decision. Last check: the definition of each solution lives in plain text on the page. A hero video is invisible to every one of them.

Set up the analytics segment before launch: ChatGPT appends utm_source=chatgpt.com to the links it hands out, Perplexity and Copilot pass a referrer, and Search Console folds AI Mode clicks into the web report without a separate filter. What I don’t have is a clean number for how much B2B pipeline arrives that way. The assistants don’t report clicks, some referrers arrive stripped and land in direct, and the studies claiming a conversion uplift for AI referrals don’t publish a method I can check. On client sites the segment has been doubling or tripling year over year from a small base, and it lands on solution pages more often than on blog posts.

Measurement: GA4 and CRM event parity, then the test plan

A redesign renames things. GA4 events get new names, the HubSpot or Salesforce form IDs change, the CRM field that stored the landing page gets a new key, the UTM convention gets tidied. Each change on its own is sensible. Together they mean that six weeks after launch, “leads by landing page” can’t be set against the baseline, and the question the whole project was supposed to answer has no answer. Before launch I want a mapping of every conversion event, old name to new, with the old names kept as GA4 key events until the first full month of the new ones exists; a test lead submitted from each form template and traced to the CRM record with the landing page attached, because a HubSpot form rebuilt in the new design gets a new GUID and every workflow keyed to the old one goes silent without an error; hidden fields on every form for landing page, referrer and the UTM set, written server-side so a consent banner can’t blank them; and an inventory of the URLs currently live in Google Ads, LinkedIn, nurture emails and the sales deck, because every one of them is about to hit a redirect or a 404.

One lag to plan around: the Core Web Vitals report in Search Console runs on CrUX field data over a rolling 28 days, so a template’s verdict after launch reflects a month of mixed old and new traffic before it says anything about the new site. For the first month, per-URL field data in PageSpeed Insights and your own RUM (the web-vitals library into GA4 is a two-hour job) are the only readings that describe the site you shipped.

Launch is the start of the test plan. Before go-live I want a 30/60/90 list: which template gets the first A/B test, which form field comes out first, which solution page gets its second case study. Without it the site freezes on launch day and the clock on the next redesign starts.

SEO migration during a website redesign: what the 301 checklist misses

This is the section I’d move to the front if I trusted you to read it there. Every migration article hands you the same line: set up your 301s. Correct. Necessary. Nowhere near the whole story. I’ve watched sites with a spotless redirect map still shed 30–40% of their traffic after a redesign, because the damage came from places the checklist never mentions.

Pages cut while still ranking. In a redesign, pages get cut for not fitting the new structure or looking dated. Some of them were earning rankings, links and traffic you never gave them credit for. Before anything gets deleted, every URL gets crawled and checked against the baseline: traffic, rankings, backlinks, conversions. Then each one gets one of four verdicts, and “delete” is the rarest:

The internal link graph, shredded by a new template. Your old site had a web of internal links telling Google how pages relate and where authority pools. Rebuild navigation and templates from scratch and you can shred that web without anyone noticing: the pages survive, the relationships don’t, rankings sag a few weeks later, and everyone blames the algorithm. The crawl export from the baseline is what you diff the new site against: Screaming Frog’s Compare mode set to the old crawl and the staging crawl, sorted by the change in unique inlinks. If a page that had forty internal links now has three, that’s the finding, and it’s invisible without the export. The usual cause is structural, not editorial: the old footer linked every solution page, the new footer links four, and a related-content block that used to be hand-curated now shows “latest posts.”

Titles and headings rewritten in a “copy refresh.” A title tag and an H1 that rank are assets. Swap them wholesale and you can reset a page’s relevance overnight. Google also rewrites titles in the results from the H1 and the anchor text pointing at the page, so changing the H1 alone changes what the SERP shows even when the title tag is untouched. Preserve first, improve second, and log every deliberate change with the date so a drop three weeks later can be matched to it.

Intent drift. An informational resource that ranked for years becomes a hard sales page, or two intents get merged into one page for cleanliness. Google was ranking the old intent; the new one ranks for neither.

Redirect mechanics done half-way. 301, not 302, for anything permanent. Google has said on the record that 3xx redirects don’t lose PageRank and that Googlebot follows up to ten hops, so chains aren’t fatal, but every hop is a place something can break; flatten them. The redirect itself carries no penalty; what looks like a 301 redirect SEO penalty after a relaunch is almost always one of the four traps above wearing a redirect’s clothes, and that article takes the myth apart case by case. Then point internal links at the final URLs instead of leaving them to ride the redirects: crawl the new site on staging and filter internal links that answer 3xx. On a rebuilt navigation that list is rarely empty. Two things from Google’s own site-move documentation that almost nobody does: keep a sitemap of the old URLs submitted for a few weeks after launch so Googlebot recrawls them and processes the redirects sooner, and leave the redirects up for at least a year. And for the launch-day check itself, the URL Inspection API allows 2,000 requests per property per day, which is enough to verify every ranking URL on a mid-sized site the same afternoon rather than sampling.

For scale: the range every checklist quotes is a 10–25% traffic dip in the first thirty days and two to eight months to recover, and as an average I’ve no reason to argue with it. What the average hides is that one variable does most of the work: the share of URLs that moved. A redesign that keeps its URLs and only swaps templates sits at the bottom of that range or below it. A rebuild that moves every URL and rewrites the navigation sits past the top, and the recovery clock on it starts only when the last redirect is right. The careless version turns the dip into a cliff you might never fully climb back. The protocol that keeps it survivable, as a checklist for the project plan:

Above all, track leads by landing page. A site can hold its total traffic steady while the three pages that drove demos flatline. The status-code mechanics under all this, when a 410 beats a 301, how canonical and redirect signals interact, how to read a redirect audit, are in the technical SEO audit guide. This section is the traps the 301-only advice skips.

Blog and content strategy during a redesign: information gain vs topical authority

A redesign is when somebody decides to fix the blog, and the default plan (map every cluster a competitor ranks for, produce a matching article for each, “own the topic”) doesn’t just fail in 2026. It can drag the whole domain down with it.

Two dates explain why. August 2022: Google ships the helpful content system, a site-wide signal reading how much of a domain is written for people versus for rankings. March 2024: that signal folds into core ranking, and the same update adds scaled content abuse to the spam policies. A blog full of rehashed articles no longer fails on its own terms and leaves the money pages alone. Google scores the domain as one object, and the solution pages that sell inherit the blog’s grade.

What replaced volume is information gain: how much a page adds to what the index already holds for that query. A cluster with no new data, no fresh case, no original connection between things is a tenth restatement of an answer that already exists. The AI surfaces are stricter still. In early 2025 Ahrefs found roughly 76% of AI Overview citations came from pages ranking in the top ten for the query, which means a quarter didn’t, picked for being a source rather than for position. Ranking well helps. Having something the other pages don’t is what gets cited.

The redesign-specific version of the mistake is cannibalization at scale. The old blog already had two thin posts chasing one intent; the topical-authority plan adds a third, and now three pages split the signal that should have gone to one. A map of the content clustered by semantic similarity shows it in seconds; the folder tree never will. The red pairs are the pages Google can’t tell apart, and each pair is a merge decision for the verdict table above. Search Console confirms it without any tooling: filter the Performance report to one query, switch to the Pages tab, and if two URLs trade the impressions between them week to week, neither is holding the position. The merge keeps the URL with the more backlinks, moves the unique passages from the other, and 301s it.

The blog plan I hand clients during a redesign is short. Grade the existing posts on what they add and ignore their length; merge or cut the rest; the grading method is in the audit guide. Then write what isn’t in the top ten. One article with a case, screenshots, numbers and a named author with a footprint you can verify will out-earn a hundred pages of filler, because in B2B the blog is where a buyer decides whether you know the work. Five to ten of those is a year’s plan. I’ve had a site under a month old show up in rankings on a handful of pieces like that and a clean build alone. That doesn’t happen with rewrites.

Timeline and cost: what moves the estimate

Timeline first, because it’s the one people get wrong by a factor of two. Content stretches it more than design does: interviews and drafts for twenty pages take longer than the templates that hold them, and the approval loop for a security page can outlast the build of the entire site.

Cost is a function of the same variables, which is why any range quoted without them is a guess dressed as a number. The factors, roughly by weight: the number of templates, not pages (a 200-page site on eight templates costs less than a 40-page site on twenty, because templates are what get designed, built and tested); whether the CMS moves, which turns a redesign into a rebuild and adds content export, field mapping and every integration re-wired; integrations with CRM, marketing automation, product data, scheduling and chat, each a project inside the project; content rewriting, the line most often underestimated because it looks like a freelancer’s afternoon and is usually two months of interviews and drafts; the SEO migration itself, URL mapping, redirect testing, the internal-link diff and the post-launch watch; multilingual architecture, which doubles the templates in practice; a design system, cheap if the site keeps growing and expensive if it never gets used; and enterprise governance, approvals, accessibility compliance, legal review, which is slow more than it is expensive, and slow is expensive.

The one number I’ll put on it: a B2B website redesign done as the sequence above, with the baseline, the migration and the measurement layer in scope, starts at around $10,000 and goes up from there with the list. I won’t quote a ceiling, because every agency in the top ten quotes a different one and none of them has seen your site; the two lines that move the estimate most are integrations and content, and both are unknowable before the audit. Budget 10–15% on top for the things nobody scoped, because there are always things nobody scoped.

SEO migration is the cheapest line on the estimate and the most expensive line to cut, on every budget I’ve seen. Removing it saves a few percent of the build and risks the entire organic channel the redesign was supposed to grow. If the budget has to lose something, lose a template.

Where the weight sits, by company type

The sequence above is the same for every B2B site. What differs by company type is which stage carries the weight, and where the redesign most often goes wrong.

Know which row you’re in before the sequence starts. It decides where the audit looks first and which template gets the design hours.

Common redesign mistakes, by stage

The traps are scattered through the sections above where they happen. Collected in one place, for the project plan:

ROI: when not to redesign

I’ll close on the section no agency wants to write, which is why it earns the most trust. A serious conversion-focused redesign stopped being “affordable marketing.” It’s expensive, complicated and high-risk, while CTR keeps sliding as AI answers swallow the informational clicks. For some companies the most useful thing I can say is: don’t do this yet. Keep your money.

Three numbers decide it. First, what a closed deal is worth to you. Second, how many qualified leads a redesign would realistically add per month, counted from your current traffic and conversion deltas you’d defend in front of the CFO. Third, the all-in cost from the section above, including the months of monitoring and iteration after launch. Multiply the first two, set them against the third, and believe the result over anyone promising you easy traffic.

If it does close, the work is the sequence above, run as one plan by people who can hold design and search in the same head. That’s what we do at Neon: UX/UI design and custom web development with the migration in the plan from the first meeting. If the math doesn’t close, we’ll say that too.

If the funnel math doesn’t close, no amount of beautiful design or clever SEO changes the answer.

My bet for the next two years: the redesigns that survive will be the ones where the migration plan and the intent map existed before the moodboard, and by 2028 the “we lost half our rankings after the relaunch” story will read the way “we built the whole site in Flash” reads now, a known mistake nobody serious makes twice.

FAQ

How long does a redesign take?

Three to six months for a mid-sized site staying on its platform; six to twelve if the CMS moves; longer if the site is multilingual. Content stretches it more than design does.

Do rankings drop during a website redesign?

Some, temporarily, almost always. The range every checklist quotes is a 10–25% dip in the first month and two to eight months to recover; where you land in it depends mostly on how many URLs moved. Skip the URL mapping or shred the internal links and the dip becomes permanent.

Can we redesign without changing URLs?

Usually, and it’s the single cheapest way to protect rankings. New templates, new design system, new copy on the same URLs keeps the migration section to title parity and an internal-link diff. URLs change when the architecture does, and then the change is deliberate and mapped.

Website redesign or rebuild: how to choose?

Redesign if the platform can render what you need and the team can edit it. Rebuild if the CMS is the bottleneck or the site can’t be crawled without JavaScript gymnastics. A rebuild is a redesign plus a platform migration, and it costs like one.

Should we move the CMS and redesign at the same time?

Only if the CMS is the reason for the project. Doing both at once moves every URL, every template and every integration in one launch, which is the worst case for the migration and the hardest to debug when something drops. If the platform is merely old, redesign on it first and migrate later, when the new architecture is proven.

How much does a B2B website redesign cost?

From about $10,000 for a redesign that includes the baseline, the migration and the measurement layer; above that it depends on templates, integrations, content and whether the CMS moves. What holds without seeing the site: cut anything from the estimate before you cut the migration line.

How do you measure the results?

Leads by landing page, MQL→SQL rate, and pipeline attributed to organic, compared against the baseline you exported before the project started. Traffic on its own answers a different question. If that export doesn’t exist, you can’t answer this one, which is why it’s the first step.