B2B SaaS Website Redesign: A Framework for Not Making It Worse

Most website redesigns lose ground before they gain any, and the reason is almost always the same: nobody wrote down what was working before it was replaced.

A redesign is the only common business project that begins by deleting the thing it is trying to improve. Everything else — a product feature, a pricing change, a campaign — is additive and reversible. A redesign replaces a system that is currently producing pipeline with one that has never produced anything, and it does it all at once.

This is a framework for doing it without that being true.

The short answer

Four phases, in order, and the first one is the one that gets skipped:

1. Record — measure and freeze what currently works, before touching anything.
2. Diagnose — decide whether you have a design problem or a positioning problem. They need different budgets and different people.
3. Rebuild in slices — one template at a time, each measured against the old one.
4. Retire deliberately — keep URLs, keep rankings, keep what converted.

The recording phase is what separates a redesign that compounds from one that resets you to zero.

First: should you redesign at all?

A redesign is the most expensive answer to most website problems, and it is frequently the wrong one. Before committing, work through the cheaper diagnoses in why SaaS websites get traffic but no signups — measurement errors, traffic quality, a headline nobody understands, a form with too many fields. Several of those are afternoons, and any of them can produce the symptom that looks like "the site needs redoing."

A redesign is genuinely the right call when:

• The information architecture is wrong, not just the styling. You have six products and a navigation built for two.
• The positioning has changed and the site describes a company you no longer are.
• The system cannot express what you need. Every new page is a fight, so nobody makes new pages.
• It fails on phones, or fails accessibility, in ways that cannot be patched.

It is the wrong call when the actual complaint is "it looks dated" and nobody can name a business consequence. That complaint is usually real, but it is a restyle, and a restyle is a fraction of the cost and risk.

Phase one: record what works

Before a single screen is designed, produce two documents. This is the phase that gets cut for time, and cutting it is what makes redesigns dangerous.

The freeze list

Every page you are not allowed to break, with the reason. Typically:

For each, record what it currently does — entries, conversion rate, the queries it ranks for. This becomes the baseline you measure the new site against. Without it, "did the redesign work?" is unanswerable, and it will be answered by whoever argues most confidently.

The current-state audit

The second document is what the current site actually says, section by section. Which claims it makes. Which proof it offers. Which objections it handles.

Do this even if you hate the current site. It is where the accumulated knowledge lives — five years of sales objections quietly answered by a FAQ nobody has looked at since. Redesigns routinely delete that FAQ and then spend a year rediscovering the questions.

Phase two: design problem or positioning problem?

These get conflated constantly, and they cost very different amounts to fix.

That last row is the most reliable diagnostic in this article. If your team cannot agree on the headline, no designer can rescue it. A headline argument that runs for weeks is not a copywriting problem; it is the company not having decided what it sells and to whom.

Positioning is upstream of design. A clearer site aimed at the wrong buyer only helps that buyer reject you faster. If phase two says positioning, stop, resolve it, and come back — otherwise you will pay for a redesign twice.

Phase three: rebuild in slices

The instinct is to design the whole site, then launch it on one date. This maximises risk and destroys your ability to learn anything.

Ship by template instead, in order of risk:

Start with the lowest-traffic, highest-clarity template. Often a secondary page — about, careers, a single feature page. It proves the system, the components and the build pipeline while nothing important is exposed.

Then the templates that convert. One at a time, each measured against the recorded baseline for two to four weeks. If a new template underperforms the old one, you find out while one template is affected rather than forty.

Homepage last. It is the most-linked, most-scrutinised, most-argued-over page. By the time you reach it, the design system is settled and the arguments are about content rather than about colour.

Blog and article templates on their own track. They have different requirements — reading measure, heading hierarchy, structured data — and they carry your organic traffic. Rushing them alongside marketing pages is how sites lose rankings.

Keep the URLs

If a URL must change, redirect the old one permanently, one-to-one, to the closest equivalent. Not to the homepage — a bulk redirect of forty article URLs to / is treated as a soft 404 and forfeits every one of them.

Keep a spreadsheet of every changed URL and its destination. Check it after launch, not before, because the ones that break are the ones nobody predicted.

Phase four: launch and watch

The first two weeks are for detecting regressions, not celebrating.

Check, in this order:

1. Every freeze-list page still resolves. Crawl the site and compare against the list. Do this on day one.
2. Redirects return 301, not 302 or 404. A temporary redirect tells search engines not to transfer anything.
3. Analytics still fires. Redesigns break tracking constantly, and a broken tracker makes a successful launch look like a catastrophe — and vice versa, which is worse.
4. Core Web Vitals. Google's thresholds are LCP under 2.5 seconds, INP under 200 milliseconds and CLS under 0.1, assessed at the 75th percentile of real page loads. New sites regress here routinely, usually through unoptimised hero imagery.
5. Conversion by template against baseline. Give it a full sales cycle before drawing conclusions.
6. Organic entries by page. A ranking drop shows up over weeks, not days. Keep watching for a quarter.

Expect a dip. Search engines need to recrawl and reassess, and a temporary decline in the first few weeks is normal even when everything was done correctly. What matters is whether it recovers to and past the baseline — which you can only know because you recorded one.

What to do when a page gets worse

It will happen on at least one template. The recording phase is what makes this recoverable rather than mysterious.

Compare the old and new versions of that specific page against the audit. Usually one of:

• A proof element was dropped in the redesign because it did not fit the new layout — a customer quote, a security badge, a specific number.
• The headline was made more elegant and less clear. Redesigns tend to move copy toward the aspirational, because that is what reads well in a design review and badly on a cold visitor.
• An objection-handling section was cut for length.
• The CTA changed wording and lost its terms — "Start free, no card" became "Get started."

Put the missing thing back. This is not a failure of the redesign; it is the recording phase doing exactly what it was for.

Frequently asked questions

How long does a B2B SaaS website redesign take?

Sliced as described, a small marketing site is typically two to four months from recording to homepage launch, with templates going live throughout. A single big-bang launch appears faster on a plan and is slower in practice, because everything argued about must be resolved before anything ships.

Will a redesign hurt our SEO?

It can, and the damage is almost always self-inflicted: changed URLs without redirects, thinner content, or removing pages that were ranking. Keep URLs, keep content depth, redirect one-to-one, and expect a temporary dip while search engines recrawl. Sites that lose rankings permanently usually did one of those three things.

Should we redesign the blog at the same time?

Separately, and usually after. Article templates have their own requirements and carry organic traffic that is expensive to lose. Doing both at once means a ranking drop has two possible causes and you cannot tell which.

Framer or a custom build for the new site?

It depends on who maintains it afterwards and what the site has to do, more than on the design itself. We have written the honest comparison — including where each one runs out of road.

How do we know the redesign worked?

Only by comparing against the baseline recorded in phase one, per template, over at least one full sales cycle. Without that baseline the question has no answer, and it will be settled by opinion. This is the single strongest argument for not skipping the recording phase.