Direct answer
Every claim is sourced below
A redesign fixes a low conversion rate only when it changes whatever is stopping visitors from converting. If the cause is a message that doesn't say what you sell, or a structure that hides the next step, a redesign that rebuilds those can raise the rate. If the cause is the wrong visitors, a weak offer or a demanding form, a redesign that leaves those alone leaves the obstacle in place. So the first job is finding the cause.
A conversion rate is a count of actions divided by a count of chances to act. The problem can be in either number. Too few chances end in the action, or the total includes many that were never going to.
Lean Labs' article on conversion problems names five problems that sit behind a low rate: not understanding the audience, a product description that leaves visitors unsure what is sold, too little reason to trust the company, a poor pricing strategy and no testing. Look at where the first four live. Audience is about who arrives, and the message is the words on the page. Trust depends on the proof you can show, and price belongs to the offer itself. Without testing, the team has less evidence for choosing what to change. Layout still plays a part, because it affects whether a visitor notices the message or the proof. But each cause has a part that a change of appearance leaves alone.
That doesn't make a redesign the wrong tool, because redesign projects differ in scope. One project starts with customer research, rewrites the message and reorganizes the pages around what buyers ask. Another keeps the copy and the forms and changes how they look. The first can reach most of the causes on that list. The second changes how the same content is presented, which helps when presentation is the obstacle. It leaves an obstacle in the wording, the audience or the offer intact. How the pages get built is a separate choice, which we cover in modular web design versus custom development.
Suppose a company sells scheduling software to dental practices. Its site gets 5,000 visits a month, where a returning person counts as a new visit, and 50 of those visits end in a demo request. That's a rate of 1%. Suppose the steps in between look like the table below.
Where the 5,000 visits go
| Step | Visits in a month | Share of the step before |
|---|---|---|
| All visits to the site | 5,000 | Starting point |
| Visits that reached the demo page | 1,000 | 20% |
| Visits with a form start | 200 | 20% |
| Visits with a submitted form | 50 | 25% |
The same table fits several different stories, and each one calls for different work.
In one story, the problem is who arrives. Suppose 3,000 of the 5,000 visits land on a blog post that offers a free holiday schedule template, and none of those visits ends in a demo request. The readers are office staff looking for a template, and they aren't shopping for software. Take them out and the site converts 50 of 2,000 visits, or 2.5%. The 1% said more about the blog post than about the rest of the site, and a redesign that keeps the post and its search traffic as they are leaves that audience unchanged.
In another, the problem is the form, where 150 of the 200 form starts end without a submission. Suppose it has eleven required fields, and recordings show people stopping at the questions about budget and current software. A reasonable guess is that the form asks for too much too early. It's still a guess, because the recordings show where people stop and not why. A redesign that restyles the form and keeps the same eleven required fields could make it easier to use, but it leaves the suspected burden of information in place. If the form is the cause, the team can try a shorter one without redesigning the site.
A third story reads the step before the form. Of the 1,000 visits that reach the demo page, 800 never start the form. Suppose a survey on that page gets replies from visitors who say they want to see pricing or try the product before talking to sales. That points at the offer, because a sales demo is the only next step the page gives them. It's a hypothesis too, since the replies come from the visitors who answered and not from every visit that left. A redesign that keeps a sales demo as the only next step leaves that obstacle in place.
The last story is about the message. Suppose interviews with recent buyers show that most of them run practices with several locations, and that several asked early whether the software handles more than one. Suppose the homepage never says. A visitor with that question has little reason to go further, which would be one explanation for why four of five visits never reach the demo page. It's a hypothesis. The interviews show what those buyers wanted to know, and they don't show that the missing answer is what ends those visits. They also say nothing yet about the page structure.
If the answer can be added to the homepage and the product pages as they stand, a focused edit is the first thing to try. A redesign becomes the fitting tool when the current site can't reasonably be changed to carry that answer, for example because it needs sections organized by practice size. That's the situation we describe in when it's time to invest in a website redesign.
The numbers show where people leave. They don't show why, and the table alone can't choose between those stories.
In an article on A/B testing and UX research, Jen Cardello of Nielsen Norman Group separates a concept from the particular design that carries it out. A test can only compare the versions a team built. Research into the cause tells the team which problem those versions should solve.
That research draws on more than one kind of evidence. CXL's guide to conversion research describes technical checks, analytics, recordings of visitor behavior, surveys, interviews and task testing. Its purpose is to produce hypotheses a team can test.
Lean Labs' growth-driven design program shows what that looks like inside an agency's process. To find problems, the program draws on page behavior, form results, CRM lifecycle data and feedback from sales. Sales feedback earns its place because reps hear objections that analytics doesn't record. In the dental example, sales can add what buyers ask on calls, such as questions about multiple locations.
Once the team has a likely cause, it can write it down as something to test. We cover that step in turning user insights into hypotheses and experiments, and the wider relationship in the difference between growth-driven design and conversion rate optimization.
Suppose the dental software company goes ahead with the redesign, and the rate moves from 1% to 1.4%. That comparison alone doesn't show the redesign caused the gain. A new campaign could have started the same month. The template post could have lost search visits, which would shrink the visit count and lift the rate while buyers behaved exactly as before.
In an article on UX benchmarking, Kate Moran of Nielsen Norman Group describes comparing a design's measures against a reference such as an earlier version, which is what a recorded baseline provides. That comparison shows that something changed. It doesn't isolate the reason, so record visits by source next to the rate.
Where a page gets enough traffic, a split test gives a cleaner comparison. HubSpot's page A/B test, for example, shows two versions of a page at the same URL and sends half the visitors to each. It's available on Content Hub Professional and Enterprise. That particular test compares versions of one page, so here it would measure a page's change and not the whole redesign. Jakob Nielsen's 2005 article on A/B testing adds two limits. The test measures what visitors did under each design without explaining their reasons, and an immediate action can leave out later effects on the business.
The later effect that matters most here is lead quality. HubSpot's lifecycle stages separate a marketing qualified lead, which marketing has judged ready for sales, from a sales qualified lead, which sales has accepted as a potential customer. A submitted form is neither one automatically. Suppose the new demo form drops the budget question and requests rise. Sales could now be talking to more practices that can't buy. So decide before launch which of the metrics you track the redesign is meant to move: demo requests, leads that sales accepts or new customers.