9 A/B tests worth running before your next website redesign
A redesign changes everything at once and tells you little. Nine A/B tests to run first, so the new site is built on what you learned rather than on opinions.
Redesigns have a habit of arriving with a lot of confidence and very little evidence. A new agency, a new brand guide, a stack of mood boards, and twelve months later a site that looks better and converts about the same. Or worse, and nobody can say which of the forty things that changed caused it.
The cheaper move is to learn first. Run a handful of focused tests on the site you already have, find out what your visitors respond to, and hand those findings to whoever designs the new one. Here are nine we’d run, roughly in order of how much they tend to teach you.
A word on traffic before we start. If your pages get a few hundred visits a month, you won’t reach a meaningful result on small changes. Test bigger differences, run them longer, or test on your highest-traffic templates only. A test that can’t reach a conclusion is just a slower opinion.
1. The primary call to action on your busiest template
Pick the template with the most traffic, whether that’s a product page, a listing or a service page, and test the main call to action. Not the button colour; the offer. “Contact us” against “Book a viewing”. “Request a quote” against “See prices for your area”. “Enquire” against “Check availability”.
This tells you what visitors actually want to do next, which is the single most useful input a designer can have. It’s also a cheap test to set up.
2. Form length
Take your main enquiry form and test a short version against the current one. If you’re asking for nine fields, try four. Then look not just at submissions but at what happens next: do the shorter-form leads still turn into appointments, or does sales spend twice as long qualifying them?
The answer varies by business, which is exactly why it’s worth testing rather than assuming. Some teams find the extra fields were doing useful filtering. Others find they were just costing them leads.
3. A form against a conversation
A step further than form length: test a static form against a conversational one that asks one question at a time. Conversational formats can feel lighter, and they let you branch, so a seller and a buyer don’t see the same questions. They can also feel slower to some visitors.
You’ll learn something about how your audience likes to give information, and that should shape the new site’s whole approach to capture.
4. Where the contact options sit
Test the placement of your contact options: in the header, sticky on mobile, inline after the key content, or all three. On mobile especially, a contact button that’s three scrolls away is a contact button most people never see.
This is a good one to run separately on mobile and desktop. The answers often differ, and a redesign that applies one layout to both will miss it.
5. Phone, form or chat as the lead option
Many sites offer all three equally and let the visitor choose. Try making one of them the clear primary option on a key page. You might find your audience strongly prefers to call, or that they’d rather message after hours.
On the subject of after hours: if your enquiries peak in the evening, as they do for plenty of consumer businesses, test an option that works when the office is closed. A phone number is no use at 21:40.
6. Social proof near the decision point
Most sites have reviews or testimonials somewhere. Test moving them closer to the moment of decision: next to the form, beside the booking button, just above the price. Then test which kind works: a star rating, a short quote, a number of completed sales.
Be careful to use only genuine reviews you have permission to show. It sounds obvious, but redesigns are where placeholder testimonials quietly become permanent.
7. Price visibility
If you sell something with variable pricing, test showing a starting price or a range against showing nothing. A lot of businesses hide price to force a conversation. Sometimes that works; sometimes it just means the people who’d have happily paid go somewhere that tells them the number.
This one can be uncomfortable for sales teams, so agree in advance what you’ll measure. Leads alone isn’t enough; look at qualified leads and closed deals.
8. The search experience
If your site is a catalogue (cars, properties, products), search is probably the most used feature on it. Test the default sort order, the filters shown first, or a free-text search against a filter-only interface. Someone typing “family car under 20k with a big boot” is telling you exactly what they want, and a filter grid can’t hear it.
We think search deserves more attention than it gets; there’s more on that in site search is a conversion tool. For catalogues like cars or properties, an AI search such as Serviceform’s AgenticSearch is one way to test natural-language search without rebuilding the whole site first.
9. A page-aware prompt against a generic one
Finally, if you run chat or any kind of on-page prompt, test a generic greeting against one that knows the page. “Can we help?” against “Want to book a viewing of this flat?” This tells you how much context matters to your visitors, which in turn tells the redesign how tightly to tie engagement tools to page templates. We wrote about the idea in more depth in page-aware conversations.
If you want a sense of what that looks like before you test it, the Uprealer demo estate agency shows a chat that changes with the page you’re on.
How to run these without losing a year
A few practical rules:
- One test per page at a time. Overlapping tests on the same template make both results unreliable.
- Decide the metric first. Write down what “better” means before the test starts, ideally something downstream like appointments booked rather than clicks.
- Run full weeks. Weekday and weekend behaviour often differs. Stopping on a Wednesday because the numbers look good is how false winners are made.
- Write down the losers. A test that changes nothing is still a finding. It tells the designer that particular thing doesn’t matter much to your visitors, which frees them to choose on other grounds.
Expect most tests to come back inconclusive or small. That’s normal. The value is in the two or three that clearly move something, and in the arguments you no longer need to have.
Before the agency arrives
A redesign brief built on nine tests is a very different document from one built on a competitor review and a brand workshop. It says: this is what our visitors want to do, this is how they prefer to contact us, this is where they get stuck. Hand that to a good designer and they’ll build something that looks new and also works. Hand them a mood board and you’ll get a new mood board, nicely coded.