A/B testing your landing page: what to test first

Why button color shouldn't be your first test

The classic advice in A/B testing articles is to change your button from blue to orange. In practice this gives a 0.1-0.3% conversion lift, and not always. For that difference to become statistically significant, you need thousands of visitors and weeks of testing.

Button color affects conversion only when the original is genuinely broken - a gray button on a gray background or text that disappears into the layout. Otherwise it's a detail that won't outweigh a weak offer or unclear headline.

If you get 200 visits a day, a button color test will drag on for a month without a clear answer. In that time you could validate three hypotheses that actually move the numbers.

Headline and offer - what kills or saves conversion first

A visitor decides to stay or leave in three to five seconds. During that window they see the headline, subheadline, and above-the-fold content. If the offer is unclear or doesn't hook them, nothing else matters.

What to test in the headline: specific against abstract. Bad version: "We'll help your business." Good version: "We'll set up ads so each lead costs under $7." The second gives them something concrete - they immediately understand what they get and what it costs.

Test the offer by shifting the emphasis: discount versus bonus, timeline versus price, gain versus fear of loss. For example, "20% off" versus "save $200." The second often works better because the brain registers a concrete amount more sharply than a percentage.

A headline test will show results fastest because it affects every visitor. A 20-40% conversion increase at this stage is normal if the original was generic.

Lead form: length versus barrier to entry

The form is the bottleneck in your funnel. Someone is already interested but can drop off at the fill-out stage. Two parameters drive conversion here: number of fields and what you ask for.

Test field count. Three fields (name, phone, comment) versus one (phone only). A short form almost always generates more leads, but lead quality can drop. If you sell an expensive service and need warm prospects, a long form filters out tire-kickers. For mass-market products, go short.

Second point - required fields. If you don't actually need email to do the work, make it optional and see how conversion changes. Often one extra required field kills 10-15% of submissions.

Button copy also works. "Submit" versus "Get a quote" or "See pricing." A button with a concrete action is usually more clickable because people understand what happens next.

Social proof: when it works and when it doesn't

Reviews, client logos, project counts - all social proof. They reduce distrust and help someone decide. But they work differently depending on your niche.

If you sell a service where risk is high (development, construction, healthcare), a testimonial block is critical. Test its presence: version with reviews versus version without. This often yields a 15-25% conversion gain.

What to test in testimonials: client photo versus placeholder avatar, short text versus long, concrete number in the review ("saved $1,100") versus generic praise ("everything was great"). Specifics and a human face work harder.

Client logos make sense only in B2B or if you have recognizable brands among your customers. For local business or mass B2C, it's wasted space.

Price: show it upfront or hide it

This is a painful question. If the price is high, there's temptation to hide it and push people into a conversation. If it's low, show it immediately to filter out freebie hunters.

Test two approaches. First variant: price on the first screen, large. Second: instead of price, copy like "We'll calculate cost in 15 minutes" and a button to a calculator or form. Which works better depends on the product.

If the service is standardized and the price is competitive, show it upfront. This filters non-target visitors and saves you time processing empty leads. If the service is complex and price is set individually, better to lead into a conversation.

In practice the version with price often wins, but you get fewer leads - warmer ones though. Test it and look not just at lead volume but at how many closed into deals.

Above the fold: one action or several

Above the fold is what someone sees before scrolling. Classic mistake - cram everything in: headline, three buttons, a form, benefits list, and a phone number.

Test focus. Variant A: one headline, one offer, one button. Variant B: headline, benefits list, two buttons ("Order" and "Learn more"). Usually the single-call-to-action variant converts better because it doesn't force a choice.

If you do need two buttons, make one primary (contrasting color, larger) and one secondary (outline, smaller). This is called visual hierarchy, and it directs attention.

Another point: the image above the fold. Test product photo versus result photo. For an ad setup service, for example: photo of a manager at a computer versus screenshot of a lead growth chart. The result hooks harder.

How to run an A/B test and get a clear result

Test one element at a time. If you change the headline, button, and form simultaneously, you won't know what worked. That's called multivariate testing, and it requires massive traffic.

For a test to be statistically significant, you need at least 100-150 conversions per variant. If you have 500 visits a week and 2% conversion, that's 10 leads. The test will take three to four weeks. Factor this in when planning hypotheses.

Don't stop a test early. If one variant is leading after three days, that means nothing yet. Wait until you have enough sample size, or you risk mistaking noise for signal.

A/B testing tools: Google Optimize (free, shut down in 2023, but many use alternatives like VWO or Optimizely), Yandex Metrica (has built-in experiments, though feature set is limited). If the landing page is on Tilda, they have A/B testing in paid plans. We usually set up tests through code when we need flexibility, or use Metrica for simple hypotheses.

When A/B testing doesn't make sense

If your landing page gets 50 visitors a week, testing is pointless. You'd wait six months for a result, and it still wouldn't be reliable. In this case, spend effort increasing traffic or get an expert review of the page and fix obvious problems.

Second case: if landing page conversion is below 1%, the problem isn't in the details. Most likely the offer is broken, you have the wrong audience, or the traffic is off-target. A/B tests won't help here - you need to rebuild the concept.

Testing makes sense when you already have baseline conversion (at least 2-3%) and want to squeeze more from it. It's an optimization tool, not a resuscitation kit.

What to do after the test

Roll out the winning variant permanently, delete the loser. Document the result: which element you tested, what gain you got, how long the test took. That's the foundation for your next hypotheses.

Don't obsess over one landing page. If you've squeezed everything you can from it, shift to other growth levers: ad targeting, lead response time, sales team performance. The landing page is one element of the funnel, not the whole funnel.

If you don't have time or expertise to set up tests yourself, we can handle it end-to-end: from hypotheses and setup to analysis. Cost depends on the number of hypotheses and page complexity - we discuss it after a brief. Our contact info and work examples are at efimovdev.ru.