19 landing page A/B test ideas ranked by expected impact

shipped test #4

Landing page test lists are usually written as a grab bag: button colors next to headline rewrites next to layout overhauls, as if they had equal odds. They do not. The expected impact of a test is roughly how much it changes what a visitor understands or believes, multiplied by how many visitors see the element. That formula puts comprehension changes at the top, trust changes in the middle, and cosmetics at the bottom.

Here are nineteen ideas in five ranked groups, with the reasoning attached — because the reasoning, not the ranking, is the reusable part.

Tier 1: what the page claims (highest expected impact)

  1. Headline: outcome vs description. "Ship experiments weekly" versus "An experimentation platform". The single most-read line on your site; clarity changes here move everything downstream. Full treatment in the headline testing guide.
  2. Headline specificity. Add the concrete noun: who it is for, what it does, in what stack. Specific claims are more believable and better matched to search intent.
  3. Subheadline: how it works in one line. Test explaining the mechanism ("scans your repo, opens PRs") against expanding on the benefit. Skeptical audiences often need mechanism before they accept benefit.
  4. Above-the-fold answer test. Can a stranger answer "what is this and why would I use it" in five seconds? Test a version rewritten solely for that, against your current hero.
  5. Price transparency on the page. Showing a starting price versus hiding it behind a pricing link. Changes lead volume and lead quality in opposite directions; decide which you want before running it.

Tier 2: what the visitor is asked to do

  1. CTA promise. "Start free trial" vs "See your first experiment" vs "Get a demo" — each sets a different expectation of effort and commitment. Verbs and friction words are covered in CTA copy experiments.
  2. Anxiety reducer under the button. "No card required. Free tier forever." versus nothing. Cheap, frequently effective, only works if true.
  3. One CTA vs two. A single primary action against primary-plus-secondary ("or watch a 2-min demo"). The secondary can catch not-ready visitors or dilute the ready ones.
  4. Form on page vs behind click. Embedding the signup form on the landing page removes a step but raises the page's ask. Direction of the effect varies by traffic temperature.
  5. Form length. Every field is a toll; audit and cut per the method in form length and conversion.
Claims (headline, value)90Ask (CTA, form)70Trust (proof)55Structure (length, layout)40Cosmetics15
Illustrative shape only: expected impact by tier, not measured data.

Tier 3: why they should believe you

  1. Proof above the fold. Move your strongest single piece of evidence — a sharp quote, a concrete usage fact — into the hero, versus keeping proof lower. Placement of proof usually beats quantity; more in social proof experiments.
  2. Specific quote vs logo wall. One named, concrete testimonial against a strip of logos. Specificity tends to carry more belief per pixel.
  3. Show the product. A real screenshot or short product clip in the hero versus abstract illustration. Buyers discount pages that will not show the product.
  4. Objection section. A short "what about X" block addressing your most common sales objection, versus omitting it.

Tier 4: structure and length

  1. Long page vs short page. More sections help unaware visitors and slow down ready ones. The right length tracks traffic temperature: cold ads usually need more page than word-of-mouth traffic.
  2. Section order. Proof-then-features versus features-then-proof. Test the order your sales conversations actually follow.
  3. Comparison section. An honest "how we differ from X" block, versus none — high value for visitors arriving from comparison searches.

Tier 5: cosmetics (test last, if ever)

  1. Button color and size. Occasionally matters when contrast was genuinely poor; usually noise. Rank it honestly: last.
  2. Imagery style. Photo versus illustration versus abstract. Weak lever unless current imagery actively confuses what the product is.

How to re-rank this list for your funnel

This ranking assumes a typical SaaS landing page with mixed-temperature traffic. Yours may differ, and the evidence to re-rank is already in your analytics. If visitors bounce in seconds, comprehension (tier 1) is the problem. If they scroll deep and leave without clicking, the ask or the belief is failing (tiers 2–3). If CTA clicks are healthy but signups are not, the form or the next page is the leak, not the landing page. Match the tier to the observed failure, and test the biggest plausible change there first — low-traffic sites especially cannot afford to spend weeks detecting a 2% cosmetic effect.

The bottleneck for most teams is not ideas — it is that every idea needs variant code, tracking, and cleanup. Trevo closes that gap: it scans your repo and your funnel, proposes the specific tests from this list that your evidence supports, and writes each variant as a pull request you review. Point it at your landing page with a free repo scan and see which of the nineteen it would run first.

Frequently asked questions

Which landing page test should I run first?

Almost always the headline, unless your analytics point elsewhere. It is the most-read element and controls whether visitors understand the product at all, so plausible effects are large — which also makes it the fastest test to read at low traffic. If visitors already scroll and engage but do not convert, start with the CTA or the form instead.

Do button color tests ever work?

Occasionally, when the original button had genuinely poor contrast or was hard to find. As a general lever it ranks near the bottom: it changes nothing about what visitors understand or believe. If you have limited traffic, spending it on comprehension and proof tests has far better expected return than cosmetics.

Should a landing page be long or short?

It depends on traffic temperature. Visitors from cold ads need more convincing — more sections, more proof, more objection handling — while referred or returning visitors mostly need a clear claim and a fast path to the CTA. Test length against your dominant traffic source rather than following a universal rule.

How many landing page tests can I run at once?

On the same page, run one at a time unless you have serious traffic and tooling that handles interactions. Two simultaneous tests on one page split your sample across four combinations and can interfere with each other's reads. Across different pages, parallel tests are fine.