Landing pages that convert: structure, proof and the offer
Landing page testing has a reputation problem because most of it tests trivia. The variables that move conversion rate by a meaningful margin are structural, and you can usually fix them without a testing tool.
Message match
The headline should repeat the promise of the ad that was clicked, in close to the same words. A visitor spends a second or two confirming they are in the right place; any friction in that moment costs you the session. One landing page per offer, not one page for a campaign.
What has to be above the fold
- 1.What this is, in plain language, without a metaphor
- 2.Who it is for, stated explicitly enough to exclude people
- 3.One primary action, repeated later but never competing with a second action
- 4.One piece of immediate proof — a named client, a number, a recognisable logo
Proof next to the claim
A wall of logos in the footer does almost nothing. The same logos placed directly beside the claim they support do a great deal. Pair each substantive claim with the evidence for it, and never invent the evidence — a fabricated testimonial is a legal problem, not a creative choice.
Form length is a pricing decision
Every field is a price the visitor pays. Short forms produce more leads of lower average quality; longer forms produce fewer, better-qualified leads. Which is right depends on whether your bottleneck is sales capacity or pipeline volume. Decide that first, then set the field count deliberately.
Pricing transparency
If you have a floor, state it. Publishing a starting figure filters out enquiries that were never going to close and raises the quality of the ones that arrive. We publish a $5,000 engagement minimum for exactly this reason: it makes the first call a conversation about fit instead of budget.
The order to test
- Offer and headline — the largest available swing
- Proof type and placement
- Form length and field ordering
- Page structure and section sequence
- Everything else, once the above are settled
Run one variable at a time, and let the test reach the sample size you calculated before you started. Stopping early because the numbers look good is the most common way teams learn something untrue.