High-conversion Landing Page

A landing page is not a homepage. It does one thing: get a visitor to complete one action. So the first screen has to say three things:

1. Who it's for: "for indie developers who don't want to maintain a backend" — not for everyone
2. What they get: "a static site that can take money, up in two hours" — write the outcome, not the feature list
3. Why believe you: "21 scenarios and 100+ steps already running", "real scripts included, not slides" — evidence, not slogans

First-screen structure (copy it):
one-line claim → subhead → primary CTA → proof strip (numbers / screenshot / one-line endorsement)

The counter-example: "committed to becoming an industry-leading solutions provider". Anyone could write that, which is why it says nothing.

CTA copy is action plus benefit: "Get the scripts free →" outperforms "Submit" by a wide margin.

Every screen answers the next question in the visitor's head:

1. What is this? (first screen)
2. What does it have to do with me? (the pain: where you're stuck now)
3. How does it work? (the solution, in 3–5 steps, with screenshots)
4. Has anyone used it? (proof: numbers, screenshots, a one-line quote)
5. Is this going to screw me over? (objections: price, do I need paperwork, how long it takes, can I get out)
6. So what do I do next? (CTA, repeated)

Ordering rule: the closer an objection sits to the decision, the later it goes — but it must be resolved before the final CTA.

What to cut: team bios, company history, news, a nav bar full of unrelated links. On a landing page the nav should be logo plus CTA, nothing else.

Mobile first: most traffic is on a phone. Get the claim and the CTA onto one phone screen before you touch the desktop layout.

The one thing you can't skip on a landing page is tracking. Three things must be recorded separately, or the data comes out as mush:

| Event | Trigger | props |
|---|---|---|
| lp_view | page visible | src (utm_source) / variant (A/B version) |
| lp_scroll | scrolled to 25 / 50 / 75 / 100% | depth |
| cta_click | any CTA clicked | pos (top / middle / bottom) / src |
| cta_submit | form submitted successfully | src |

Why source has to be recorded: traffic from a chat app and traffic from search can convert three times differently. Mix them and you'll wrongly conclude "the page is fine". Put ?src=wx-style parameters in the URL and write them into props.

Record the definitions in data/conversion.json, so before-and-after changes are auditable.

The smallest closed loop: lp_view → cta_click → cta_submit. Connect those three and you have conversion observation at all.

Don't change five things at once — if the number goes up, you won't know who to thank.

The least-work A/B, which a static site can do:

1. Copy the page to lp-b.html and change exactly one variable — only the main headline, say
2. Split traffic by ?variant=b or by source: half the links point at A, half at B
3. Record lp_view / cta_click / cta_submit separately per version
4. Compare only after 100+ visits per version. Below that it's noise

Three things worth testing, in order of impact: the main headline, the CTA copy, the proof strip.

How to judge: look at two ratios — click rate = cta_click / lp_view, and submit rate = cta_submit / lp_view. Not the absolute counts: when traffic rises, clicks rise too, and that was the traffic's doing, not the page's.

Keep a record: write the conclusion into the notes in data/conversion.json, so nobody re-tests the same variable next quarter.

Speed is conversion: past three seconds of mobile load time, bounce rate doubles. At minimum:

• Compress every image and use WebP or SVG; the first-screen image gets loading="eager", the rest lazy
• Inline critical CSS; JS at the bottom or defer
• Put it behind a CDN (see the CDN + HTTPS Acceleration scenario) with gzip or br on
• Run PageSpeed Insights once and get LCP under 2.5 s first

Launch checklist:

• HTTPS with forced redirect
• canonical pointing at the main version — the B variant canonicals to A, so it isn't treated as a duplicate page
• Submit the sitemap; add the landing page to data/index-observation.json
• Run bash scripts/seo-check.sh

Weekly review (15 minutes):
Look at lp_view → cta_click → cta_submit and fix whichever stage drops hardest. Losing view→click means fix the first screen; losing click→submit means fix the form.

Compliance, easy to forget: promises have to be keepable, state the conditions on any pricing or earnings claim, and put a disclaimer in the footer.