LiftCRO

Sample work

Read the code before you send us a brief

We are a new team, so instead of client logos here are three builds you can read line by line, with the QA report each one shipped with. All three run on our own demo store; client work is never shown without permission.

Read a sample QA report
  • Plain JavaScript and CSS, no libraries
  • Every sample backs out if its element is missing
  • 14 QA checks per build
A/B test · product page

Sticky buy bar on mobile

Hypothesis. On phones the buy button scrolls out of view within a second. Keeping it on screen will lift add-to-cart rate.

Approach. The variation clones the existing buy button into a fixed bar at the bottom of the screen, so the click handler and tracking the store already has keep working. It only runs under 768px, waits for the button to exist, and does nothing if it is missing.

  • No libraries: 20 lines of plain JavaScript and a 6-line stylesheet.
  • Backs out cleanly if the buy button never appears, so the control is never broken.
  • Reuses the store's click handler instead of re-implementing checkout or tracking.
variation-b.jsSample
// Variation B: sticky buy bar on mobile
if (window.innerWidth > 768) return;

waitFor('#buy', function (buy) {
  if (document.querySelector('.lc-sticky-buy')) return;

  var bar = document.createElement('div');
  bar.className = 'lc-sticky-buy';
  bar.appendChild(buy.cloneNode(true));
  document.body.appendChild(bar);

  // Hide the bar while the real button is on screen.
  new IntersectionObserver(function (entries) {
    bar.hidden = entries[0].isIntersecting;
  }).observe(buy);

  // The clone keeps the store's own click handler via event delegation.
  bar.addEventListener('click', function () { buy.click(); });
});
QA checkDesktopMobile
Chrome, Firefox, Edge (desktop)Not shown, by design—
Safari (macOS)Not shown, by design—
Safari (iPhone 13, 15)—Pass
Chrome (Pixel 7, Galaxy S23)—Pass
Flicker / layout shift on slow 3GNoneNone
Survives SPA navigationPassPass
Goals: purchase fires oncePassPass
Console errors00
A/B test · homepage

Benefit-led hero headline

Hypothesis. The hero describes the product; a headline that states the outcome will raise clicks on the primary button.

Approach. Text swaps are the most common test and the most common source of flicker. The variation changes the headline and button copy as soon as the elements exist, before first paint on most connections, and sends a click event so the result can be read in the testing tool.

  • Each change waits for its own element, so a slow-loading hero does not block the button change.
  • `once: true` stops the goal double-counting on repeated clicks.
  • The CSS is a separate file, applied by the testing tool before the page renders.
variation-b.jsSample
// Variation B: lead with the benefit
waitFor('.hero h1', function (h1) {
  h1.textContent = 'Ship faster. Sleep better.';
});

waitFor('.hero #checkout', function (btn) {
  btn.textContent = 'Start free trial';
  btn.addEventListener('click', function () {
    track('hero_cta_click');
  }, { once: true });
});

// variation-b.css
// .hero h1 { letter-spacing: -0.03em; }
QA checkDesktopMobile
Chrome, Firefox, Edge (desktop)Pass—
Safari (macOS)Pass—
Safari (iPhone 13, 15)—Pass
Chrome (Pixel 7, Galaxy S23)—Pass
Flicker / layout shift on slow 3GNoneNone
Targeting: homepage onlyPassPass
Goals: hero_cta_click fires oncePassPass
Console errors00
A/B test · cart

Free-shipping progress banner

Hypothesis. Visitors below the free-shipping threshold will add more when the banner shows how far they are from it.

Approach. The store already prints the cart total, so the variation reads it from the page, works out the gap to the threshold and shows a banner with the remaining amount. It re-runs on cart changes and hides itself once the threshold is met, so it never shows a wrong number.

  • Reads the number the store shows the visitor, so the banner can never disagree with the cart.
  • A MutationObserver keeps it right after quantity changes without any knowledge of the store's JavaScript.
  • Hides itself at the threshold instead of showing "$0 away".
variation-b.jsSample
// Variation B: free-shipping progress
var THRESHOLD = 50;

function update() {
  var el = document.querySelector('.cart-total');
  var banner = document.querySelector('.cart-banner');
  if (!el || !banner) return;

  var total = parseFloat((el.textContent.match(/[\d.]+/) || [0])[0]);
  var gap = THRESHOLD - total;
  banner.hidden = !(gap > 0);
  if (gap > 0) {
    banner.textContent = "You're $" + gap.toFixed(0) + " away from free shipping";
    banner.style.display = 'block';
  }
}

waitFor('.cart-total', update);
new MutationObserver(update).observe(document.body, { subtree: true, childList: true, characterData: true });
QA checkDesktopMobile
Chrome, Firefox, Edge (desktop)Pass—
Safari (macOS)Pass—
Safari (iPhone 13, 15)—Pass
Chrome (Pixel 7, Galaxy S23)—Pass
Updates after quantity changePassPass
Hidden at or above thresholdPassPass
Control untouched (no style leak)PassPass
Console errors00

The QA checklist

14 checks before any test is handed over

A second person, not the developer who built it, runs every check on the browsers and devices agreed in the contract. The list below is the one we use.

Browsers

  • Chrome, Firefox and Edge on desktop
  • Safari on macOS
  • Safari on iPhone
  • Chrome on Android

Behaviour

  • No flicker or layout shift, tested on throttled 3G
  • Survives client-side navigation and back/forward
  • Backs out cleanly when an element is missing
  • Control untouched: no styles leak from the variation

Targeting and goals

  • Runs only on the agreed pages
  • Audience and device conditions hold
  • Every goal fires once, in every variation

Quality

  • Zero console errors
  • Keyboard focus and contrast kept where the UI changed
  • Screenshots of each variation attached to the report

Pilot test

Start with one test

One A/B test, built and QA'd in your tool within 5 business days. Free for agencies. $500 for brands, credited against your first month. If the code is not good enough to launch, you owe nothing.

Request a pilot test