A Faster Website Makes You More Money: How We Took a Landing Page From 75 to 99
You pay for every click. Whether that person ever sees your page is a separate question, and on a phone the answer depends mostly on how fast the page shows up.
Google studied this on mobile sites in 2016 and found that 53% of visits are abandoned when a page takes longer than three seconds to load. If you run ads, think about what that means. You paid for the click, the person tapped, and a good share of them left before your headline appeared. Your ad report still counts every one of them as a click.
A couple of weeks ago we had a landing page in that spot. It gets paid traffic, it has a sales video at the top, and it scored 75 out of 100 on Google's mobile speed test. After a few hours of work it scored 98 to 99, on all three versions of the page. Below is what we changed, in the order that moved the score.
Why speed is a money problem
- You pay per click, not per page view. Meta reports both numbers: link clicks and landing page views. The gap between them is people you paid for who never saw the page. A slow page makes that gap bigger.
- It's the cheapest lift you have. You don't need a new ad, a new offer or more budget. The same traffic sees the same page, and more of it stays long enough to read it.
- Google looks at it too. Page speed on mobile is one of the signals Google uses when it ranks pages, so a slow site also costs you free traffic.
I'm not going to tell you this doubled sales. Other things changed on that page in the same weeks, so I can't isolate what speed did on its own. What I can give you is the number to watch on your own account, and the steps.
Step 1: Find out where you're losing points
Go to pagespeed.web.dev, paste your URL and look at the mobile tab. The score out of 100 is built from five measurements, and they don't weigh the same:
- Total Blocking Time: 30 points
- Largest Contentful Paint: 25 points
- Cumulative Layout Shift: 25 points
- First Contentful Paint: 10 points
- Speed Index: 10 points
Our page was losing 21 of its 25 points on one line: Largest Contentful Paint, which is how long the biggest thing on the screen takes to finish loading. The long list of warnings further down the report, about caching and image dimensions, wasn't costing us anything. So find the measurement that's losing points and fix that one first. The warnings list will send you in ten directions.
Step 2: Load your biggest element first
On a page with a video at the top, the biggest element is the video's cover image. The browser was finding ours late, after the styles and the fonts. We made two changes.
- Tell the browser to fetch it first. One line in the head of the page:
<link rel="preload" as="image" href="cover.webp" fetchpriority="high">. Then remove the high-priority hints from images further down the page. If everything is marked urgent, nothing is. - Make the file smaller. We saved the cover as a WebP, 960 pixels wide, at quality 75. It went from 64 KB to 19 KB and looks the same on a phone.
Step 3: Don't load the video until someone presses play
A sales video is tens of megabytes, and by default most browsers start downloading part of it as soon as the page opens. Adding preload="none" to the video tag stops that. No video data moves until the person presses play.
This step is the one that bit us. With that setting, clicking the middle of the video on desktop Chrome did nothing. The click registered and the video stayed paused. We shipped it that way and didn't catch it for three days. The fix was our own play button on top of the video that starts it directly.
If you make this change, add the button, then test the live page by clicking it yourself on a real phone and a real computer.
Step 4: Host your own fonts
Fonts loaded from Google mean the browser has to connect to another server before your text can appear. We downloaded the font files and served them from our own domain, one file per font family covering the weights we use. We also kept only the character sets we need. The basic Latin sets cover English and Spanish.
Step 5: Make your tracking wait its turn
The Meta Pixel and the session-recording script are heavy, and they were loading before the page had finished. We didn't remove them. We kept the small piece of code that queues up events and moved the download of the heavy script to after the page finishes loading. Any event fired in the meantime waits in the queue and sends once the script arrives.
Check this one before you trust it. We opened the network log and confirmed the PageView event still went out. If you skip that check you can end up with a fast page and broken tracking, which is worse than where you started.
Step 6: Send phones phone-sized images
The images further down the page were one large file for every screen. We made three sizes of each one (400, 600 and 800 pixels wide) and let the browser pick with srcset. Generate the smaller sizes from the original file, not from a copy that's already compressed.
How to check your own page this week
- Run the test. Use pagespeed.web.dev on the mobile tab. Run it three times, because the score moves a few points between runs.
- Find the losing measurement. Of the five, which one is red or orange? Start there.
- If it's Largest Contentful Paint, find the element. The report names it. It's usually the hero image or the video cover.
- Compare clicks to page views. In Ads Manager, add the "landing page views" column next to "link clicks" for the last seven days. Write down the gap, then check it again a week after you make changes.
The page looks the same as it did before. It has the same video, the same copy and the same offer. The difference is that it now appears before the person gives up, and that's the part you were already paying for.
Want to know where your funnel is losing the people you paid for?
The Audit covers the whole path from ad to booked call: page speed, tracking, follow-up and the numbers in between. You get a written list of what to fix and in what order.
See the Packages →