Page speed tends to be owned by developers and measured in milliseconds, which is why it rarely gets prioritised. Framed as a conversion problem instead, it competes for attention on equal terms with campaigns and creative.
What the three metrics actually measure
- Largest Contentful Paint: how long until the main content appears. This is the one visitors experience as the page being slow.
- Interaction to Next Paint: how quickly the page responds when someone taps or clicks. Poor scores feel like a broken page rather than a slow one.
- Cumulative Layout Shift: how much the page moves while loading. This is what causes the mis-tap that sends someone to the wrong screen.
Why it affects revenue
Every additional second before content appears gives a visitor another opportunity to leave, and the visitors most likely to leave are the ones arriving on mobile connections from paid campaigns you have already paid for. Slow pages do not only lose conversions; they raise the effective cost of every channel that sends traffic to them.
Where the time usually goes
- Uncompressed hero images, or images served far larger than the space they occupy.
- Render-blocking CSS and JavaScript loaded before anything can paint.
- Third-party scripts for analytics, chat, testing and advertising, each with its own connection cost.
- Fonts loaded late, causing text to appear, disappear and reappear.
- Layout that reserves no space for images and embeds, so everything jumps when they arrive.
A sensible order of work
- Measure real users, not just lab tests. Field data reflects the devices and networks your visitors actually have.
- Fix the largest contentful element first, which is usually one image or one stylesheet.
- Audit third-party scripts and remove anything nobody is using.
- Set explicit dimensions on images and reserve space for anything injected after load.
- Only then start tuning code.
Set a target you will defend
Speed work decays. A tag gets added, an image gets uploaded at full size, a plugin arrives. Agreeing a budget in advance, and treating a regression as a bug rather than a nice-to-have, is what keeps the gains.