Site speed is a conversion input, not a compliance task. The clearest published example: QuintoAndar reduced Interaction to Next Paint by 80% and saw conversions rise 36% year on year. redBus improved the same metric and increased sales by 7%.
Those are other people’s businesses and other people’s audiences, so do not adopt the percentages. Adopt the direction, then work out what it is worth on your own numbers — which takes about five minutes and is the last section of this article.
The three metrics, and what they actually measure
Google’s Core Web Vitals are three separate complaints a visitor might have, each with a published threshold measured at the 75th percentile of page loads.
| Metric | The complaint | Good |
|---|---|---|
| LCP Largest Contentful Paint |
“It took ages to show me anything.” | Under 2.5s |
| INP Interaction to Next Paint |
“I tapped it and nothing happened.” | Under 200ms |
| CLS Cumulative Layout Shift |
“It moved and I tapped the wrong thing.” | Under 0.1 |
The 75th percentile matters more than it sounds. It means three quarters of visits must be good, so an average pulled up by fast desktop users hides a mobile experience that fails.
INP replaced First Input Delay and is the harshest of the three, because it measures every interaction rather than the first one. Sites that passed comfortably before often do not now.
Why slow pages cost money twice
Directly: people leave. Every second between the tap and the content is a chance to reconsider, and on a paid click you have already spent the money.
Indirectly: a slow page lowers your conversion rate, which raises your cost per acquisition, which lowers your return on ad spend across every campaign at once. The ad account looks like it is underperforming. The ad account is fine.
What is usually wrong
In the sites we audit, the same handful of causes come up repeatedly.
Images nobody resized
A 3,000-pixel photograph displayed in a 600-pixel space, sent in full to a phone. Serve modern formats, size them to their container, and let the browser pick with srcset. This is usually the single largest LCP win and it needs no rebuild.
Too many plugins doing too little
Each one loads its own CSS and JavaScript on every page, whether that page uses it or not. A WordPress site with forty plugins is carrying most of that weight on the homepage. Audit what each one earns its place with.
Third-party scripts
Chat widgets, heatmaps, three analytics tools, a review badge, two pixels. Each blocks the main thread and each one was added by someone who reasonably thought it was small. Together they are the most common INP failure we see.
Fonts that hold up the text
A custom font loaded without font-display: swap leaves the page blank until it arrives. Preload the one face you need above the fold and let the rest swap in.
Layout that settles after it loads
Images without width and height, ads injected into the flow, a cookie banner pushing everything down. Reserve the space before the content arrives. CLS is the cheapest of the three to fix and the most irritating to experience.
Work out what it is worth to you
Take a site with 20,000 monthly sessions, a 2% conversion rate and a $60 average order — $24,000 a month.
Assume a speed fix lifts conversion from 2% to 2.2%. That is a 10% relative improvement, far below the published case studies.
- Orders: 400 → 440
- Revenue: $24,000 → $26,400
- Gain: $2,400 a month, $28,800 a year
Against a one-off fix, that is a straightforward decision. And unlike media spend, it does not recur — you pay once and keep the uplift on organic, paid, email and direct traffic alike. The ad budget planner shows the paid-side half of that: how much less budget the same revenue target needs once conversion improves.
How to measure it honestly
- Field data first. The Core Web Vitals report in Search Console uses real visits from real devices. That is the number Google uses.
- Lab tools for diagnosis only. PageSpeed Insights and Lighthouse tell you what to fix. The score is not the goal; a 100 that nobody experiences is decoration.
- Test on a mid-range phone on mobile data. Not your laptop on office wifi. That is where the failures live.
- Measure before and after, on the same pages. Speed work is easy to claim and easy to verify.
Frequently asked questions
Is speed a ranking factor?
It is part of Google’s page experience signals, and it is a tiebreaker rather than a lever — better content at a slower speed still wins. The conversion argument is far stronger than the ranking argument, which is why this article leads with it.
Will a caching plugin fix it?
Caching helps LCP on a slow server and does almost nothing for INP or CLS, because those are caused by your own scripts and layout. It is a reasonable first step and rarely the whole answer.
My score is 95 on desktop and 40 on mobile. Which counts?
Mobile, for almost every business. It is where most traffic is, it is the harder test, and it is what Google evaluates for most sites.
How much work is this usually?
Images, fonts and layout shift are typically days rather than weeks. Third-party scripts are a negotiation with whoever wanted them. A theme built badly from the start is a rebuild, and it is better to know that early.
Where to go from here
Pull your Core Web Vitals report in Search Console and look at the mobile column at the 75th percentile. If any of the three is failing, work out what a 10% conversion improvement is worth on your numbers and compare it with the cost of fixing. That calculation is usually short. Speed work sits inside our web design service, and it is one of the first things a free audit reports on.
[…] The published case studies are specific. QuintoAndar reduced INP by 80% and saw conversions rise 36% year on year. redBus improved INP and increased sales by 7%. We look at this in more detail in what page speed is worth in money. […]