There's no universal threshold, but there are public benchmarks and a measurement you can run on your own site. Google considers a page good when its main content loads in 2.5 seconds or less, and poor beyond 4 seconds.
The benchmark
Google's thresholds, measured on real visitors, for three out of four page loads (the 75th percentile), over the last 28 days:
| Metric | Good | Poor |
|---|---|---|
| Main content display (LCP) | 2.5 s or less | more than 4 s |
| Responsiveness to interactions (INP) | 200 ms or less | more than 500 ms |
| Visual stability (CLS) | 0.1 or less | more than 0.25 |
Your own threshold comes from comparing the conversion rate of fast visits and slow visits, on the same pages.
At a men's fashion brand the authors work with, the homepage triggered 417 requests and the product page 285, including 119 third-party code requests ("scripts"): the email tool loaded twice, the session recording tool twice, a second tool of the same kind on top of that, and two measurement properties.
The authors add duplicates like this too, such as an audit tool installed for three weeks and never removed. On that site, the target is fewer than 150 requests and fewer than 40 third-party scripts on the homepage.
A 2020 Deloitte study for Google linked a tenth of a second gained on mobile to 8.4% more conversions in retail, across 37 brands. It's a correlation, but it gives you the order of magnitude.
The rule
First fix what's "poor" for your real visitors, on mobile, on your product pages and collection pages.
The perfect score can wait. An SEO consultant told a shared client in April: "The goal is to reach 65 out of 100. That's the minimum. Developers will spend a day to gain two or three points, and that's pointless."
He added a cause people forget: "If you have an A/B testing tool, of course it takes time."
The steps
- Field data. Pull it for your three main templates.
- Third-party scripts. List the ones loaded on the product page, and count the duplicates.
- Apps. Remove the ones you no longer use, then their code: Shopify warns that some apps leave code in the theme after being uninstalled.
- Images. Lighten the ones above the fold, and reserve their space: a price or button that shifts just as a thumb lands causes missed taps, and it can be fixed with one line of code in the theme.
- Remeasuring. Wait 28 days.
A sentence to say to your leadership, to adapt with your own templates:
"We're not aiming for 100. We're aiming for green on phones, on the product page and the collection page, in field data, and we're removing two third-party tools a month."
The trap
Trusting the lab score is the first trap. It simulates a phone and a network that aren't your customers'. Field data is what counts. The opposite case exists too: a site that's green everywhere but converts poorly.
At an accessories brand, all three metrics were good: 1.5 seconds, 128 milliseconds, 0.02, and engagement was capped at 52%. Speed was ruled out as the cause, and the work lay elsewhere, in the content.
Finally, on Shopify, look for the culprit in the stack of apps before suspecting the theme.
