SingRankYour Site Scored 16 on PageSpeed. Don't Rebuild Yet.
Contents7
A developer sends you a screenshot. Mobile score 16 out of 100. Largest Contentful Paint 22.5 seconds. The quote that follows is for a new website, and the number in the screenshot is doing most of the selling.
Before you sign, look for the other number on the same page. On 30 September 2026 we ran PageSpeed through our audit tool on nine Singapore service websites. All nine scored under 70 on mobile. Every one of the nine failed Google's 2.5-second threshold for Largest Contentful Paint in the lab test. Seven of the nine also had enough real Chrome traffic for Google to report what actual visitors experienced, and six of those seven were inside the threshold. On the same sites, at the same minute.
One number said the site was broken. The other said it was fine. Only one of them describes your customers.
The two numbers PageSpeed gives you, and which one is about your visitors
PageSpeed Insights reports a lab test and, when it can, field data. The lab test loads your page once on a throttled simulated phone in a data centre. The field data is the Chrome UX Report, which Google describes as "a dataset that reflects how real-world Chrome users experience popular destinations on the web", collected "from real browsers around the world".
The thresholds are the same for both. Google's Core Web Vitals guidance says Largest Contentful Paint "should occur within 2.5 seconds", Interaction to Next Paint should be "200 milliseconds or less", and Cumulative Layout Shift "0.1. or less" — and that these are assessed at "the 75th percentile of page loads, segmented across mobile and desktop devices".
That 75th percentile line is the one people skip. It means three in four of your real visits must clear the bar. It also means the field number already accounts for your visitors' actual phones, actual connections and actual location, which in Singapore is usually a good phone on a fast network — while the lab test deliberately simulates a slow one.
Our Core Web Vitals guide covers what each metric measures. This piece is only about the decision.
Nine Singapore sites, two numbers each
These are the nine sites our audit completed on 30 September 2026, sorted by mobile score. They are all Singapore veterinary clinic sites, drawn from the sample in our vet website study published the same day. No business is named here, and nothing in this table is a comment on any company.
| Site | PageSpeed mobile | Lab LCP | Real-user LCP | Real-user INP |
|---|---|---|---|---|
| 1 | 16 | 22.5 s | 5.72 s | 158 ms |
| 2 | 38 | 38.3 s | 1.71 s | — |
| 3 | 43 | 15.0 s | not enough data | — |
| 4 | 45 | 28.7 s | not enough data | — |
| 5 | 49 | 11.5 s | 1.87 s | 195 ms |
| 6 | 54 | 10.9 s | 1.10 s | 111 ms |
| 7 | 54 | 5.5 s | 2.26 s | — |
| 8 | 64 | 6.3 s | 1.07 s | — |
| 9 | 65 | 7.5 s | 1.02 s | 69 ms |
Read down the two LCP columns. The median lab figure across the nine is 11.5 seconds. The median real-user figure across the seven that have one is 1.71 seconds. Site by site, the lab number ran between 2.4 and 22.4 times the real-user number.
Four of the nine also reported real-user Interaction to Next Paint. All four were inside 200 milliseconds, one of them at 195, which is inside but not comfortably.
Site 1 is the exception that makes the rule useful. Its lab score was the worst in the set at 16, and its real-user LCP of 5.72 seconds is the only one that fails. Its lab Cumulative Layout Shift was 0.806, eight times the threshold. That is a site with a genuine problem, and the two measurements agree about it. Site 2, with a worse lab LCP of 38.3 seconds, has real visitors loading it in 1.71 seconds.
If you had used the mobile score alone to decide who needed a rebuild, you would have ranked site 2 as the second-worst case in the room. On the evidence of its own visitors, it is one of the healthiest.
The rule I would use before signing anything
Five steps, in order. It takes about ten minutes and needs no agency.
- Put your busiest page through PageSpeed Insights and scroll past the score. Look for the real-user section. If it is there, that is your answer and the lab score below it is a diagnostic aid, not a verdict.
- If real-user LCP is 2.5 seconds or under and INP is 200 milliseconds or under, your site is not slow for the people using it. A rebuild cannot sell you speed you already have. Spend the money on what the page says instead.
- If there is no real-user section, that is a traffic finding, not a speed finding. Google has not seen enough Chrome visits to your page to report on it. A new website will not create that traffic. Two of our nine sites are in this position.
- If real-user LCP is above 2.5 seconds, you have a real problem — now find whether it is the house or the furniture. Check the server response time first, then look at what the lab test names as the largest element. In our nine, one site answered our own plain request in 75 milliseconds and another took 3,024 milliseconds. A hero image, an embedded video or a chat widget is furniture, and furniture can be moved in an afternoon.
- Rebuild only for a reason that is not speed. You cannot edit your own pages. The platform cannot produce the page types you need. The thing is unusable on a phone. Those are real reasons. A score of 16 is not one, on its own.
What a rebuild costs that the quote does not mention
The invoice is the cheap part. In our view four other costs land on you, and none of them appears in a proposal.
Your URLs. Every address that changes needs a redirect, and the redirect map is the first thing a rushed rebuild gets wrong. Pages that ranked stop ranking, and nobody notices for six weeks because nobody was watching the old list.
Your measured history. A new site starts with no real-user data, so for the first month or two you cannot tell whether a change in enquiries came from the new design or from the season. You have traded a known baseline for a blank one.
Your attention. Three to six months of decisions about fonts is three to six months not spent on the pages that answer a buyer's actual question. For most of the businesses we look at, that is the larger loss.
The thing that was actually wrong. A rebuild moves the same missing answers onto a faster page. In the sample above, not one of the ten homepages showed a price, and five of ten did not state their opening hours. A rebuild that reproduces both misses has cost you money to change nothing a customer cares about.
Here is the part that costs us something to write. We sell search work, and a slow-site diagnosis is an easy thing for any agency to sell against. On the evidence of these nine sites, most of them need neither a rebuild nor a speed project. They need two paragraphs and a phone link. We have said elsewhere that there are three times SEO is the wrong thing to buy; this is the fourth cousin of that argument, pointed at our own industry's favourite screenshot.
What the current answers to this question leave out
We ran "should i redesign my website or improve seo" through our own metasearch on 30 September 2026. It returned 41 results. Of the 15 we read, 12 were web design or digital marketing companies whose business includes selling one of the two answers. None of the 15 gave a numeric threshold, and none told the reader to separate the lab test from the field data before deciding.
The demand is real and badly served. Autocomplete returns "website redesign singapore" as a query people type, and a seed of "slow website" came back with 40 suggestions including "how to fix slow website", "how slow is my website", "slow website check" and "slow website fix" (Google Autocomplete, pulled through our keyword tool, 30 September 2026). Thirteen of those 40 were not about the searcher's own site at all — slow cookers, slow travel, slow-motion video and one large overseas retailer's checkout — which is a reminder to read what a keyword tool hands you before acting on the count.
A caveat on our own table. Field data is reported per origin or per page depending on how much traffic Google has, it lags by weeks rather than minutes, and "not enough data" is not the same as "fast". Two of our nine sites sit in that gap and we cannot say either way. An SEO audit is where the server-side half of this gets tested properly.
One next step
Send us the URL and the PageSpeed screenshot somebody used to quote you. We will tell you which of the two numbers your quote was built on, whether your real visitors are actually waiting, and what we would fix instead — including the answer "nothing, your site is fine, your problem is the copy". It starts as a free visibility check through our contact page.
What we will not do is promise a ranking, a lead volume or an AI citation from a speed change.
FAQ
Does a low PageSpeed score mean my website needs rebuilding?
No. The score comes from a single throttled lab test, not from your visitors. On nine Singapore sites we audited on 30 September 2026, all nine scored under 70 on mobile, yet six of the seven with real-user data met Google's 2.5-second LCP threshold. Check the real-user section before you accept a rebuild quote built on the score.
What is the difference between lab data and field data in PageSpeed Insights?
Lab data is one simulated load on a throttled phone in a data centre, useful for finding which element is heavy. Field data is the Chrome UX Report, which Google describes as reflecting how real-world Chrome users experience the web. Field data describes your actual customers. Lab data describes a test.
What counts as a good Core Web Vitals result?
Google's guidance is Largest Contentful Paint within 2.5 seconds, Interaction to Next Paint of 200 milliseconds or less, and Cumulative Layout Shift of 0.1 or less, assessed at the 75th percentile of page loads and split between mobile and desktop. Three in four of your real visits have to clear the bar, not the average visit.
PageSpeed says there is not enough data for my site. Is that bad?
It means Google has not recorded enough Chrome visits to your page to report on them, which is a traffic finding rather than a speed one. Two of the nine sites we audited were in this position. A new website does not create visits, so a rebuild is the wrong answer to this particular message.
When is a website rebuild the right decision?
When the reason is not speed. If you cannot edit your own pages, if the platform cannot produce the page types your customers search for, or if the site is genuinely unusable on a phone, rebuild. Those problems survive every optimisation. A slow lab score usually does not.
How much of my search budget should go to speed work?
Less than most proposals suggest, if your field data already passes. In the sample above, the fixes we would make first cost an afternoon each: a fee band, opening hours, and a tap-to-call link. Speed work earns its place when the real-user numbers fail, as they did on one of our nine sites.