10 Singapore Renovation Websites, Measured (2026 Study)
Contents15
- The numbers from the 19 September sample
- What we measured, and on whose sites
- How fast are these sites on a phone?
- Where our own speed number overstated the problem
- Seven of the ten had no real-user data at all
- 314 of 767 images carried no alt text
- The portfolio page problem: 230 words and 23 photos
- Four sites had a dead internal link, and three were the same one
- Nine dead external links, four of them pointing at HDB
- What every site in the sample already got right
- What I would do this week
- What it costs
- Method and limits
- Send us the page you would defend
- FAQ
A homeowner with a 4-room resale flat opens six renovation websites on her phone in one evening. She is comparing quotes, not admiring layouts, and she closes whichever page makes her wait. On 19 September 2026 we measured ten Singapore renovation and interior design websites to see what she would have run into.
The basics are done. The photos are the problem. Every site in our sample was on HTTPS, served a readable robots.txt, and put exactly one H1 on every page we opened. None of the 44 pages needed JavaScript to show its content. But 314 of the 767 images we sampled (41%) carried no alt text, nine of the ten sites scored under 90 on Google's mobile PageSpeed test, and the median page took 7.1 seconds to paint its largest element in the lab.
One finding cut against us, and we report it in full below: on three of the sites, real visitors were loading the page in about a second while our own speed test called it slow.
The numbers from the 19 September sample
Every figure below comes from one run on 19 September 2026: our own audit crawler fetched up to five pages per site, verified every link it found, and pulled each site's mobile score from Google's PageSpeed Insights. Counts are always given with their base.
| # | What we measured | The number |
|---|---|---|
| 1 | Sites scoring under 90 on mobile PageSpeed | 9 of 10 |
| 2 | Sites whose lab LCP was above Google's 2.5-second threshold | 10 of 10 |
| 3 | Sites with a lab LCP of 8 seconds or worse | 5 of 10 |
| 4 | Sites with any real-user (field) LCP figure available | 3 of 10 |
| 5 | Sampled images with no alt text | 314 of 767 (41%) |
| 6 | Sampled pages under 300 words | 7 of 44 |
| 7 | Sites with at least one internal link returning 404 | 4 of 10 |
| 8 | Sites with at least one dead external link | 4 of 10 |
| 9 | Homepages declaring a business entity in structured data | 8 of 10 |
| 10 | Sites with FAQ markup anywhere in the sample | 5 of 10 |
Row 5 is the one we would act on first, and it is the cheapest to fix.
What we measured, and on whose sites
We ran four searches a Singapore homeowner actually types, then audited the businesses that came back. The searches were "hdb renovation contractor singapore", "best hdb renovation contractor singapore", "interior design singapore hdb" and "hdb kitchen renovation singapore".
All four are real: Google Autocomplete returned each of them as a suggestion on 19 September 2026.
From the results we kept Singapore renovation and interior design businesses only. We dropped directories, "top 10" blogs, loan comparison sites, Pinterest, TikTok, YouTube and Carousell. We also dropped one business SingRank works for and replaced it with the next result. That left ten websites, and we audited five pages of each where the sitemap offered five.
One site's audit timed out twice and was excluded; we replaced it with the next renovation business in the same search. Nothing in this study names a business, and nothing here is a rating of any company.
How fast are these sites on a phone?
Slow, by the lab measure. Mobile PageSpeed scores ran from 42 to 90, with the middle of the sample around 60. Nine of the ten sites came in under 90.
The Largest Contentful Paint numbers tell the same story more usefully. Google's Core Web Vitals documentation says to "strive to have LCP occur within the first 2.5 seconds of the page starting to load". In our lab tests, all ten sites were above that. Five were at 8.0 seconds or worse, and the slowest painted its main content at 14.9 seconds on an emulated mid-range phone.
The cause is not mysterious. These are portfolio businesses. One homepage in our sample loaded 113 images. Another loaded 108. A third carried 33 on a page of 429 words. Photographs are the product, so nobody wants to remove them, and almost nobody resizes them.
Our guide to Core Web Vitals for Singapore websites covers the three thresholds and how to measure your own for free.
Where our own speed number overstated the problem
Here is the part that costs us something. PageSpeed Insights reports two different things, and only one of them describes your customers.
The lab score is a simulation: a throttled mid-range Android on a slow connection. The field data is the Chrome User Experience Report, built from what real Chrome users actually experienced on your site.
Three of our ten sites had enough traffic for field data. Their real-user LCP figures were 1.0 seconds, 1.6 seconds and 1.0 seconds. Every one of those sits comfortably inside Google's 2.5-second threshold. Two of those three sites scored 42 and 44 in the lab.
So if we had opened a sales call by saying "your site takes 14.9 seconds to load", we would have been wrong about what visitors experience. In our view, an agency that quotes only the lab number is selling you a fright. Check the field data first. If it exists and it is green, your speed problem is smaller than any audit tool will tell you.
Seven of the ten had no real-user data at all
Only 3 of 10 sites returned a real-user LCP, and only 2 of 10 returned a real-user INP. On the other seven, Chrome had not seen enough visits to form a sample.
That absence is itself a finding, and it is the more serious one. A renovation firm with no field data is a renovation firm with too few visitors for Google to measure. Speed is not that site's problem. Being found is.
If that is you, the fix is not a faster image. It starts with a complete Google Business Profile and the local work set out in our renovation SEO guide.
314 of 767 images carried no alt text
Across the 44 pages we sampled, 767 images loaded and 314 of them (41%) had no alt attribute at all. Six of the ten sites had at least one. On three sites the majority were missing: 76 of 86 sampled images on one, 70 of 101 on another, 143 of 225 on a third.
Four sites had none missing, which shows this is a habit, not a technical limit.
For a renovation business this matters more than it does for most trades, because your photographs carry the information. A picture of a kitchen you built in Tampines with a sintered stone top and a tall pull-out pantry is a fact about your work. Without alt text, that fact is invisible to a search engine, to an AI assistant reading the page, and to a customer using a screen reader.
Alt text is also the cheapest item on this entire list. One line per image, written by whoever chose the photo, describing what is in it and where. No developer required.
The portfolio page problem: 230 words and 23 photos
Seven of the 44 pages we sampled ran under 300 words. Six of those seven were project pages.
One site showed the pattern completely. Four of its five sampled pages were individual project write-ups of 227 to 237 words, each carrying between 11 and 23 photographs. Several of those pages also had no structured data of any kind, page titles between 77 and 121 characters, and meta descriptions running to 403 characters.
A project page like that is a slideshow with a caption. It tells a machine almost nothing, and it tells a homeowner less than it could. The flat type, the town, the scope, the rough cost band and how long it took are all facts the business already knows and did not write down.
It is also the easiest page on the site to improve, because nobody has to invent anything. The words come first; the markup can wait.
Four sites had a dead internal link, and three were the same one
Four of the ten sites had at least one internal link returning a 404. On three of those four it was the identical path: Cloudflare's email-protection URL, produced automatically when Cloudflare obfuscates an email address on the page.
We flag it because every crawler and every audit tool will report it, and because it is not really a content problem. If your audit report shows one broken internal link and it ends in /cdn-cgi/l/email-protection, that is Cloudflare, not neglect. Do not pay anyone to fix it.
The fourth was a different matter: a link in a blog index built from a template variable that never rendered, so the anchor pointed at the raw placeholder text. That one is a genuine defect, and it sat on the page that lists every article the firm has written.
Nine dead external links, four of them pointing at HDB
Four of the ten sites carried at least one external link returning a 4xx, nine across the sample.
The pattern inside them is worth a renovation owner's attention. On one site, four of its six dead external links pointed at pages on hdb.gov.sg that no longer resolve. The firm had done the right thing years ago, citing the housing authority's own renovation guidance, and the authority reorganised its website.
The live pages exist. MyNiceHome's HDB renovation guide states that "the renovation contractor you engage must be listed in the HDB Directory of Renovation Contractors", and that "as the HDB flat owner, you are responsible for obtaining all HDB renovation permits required for your flat's renovations before renovation works begin". The same page gives the permitted hours: general renovation on Monday to Saturday, 9am to 6pm, not on Sundays or public holidays, and noisy renovation on Monday to Friday, 9am to 5pm only.
If your site cites HDB anywhere, open those links today. A dead link to the regulator is worse than no link at all, because the homeowner who clicks it was checking whether to trust you.
What every site in the sample already got right
We should be fair about the base. On 19 September 2026, all ten sites were on HTTPS and all ten served a robots.txt our crawler could read. Every page we opened had a viewport tag and exactly one H1. None of the 44 pages required JavaScript to render its main content, which matters because AI crawlers generally do not run it.
Eight of the ten homepages declared a business entity in structured data, most often HomeAndConstructionBusiness or LocalBusiness. Five of the ten had FAQ markup somewhere in the sample.
This mirrors what we found in our wider Singapore AI search readiness study last week: the technical fundamentals are largely handled across Singapore's local trades, and the gaps that remain are gaps in words and in images.
What I would do this week
Three calls, in this order, for a renovation or interior design firm.
- Write alt text for every photo on your three most important pages. Name the room, the flat type and the town. "Kitchen carpentry in a 4-room resale flat, Ang Mo Kio" beats "IMG_4471". Budget an afternoon.
- Give every project page 150 more words of fact. Flat type, town, scope, duration, and a price band if you are willing to publish one. You already know all of it.
- Check your field data before you spend a cent on speed. Run your homepage through PageSpeed Insights and look for the real-user section at the top. If it is there and green, leave the images alone and go back to point one.
Then, and only then, compress your photographs.
What it costs
Alt text and project copy cost you time, not money: roughly a day of somebody's attention for a three-page pass, and that somebody has to be a person who was on site, not a writer guessing from photos. That is the real cost. The work is boring and it cannot be outsourced cheaply without becoming wrong.
Image compression costs more. Done properly it means regenerating your portfolio at sensible dimensions, which on a site with 113 images on the homepage is a developer job, not a plugin job.
And you give something up either way. An afternoon spent on alt text is an afternoon not spent quoting a job. If your enquiries are already more than you can service, none of this is your best use of a Tuesday.
Method and limits
Read these before quoting any figure above.
- One day, one run. All measurements were taken on 19 September 2026. A snapshot, not a trend.
- Ten sites, 44 pages. Small numbers. One site more or less moves a "of 10" figure by 10 points.
- Metasearch, not Google. The four searches ran through SingRank's self-hosted SearXNG. Its blended ordering is not Google's ranking and not the Maps pack. Most of its engines returned errors that day; the usable results came from a single engine.
- Up to five pages per site. One site's sitemap offered only one page in our sample, so 44 distinct URLs came from ten sites.
- Lab speed is a simulation. PageSpeed mobile scores and lab LCP come from an emulated device. Where real-user data existed, we report it and it disagreed.
- We did not measure AI crawler access beyond robots.txt readability, Google Business Profiles, reviews, backlinks, or anything behind a login.
- One exclusion and one replacement. A business SingRank works for was replaced with the next result, and one site whose audit timed out twice was replaced the same way.
Our method page sets out how we separate what we measured from what we think.
Send us the page you would defend
Pick the one page you would want a homeowner to land on, and send it to us. We will run the same checks on it, tell you what your real-user data says, and tell you which of the three fixes above is worth your time. That starts as a free visibility check through our contact page.
If you want us to do the work afterwards, our SEO, AEO and GEO programme starts from S$500/month as the live page states it today. What we will not do is promise you a ranking. Google and the AI assistants choose their own answers, and no agency controls them.
These findings describe ten websites sampled on 19 September 2026. This page names no business and is not a rating of any company.
FAQ
Do renovation websites need to be fast to rank in Singapore?
Speed helps, but it is rarely the biggest lever for a renovation firm. In our sample, seven of ten sites had too little traffic for Google to collect real-user speed data at all. A site nobody visits does not have a speed problem. It has a visibility problem, and those are fixed differently.
What is a good LCP for a renovation portfolio page?
Google's documentation says to aim for the Largest Contentful Paint to occur within the first 2.5 seconds of the page starting to load. All ten sites we measured exceeded that in lab conditions, with a median of 7.1 seconds. Real-user figures, where they existed, were far better at around 1 second.
Does missing alt text hurt my renovation website?
It costs you description, not ranking directly. Your photographs carry the facts about your work: the flat type, the finish, the town. Without alt text, a search engine, an AI assistant and a screen reader all see an empty box. In our sample, 314 of 767 images had none.
Should I pay someone to fix a broken cdn-cgi/l/email-protection link?
No. That path is generated by Cloudflare when it hides an email address on your page, and it appeared as the only internal 404 on three of the four affected sites in our sample. It will show up in every audit tool. It is not a sign that your website is neglected.
How many words should a renovation project page have?
More than the 227 to 237 we found on one site's project pages. There is no magic number, but a page carrying 20 photographs and 230 words is a slideshow. Add the flat type, the town, the scope, the duration and a price band, and it becomes something a buyer and a machine can both use.
Can I cite these figures?
Yes. Please cite "SingRank, Singapore Renovation Website Study 2026" with a link to this page and the collection date, 19 September 2026. Quote each percentage with its base, such as 314 of 767 images, and carry the metasearch and sample-size caveats with it. Corrections are welcome, and the contact page above is the fastest route.