Dirty South Power Wash Ltd is a pressure washing and exterior cleaning business based in Horley, Surrey, working across Surrey and Sussex. Driveways, patios, roofs, gutters and solar panels, for homes and businesses. Companies House number 14945121. The site is at dirtysouthpowerwash.co.uk.
This case study covers the first month. It is deliberately early, because the useful thing to show a business owner deciding whether to spend money on a website is not a polished result eighteen months later. It is what actually happens in the first four weeks, including the parts that have not happened yet.
The site it replaced
Nine pages on a GoDaddy Websites and Marketing builder. It worked in the sense that it loaded and it had a phone number on it. What it did not have was anything for Google to rank.
There was no page about driveway cleaning. No page about roof cleaning. No page for Reigate, or Redhill, or Caterham, or any of the other towns the van drives to. A search engine can only put a site in front of someone if there is a page that answers what that person typed, and for every search except the business name itself, there was nothing to offer.
That is the normal position for a small trade business, and it is not the owner’s fault. Builder platforms are sold on how quickly you can get a site up, not on whether the site can win work.
Moving it, without breaking the email
The rebuild went live on the client’s own domain on 3 August 2026.
The part that worries owners most is email, and rightly. The domain’s nameservers moved to Cloudflare the same night, and all eleven Microsoft 365 DNS records were carried across and verified before the switch, so mail kept flowing throughout. Nine redirect rules covered the old addresses, and each was tested end to end as a 301 landing on a live page before go-live rather than after it.
What we built
Ninety pages, static, built with Astro and served from Cloudflare:
- 7 service pages, one for each job the business actually does
- 12 town pages
- 48 pages pairing a job with a town, written only where there is something specific to say, so “driveway cleaning in Reigate” is genuinely about driveway cleaning in Reigate
- 15 area pages
- 6 guides, plus the standard pages a trade site needs
Those town-and-job pages are the part that deserves a word of caution, because scaled location pages are exactly the pattern Google treats as spam when there is nothing specific on them. These exist only where there is something real to say about the town and the job. Where there is not, we would rather not have the page, and we review the set with the owner as we learn which areas the work genuinely comes from.
We rewrote the opening of every page to answer the search question first. Eighty-five page openings on the first pass, eighty-seven of ninety when we re-measured on 17 August, each one unique rather than a template with the town name swapped in. If someone searches “will pressure washing damage block paving”, the guide’s first two sentences answer it. That is the whole trick, and it is not a trick: you can read it working on the block paving guide.
The engineering nobody sees
Three things ran under the copy work, each of them now enforced by a gate that fails the build rather than by anyone remembering.
Internal links. 3,712 links in the site were missing their trailing slash, which means every internal click paid for a redirect before it got anywhere. We fixed all of them, and a link gate now checks every internal link on every page on every build. On the 8 August build it reported 4,205 links across 87 pages, all resolving.
Images. 734 image tags were given alt text and real width and height attributes, so the page stops jumping about while it loads and so the images are described for anyone who cannot see them. An image gate now fails the build if a single one regresses.
Weight. The home page hero was a single 675 KB image. It now ships as responsive sizes at 400, 640, 960 and 1280 pixels wide, along with 21 gallery photographs and 13 before-and-after pairs.
What we took off the site
This is the part we are most willing to be judged on.
We removed the self-published star rating markup. A business that controls the reviews shown on its own pages is not eligible for Google’s star review feature on those pages, so publishing that markup was asking for a rich result the site could never have earned. It came off.
We removed the FAQPage markup. Google stopped showing FAQ rich results on 7 May 2026, so the markup no longer buys anything, and carrying schema that does nothing is just weight and a maintenance liability.
We built a claims gate. It reads the rendered pages after every build and fails the deploy if wording appears that no document in the client’s file supports. The rule it enforces is the one we run to everywhere: we publish what a certificate or a record proves, and nothing else. On insurance the site carries exactly one line, “£1,000,000 public liability with Admiral Business”, against a certificate we have seen and dated.
The seven Google reviews shown on the site were copied word for word from the live listing on 1 August 2026. None were written by us.
Month one in search
All of the following comes from the client’s own Google Search Console, the 28 day window to 15 August 2026, which holds data from 2 August because that is when the site started being seen on its new domain.
Google showed the site 8,769 times and sent 13 clicks. Weekly impressions rose from 2,612 in the week of 2 to 8 August to 6,157 in the week of 9 to 15 August, 2.36 times in a week.
Of the 630 searches Google names in the export, about sixty already average position 10 or better, and a further 106 sit between 11 and 20. At 15 August 2026 that included “power wash paving caterham” at position 3, “jet wash near me” at 2, “driveway pressure washing near me” at 2, “jet wash driveway caterham” at 8, “pressure washing nutfield” at 7, “pressure washing redhill” at 9, “pressure washing surrey” at 9 and “patio cleaning horley” at 10. The guides are the strongest single group. Taking the same 28 day window to 15 August 2026 by page rather than by search, the block paving guide averaged position 10.0 across 185 impressions and the render guide averaged position 7.7 across 91 impressions, and guide pages took 5 of the 13 clicks the site won.
Now the caveat, which matters more than any of the above. 91.8 per cent of impressions, counted on the page dimension, are still at position 21 or worse, meaning page three and below, where almost nobody clicks. Thirteen clicks from nearly nine thousand impressions is what that looks like. A site in its first month being shown a lot and chosen rarely is the ordinary sequence, not a failure, but it is not a result yet either and we are not going to present it as one.
The biggest single theme in the data is roof cleaning, sitting at positions 14 to 24 across seven towns. That is the obvious next month of work, along with the town searches parked on page two.
What we would want you to take from this
Ninety pages of genuinely different content beat nine pages of nothing, and Google found them within days. That part is repeatable, and it is most of what the first month of any site like this consists of.
The rest of it is discipline. Gates that fail the build are worth more than good intentions, because good intentions do not survive the fourth week. Removing markup that has stopped working is worth more than adding markup that never did. And publishing the 13 clicks alongside the 8,769 impressions is the only version of a results page anyone should trust.
The site runs on the Essentials plan. What is in each tier is on the pricing page, and the shape of the work is on the services page.