0431 124 128
Website Diagnostic
Kestrel Joinery
An independent review of kestreljoinery.example, and what I would do about it.
Read this first
This document records what I observed and what I think about it, and the two are kept apart throughout. Everything under Observed is a fact I measured or read directly from a publicly available page on the date shown, and you can reproduce it yourself. Everything under My assessment is my opinion as a web developer, and reasonable people in my trade could disagree with some of it.
Nothing here is legal, accounting, tax or regulatory advice, and I am not qualified to give it. Where I mention privacy, consumer law or accessibility, I am flagging something to put to your own adviser, not telling you what the law requires of you.
I make no promise about rankings, traffic, enquiries or revenue, and you should treat anyone who does with suspicion. A review is a measurement, not a lever. The terms it is issued on are at the end and they are part of it.
The basis of the review
So you can judge the findings, and repeat any of them yourself.
What I did
- Loaded the public pages in an ordinary browser and read the source the server sent back, exactly as a visitor or a search engine receives it
- Timed the server’s first response three separate times, and recorded the page weight and the number of files each page loads
- Read the published robots.txt and XML sitemaps
- Viewed every page at desktop width and at 375 pixels, which is roughly a phone
- Read publicly published listings, reviews and prices for the nearest comparable businesses
What I did not do, and want to be explicit about
- No security testing of any kind. No attempt to reach a restricted or password-protected area, no login tested, no scanner or intrusion tool run against the site
- I hold no credentials for the site, its hosting, its domain or any connected account, and I did not ask for any
- I did not submit a form, place an order or contact the business through the site
- Where a security matter appears below, it describes something the server publishes to anybody who asks for it
What I could not check
- Real-world loading speed for actual visitors. That needs Search Console access, which only the owner can grant
- Where the enquiry forms deliver to, and whether anybody reads them
- Whether off-site backups exist. I would want that answered before anybody touches anything
- Enquiry volume and where it comes from. There is no measurement installed, so I do not believe anybody currently knows
Anything I could not test is named as untested. A report that quietly leaves out what it could not reach is indistinguishable from one that found nothing there.
What I think, in short
Start here: the business is in better shape than the website is. Twenty-two years of work, a workshop people drive past, and a body of finished jobs that nobody else in the region can copy. The site’s plumbing is sound, nothing blocks it from search engines, and the pages do not jump about while they load, which is rarer than it should be.
So the problem is not the business. It is that the website is currently the weakest part of it, and my honest view is that it does not need updating, it needs rebuilding. I say that because of the number of separate structural problems rather than because the platform is wrong.
Almost none of it is hard, and a fair bit of it is free.
My score
38out of 100
Rebuild, not repairThe score is my own weighting across the areas I check on every build. It shows you where the weakness sits. It is not an industry standard, and another developer using their own method would arrive at a different number.
- 6 Critical
- 9 High
- 11 Medium
- 5 Low
- 4 already right
Four of the findings
Every finding in a report is built this way: what was measured, what I make of it, and what I would do. Nothing is asserted without the evidence sitting next to it.
The server takes about nine seconds to answer when the cache misses
I timed the server’s first response three times: 9,140 ms on the home page, 8,902 ms on the kitchens page after a forced reload, and 10,310 ms on a request written specifically to bypass the cache. Requests the cache did answer came back in 38 ms to 160 ms.
This is the most important thing in the document and it is almost certainly invisible to you, because once you have loaded a page the cache answers instantly for the next week. The people getting the nine seconds are first-time visitors to a page nobody has visited recently, everybody for a while after you edit anything, and search engine crawlers, which request uncached pages constantly. A widely used benchmark for a good experience is the main element painting within 2.5 seconds, and a nine second first response makes that arithmetically impossible no matter what else is done. Nothing else on this list is worth spending money on until this is resolved.
Diagnose before spending. Find which query or plugin is consuming the time, check whether object caching is on, and check the worker and memory allocation with whoever administers the server. My working theory is a single slow query running on every page load rather than underpowered hardware, which makes this a cheap fix rather than a hosting move. I would not change hosts until that has been ruled out.
There is no analytics or conversion tracking installed anywhere
I searched the home page source for the identifying code of every common measurement tool. None of them is present on the page. There is no tag manager, no analytics property and no conversion definition of any kind.
Nobody can currently say which pages, which suburbs or which channels produce enquiries. It also means that if you spend money on the website this year, there is no baseline to judge whether it worked. The one thing that cannot be recovered later is history: analytics only records from the day it is installed, so every week without it is a week that can never be reported on. I would rather be straight about the boundary here as well. Measurement tells you what happened. It does not make anything happen, and I am not suggesting that installing it produces enquiries.
Install a tag manager and an analytics property in accounts opened in your own name, that you own and pay for, not mine. Define the events that matter rather than leaving the defaults: enquiry sent, quote form sent, phone tapped, directions tapped. Verify the site in Search Console, also in your own account, and lodge the sitemap. A day’s work, and I would do it before anything else regardless of whether a rebuild goes ahead.
Your finished work is published as decoration rather than as photographs
The home page contains two image elements, both of which are the logo, both with empty alternative text. All fourteen photographs of finished joinery are applied as background decoration in the stylesheet instead. No page serves a modern image format, and the four largest files total 612 KB.
Your photography is your product, and at the moment it is technically wallpaper. Applied this way it carries no description, cannot be sized for different screens, and is only discovered by the browser after the stylesheet has been read, which is part of why the page paints late. Nothing you have built in twenty-two years is findable in an image search.
Republish every photograph as an image with a written description, sized for the screen that asks for it, in a modern format, with the job it came from named. That last part is the bit that pays: a caption naming the timber, the room and the suburb is worth more than the file size saving.
Nothing is blocking search engines, and the pages do not shift while they load
The robots.txt permits every crawler, the XML sitemap is complete and current, no page carries a stray noindex, and layout shift measured at or near zero on every page I loaded.
A report that only lists faults is not a measurement, it is a sales document. These are genuinely right and they would survive a rebuild, so nobody should be paid to fix them.
Findings graded Medium and Low are reported the same way. A full Diagnostic carries twenty or so of them, each with its own evidence.
What the businesses around you are doing
Read from publicly published pages on a single dated day. Prices and review counts move constantly, so they are a snapshot, and anything you intend to act on should be checked at the source.
| Business | Price shown publicly | How you enquire | Reviews shown |
|---|---|---|---|
| Kestrel Joinery | None published | A form, then a call back | 7 |
| Competitor A | From a figure per linear metre | Instant online estimate | 184 |
| Competitor B | Three packaged kitchens, priced | Booked consultation | 96 |
| Competitor C | None published | Phone, prominent on every page | 211 |
Two differences stand out, and neither is about the joinery, because every business in that table makes good cabinets. They answer the price question and you do not, which means a buyer comparing four tabs late at night has a number from three of them and a form from you. And the review counts are not close: seven against a hundred and eighty reads to a stranger as a business that might not still be trading, whatever the quality of the work.
On the businesses named above
In a real report the comparable businesses are named, so the comparison is checkable. I have no relationship, commercial or otherwise, with any business, platform or supplier named in a Diagnostic. Naming one is neither a recommendation of it nor a criticism of it.
What I would do, in what order
Every Diagnostic ends in three stages, and the first one is worth doing whether or not you ever engage me. I would rather say that than pretend otherwise.
Stage one · this week
Stop the bleeding
Cheap, quick, no design decisions, all on the site as it stands.
- Diagnose the slow server response with your host
- Install measurement, in your own accounts
- Publish a privacy policy and a collection notice at the form
- Put a tappable phone number on every page
- Claim and answer the review listings
Stage two · the build
The part that is a project
This is where the money goes, and it is quoted separately.
- The home page rebuilt around what you actually sell
- Photography republished properly, described and compressed
- A price or a price shape on the page
- Business data published, consistent across every listing
- An accessibility pass: buttons, labels, descriptions, form behaviour
Stage three · ongoing
The part that compounds
The pages that should exist and do not, written at a pace you can actually keep, and the habit of asking every finished client for a review. In a real report this stage names the pages, the question each one answers, and the order I would build them in.
I will not quote you a ranking, a traffic figure or an enquiry number, now or later, and I would encourage you to be careful with anybody who does. What I will do is measure where you start, make the changes we agree, and report honestly on what moved and what did not.
The basis it is issued on
Thirteen terms travel with every report and they form part of it. They are the reason a Diagnostic can be handed to an accountant, a business partner or a prospective buyer without anybody having to guess what it does and does not claim. In full, they cover:
- Opinion, clearly separated from fact
- Not legal, financial, accounting, tax or regulatory advice
- No representation about results
- Publicly available information only, and no security testing
- A snapshot of one day
- Third-party information is unverified
- No endorsement, no affiliation, no disparagement
- Quoted reviews and published material
- Prepared for one recipient, and not for anyone else to rely on
- Incomplete by design
- Limitation of liability
- Copyright and confidentiality
- Corrections, and a right of reply
Two of them in full, because they are the two that matter most
Incomplete by design. A review of this kind cannot be exhaustive. The “What I could not check” section sets out the known gaps. There may be faults, risks or opportunities in this website that I did not find, did not look for, or could not see from the outside, and the absence of something from this report is not a statement that it is correct.
Corrections, and a right of reply. I would rather be corrected than be wrong. If you believe any observation in this report is factually inaccurate, tell me what and why, and I will check it and issue a corrected version in writing if I am satisfied it is wrong. The same offer is open to any business named in it. Corrections are recorded against the version number.
Nothing in a Diagnostic excludes, restricts or modifies any guarantee, right or remedy that cannot lawfully be excluded under the Australian Consumer Law.