We are less than 10 weeks out from our biggest launch in A&D history. This report covers the progress to date of this newly created role, and my observations of what's working, and what isn't.
That question, going into the new-site launch, came down to what we build, what we carry forward, and what we leave behind. Here's what I've shipped and set in motion while answering it.
Wrote A&D's first tone-and-voice guide and rolled it out across the business, so everything we publish finally sounds like one company.
A live revenue-insights tool for sales: pipeline, performance and forecasting in one place, replacing the spreadsheets nobody believed.
The master sheet the entire Showcase migration runs from; every listing, status and mapping reconciled in one place.
A drag-and-drop content builder for Showcase, complete with our own demo property, El Edificio, so we can prototype and ship product marketing without ever touching a live listing.
Updated A&D's Showcase content-specs document so every project brief now points to the current Showcase Drop production PDF; one source for exactly what each listing needs.
Hired an external agency specialising in SEO, SEM, CX and strategy to pressure-test and bullet-proof the new-site launch.
Surfaced that we have no accurate single source of truth for paid versus unpaid listings. It isn't a tech bug; it's a business-process gap, and it's been costing us quietly.
Co-designed a CMS / CRM / SEO hybrid listings-management system for the new A&D. As far as I can find, nothing like it exists in market.
From here, Dan, Mahamud and I run a fortnightly market-led roadmap and prioritisation session, so the build keeps answering to the market, not the other way round.
This is the first full quarter we've had a specialist partner working alongside us, and the first time the way we run has been looked at from the outside in. Fresh eyes notice what familiar ones stop seeing.
None of what follows is a verdict on anyone's work. When you live inside a business every day, the things most obvious from the outside are the hardest to spot from within; that isn't a failing, it's just proximity. Three months of an outside perspective have surfaced things about our website and our market we genuinely didn't know to look for.
So read this as a set of discoveries rather than a scorecard: things we understand now that we couldn't before, and, just as importantly, still have time to act on before launch.
We have some of the sharpest minds in the industry. The question is whether our tech is letting them do their best work.
Three months, dozens of deliverables. What's falling behind these words is the paper trail: audits, maps, dashboards, rebuilds.
This industry has changed more in the last two years than in the decade before, and the change is measured, public and on the record.
“The hottest new programming language is English.”
Two things are true at once. The tools are in everyone's hands now, and they still only produce reliable work in skilled ones: Google's AI-written code is “reviewed and accepted by engineers,” and in that same 2025 survey more developers distrusted raw AI output than trusted it. The craft didn't vanish; it moved up a level, from typing code to directing it. So the real question for us isn't who works hard; it's whether we're set up to capture that speed, or still doing by hand what the tools now do in days. The new platform is our chance to close that gap.
Three months of groundwork is starting to show in the numbers. Every item here either lowers a launch risk or puts points on the board.
The new website fixes many of these. It does not explain why a listings business let the fundamentals break and stay broken, in some cases for years. None of what follows is exotic engineering; it's basic publishing hygiene, on a platform built to let the mistakes through.
The sitemap is the index Google leans on to find what's new and drop what's gone. Ours sat unchanged for roughly eight months, on a site that lists and retires properties every week.
Google treats sitemap freshness as its most-used discovery signal; a stale one means projects launched since October weren't reliably found, while sold-out ones kept being crawled. For a business whose entire product is fresh listings, keeping the sitemap current is the most basic job there is, and it was left for months.
To look fast to search engines, the site serves crawlers a separately-rendered version of each page. Misconfigured, it leaves 2,121 retired listings answering "live" to Google long after they have sold out and disappeared for buyers.
Google keeps crawling and ranking pages that lead nowhere, spending its budget on our dead stock and burying the listings that are actually for sale. A page that has gone should return "not found" or redirect; answering "200 OK" and then bouncing the visitor is what Google calls a soft 404, and it's a documented mistake to avoid.
The same rendering setup, meant to speed pages up, does the opposite when it is misconfigured. On ORO Newstead, the brochure viewer renders an entire ~15 MB, 100-plus-page brochure into the page's crawlable code.
Google then has to load a page several megabytes heavy; it crawls and renders slower, and slow pages rank worse and burn crawl budget. A feature built to impress buyers ends up dragging the listing down, because no one configured how the crawler sees it.
Links across the site point at pages that don't exist, and because nothing checks them, the mistakes ship straight to production. Some live links have URLs that are fragments of internal A&D staff conversations, pasted in and never validated.
Every broken link is a dead end for a buyer and a wasted crawl for Google; the staff-message URLs are the kind of thing a competitor or journalist screenshots. None of it should be able to reach a live page in the first place.
There is no validation on the basics, so plain errors go live: a Chermside project (a Queensland suburb) filed under Victoria, New Farm with the wrong postcode, suburbs simply misspelt. Dozens of listings carry one of these.
A Brisbane apartment in the Melbourne drawer is a listing nobody finds. The issue is not the typos; it is that the platform is built to allow them, with no check sitting between a person's slip and the live website.
The homepage tells Google to rank a www address that immediately redirects elsewhere; around 25% of pages have no canonical tag at all.
The homepage is the worst place to get this wrong, it is close to a one-line fix, and it is still open.
Our project pages carry no schema at all, while apartments.com.au ships five types; plus 5,017 schema errors elsewhere on the site.
Schema is the label on the tin; our best products sit unlabelled, so Google can't tell what is inside and won't surface them in the rich results or AI answers buyers increasingly start from. A basic, standard thing, missing entirely.
The site runs on Angular 11, which stopped receiving security patches in 2022, and sends no content-security policy while handling buyer enquiry data. Terms and Privacy also cite two different legal entities.
No patches arrive for known vulnerabilities at all; it is a named OWASP Top-10 exposure on a page collecting personal data, and the kind of thing a partner or regulator notices. Unpatched since 2022 means it has been sitting open for years. (A web application firewall sits in front; the legal items were captured in May, re-verify before external use.)
The platform is the bottleneck. It keeps breaking the good work our people put in.
For years, the case for offshore development and performance support was simple: lower cost. That case was made before AI changed what a small, senior team can produce, and before we started adding up what the errors actually cost us.
The pattern is the tell. The same kinds of mistakes get flagged, acknowledged, and then reappear: SEM errors Luster catches in our own accounts, listing data going live wrong, fixes that don't hold. Each time, a recurring error is acknowledged and pushed back to the offshore team; and each time, it comes back.
This isn't about the people; it's about the arrangement. A model that ships errors faster than it saves money isn't a saving. Once the rework, the lost rankings, and the hours our own team spends catching mistakes are counted, the cost advantage thins out, and the quality gap doesn't close.
So the question worth putting on the table: does an offshore-for-cost model still earn its place in 2026, when a senior partner with the right tools can deliver more, with fewer errors, at a cost far closer than it used to be?
Apartments.com.au; Urban.com.au is its old domain, now redirecting into it. Same external audit, run on both sites. Here's the scoreboard.
Higher is better on every line, and we're behind on all of them. Our numbers are in grey; theirs in black.
Traffic figures are third-party estimates, not measured sessions. Domain Rating, keyword and backlink counts are tool-measured. · Head-to-head 30 May; Ahrefs API refresh 12 Jun 2026.
86% of their top 2,500 backlinks are redirects from old domains; 1,561 from Urban.com.au alone, 440 from Property Observer. Only 345 are earned directly to the brand. It's like taking over a well-known old shop and keeping its sign over the door: “Urban racing ahead” is that inherited reputation, switched on with a redirect.
So the gap is structural. That, we can build.
None of these need us to out-spend anyone; they need us to out-build them structurally, which is what the new site is for.
Their most valuable asset is permanent pages for completed projects; 803 exclusive keywords. We delist and redirect ours. We're deleting our own authority.
They own a built /new-townhouses/ section. We have nothing there; “townhouse” alone is ~4,100 searches/mo at very low difficulty.
Our differentiator. Industry profiles already win keywords they can't touch. This is the ground we should own.
They out-structured us. And structure, we can fix.
The audit shows how the current site has been broken, and for how long; some of it goes back years. The new site is a clean slate, and the job is to not inherit a single one of those problems. Before we sign off the build, these are the questions we need answered. Tobey has put the full set to Dan; this is the shortlist that decides whether the new site launches clean. Green is already handled by the platform; orange needs a commitment from whoever builds the site.
Today Google indexes only 8,710 of the 22,410 URLs it holds for us, against ~13,000 real pages. The new build has to serve real, crawlable HTML and schema so our genuine pages get indexed; this is the single biggest question on the list.
Dead pages return a proper 404 and moved pages a server-side 301, and every URL is built correctly or not at all; no malformed links, no Brisbane listings filed under /vic/. It's what stops the 2,121 ghost pages, and the wrong-location URLs, from happening all over again.
Sub-second responses, so mobile speed passes instead of failing. Today a buyer waits over a second before anything loads.
Structured data, clean titles and self-referential canonicals, all checked before a page can publish. This is the one guardrail on the list we don't have to build or wait for; Atrium already does it today, which is a large part of why it belongs in the launch.
A definitive list of live URLs, and a check that confirms a fix actually shipped before it's marked done. Both were missing this quarter, and both are cheap to put in place.
A clean slate only stays clean if we refuse to carry the old problems into it.
We've long known A&D needs a proper content management platform; for our customers, and for us. But why stop at listing content…
Co-designed with Luster, Atrium is a CMS powerhouse meets SEO machine. It generates and governs our content, it takes the manual grind off our team and our customers, it prevents errors and promotes proactivity.
Atrium would be a backend worthy of Jess's designs.
This is how A&D moves from chasing the tech pack to setting the pace.
A keyword engine derives themed pages from every live listing's location and developer. It has already generated a 57-page education library across 8 topic clusters; roughly 46,000 monthly searches; a whole new top-of-funnel section, built automatically.
Pages can't go live until broken links clear and hygiene rules pass; schema coverage reaches 14+ types, where the category sits around 4–5; and a render-check confirms Google sees the same page a browser does, which is the exact failure behind today's indexing problem.
An overnight pass re-checks every page for the exact faults in section 05; broken canonicals, soft-404s, missing schema, blocked crawlers, and lays out what changed on a "caught overnight" dashboard, with alerts the moment something breaks. The reason those problems went unseen for months is that nothing was watching the site between reports. Atrium watches every night, so the next one gets flagged in a day, not a quarter.
Keyword gap and cannibalisation checks, AI-written ad copy through a brand-voice gate, and a negative-keyword builder grounded in our own search data that already identified ~$2,300/month of wasted spend.
It replaces the hand-built 68-page PDF with a live drill-down micro-site; and its maths validator caught 12 numerical errors in the FY26 Q2 deck, including a chart that summed to 110%.
It replays the full 14,093-row redirect map, tracks indexation, flags blocked crawlers, and buckets junk URLs; the launch safety net we don't currently have.
The five above are the engine. These are the parts that, the first time you see them, make the rest of the category look like it's standing still; and every one is already running in the build, not sitting on a roadmap.
Sense-check runs each page past three models at once (Claude, GPT and Gemini) and scores where they agree; a cross-vendor audit no single tool, and no tired human at 5pm, can match. Problems get caught before a buyer or Google ever sees them.
AI Improve proposes the new copy and shows it as a clean before-and-after diff, then a person clicks approve or reject. This is the pipeline that clears the 5,945-page backlog, with nothing going live unread.
Counterparts puts our page four-up next to REA, Domain and apartments.com.au, with an AI summary of what they do that we don't. The market gap from section 06 stops being a chart and becomes a working to-do list.
An editorial queue watches every page's score and traffic trend and drafts refresh proposals as pages age. It's the part that stops the new site quietly sliding back to where the current one is; SEO that maintains itself.
What took days by hand becomes a generated micro-site; with a validator that catches the maths errors a tired human misses.
The ~35-campaign manual sprawl becomes structured, auto-built campaigns; ad copy and meta drafted through a brand-voice gate.
This is the whole point: stop spending our best people's time on work a system should do, so they can do the work only they can.
We have only the weeks before launch to ship the new website and the campaign behind it. The prototype already exists, so this is scope and integration, rather than a design & build from scratch.
My recommendation is that we engage Luster for software design and development of Atrium. I do not think this one should sit with the devs. This one sits with us.
It comes down to this. We are slower than we should be.
Our people aren't the problem; our tech is. The gap with the market is structural, which means it's closable. And we have a working prototype that closes it, weeks before launch.
None of this asks you to trust me blindly. Every figure is checked against the live site and against published industry standards.
A&D surprises and delights our customers, instead of catching up to our competitors.
Indexation, clicks, Core Web Vitals: Google Search Console, a-d.com.au domain property, re-checked 8 Jun 2026.
Web-health audit: 83-finding register, baseline 22–28 May 2026, re-checked 8–9 Jun 2026 against the live site.
Migration & URL figures: Smoosh migration database, reconciled live 11–12 Jun 2026. Governing values: 12,733 live URLs to migrate (12,727 covered, 6 outstanding, 2 of them genuine gaps); a 14,093-row redirect map, larger than the live set because it also catches old, duplicate and variant URLs; 7,817 was the April mapping baseline, now superseded.
Crawl-based counts (orphans, schema, redirects, content): Ahrefs site crawl, 9 Jun 2026, URL-level counts.
Industry standards referenced: Google Search Central — soft 404s & expired/removed content (developers.google.com/search); web.dev — Time to First Byte thresholds (web.dev/articles/ttfb); OWASP — vulnerable & outdated components and Content Security Policy (owasp.org).
Industry context (2026 section): Alphabet Q3 2024 earnings call (S. Pichai, “quarter of all new code”); Microsoft at LlamaCon, Apr 2025 (S. Nadella, 20–30%, via CNBC/TechCrunch); Stack Overflow Developer Survey 2025 (survey.stackoverflow.co/2025); GitHub developer-productivity study (arXiv 2302.06590); A. Karpathy, 2023.
Competitor head-to-head: external capture 30 May 2026; Ahrefs API live refresh 12 Jun 2026. Traffic and dollar figures are third-party estimates, labelled as such.
Server-log root-cause review: production access logs, both frontends, 26 May–9 Jun 2026 (1.2M requests).
Paid search: Google Ads, QLD review 25 Apr–5 Jun 2026; NSW/VIC structures delivered 5 Jun 2026.
The platform (Atrium) & new PDP: Luster build log and in-browser verification, Jun 2026; master schema matrix; First Light design exports.
New-build safeguards: Tobey's new-build questions for Dan (dev-questions review), Luster, Jun 2026.
Prepared from: the Q1 findings pack, 12 Jun 2026. Standard applied: accuracy over impressiveness; anything that couldn't be verified was left out.