Rebuilding Six-Year-Old Sites on a Modern Measurement Stack

Daniel Sorenson Daniel Sorenson
flywheel development modernizing the agency technical stack

The short version

  • Some of the sites we run today we first built six years ago. They still work. The web underneath them did not stay still.
  • Core Web Vitals became a ranking input, GA4 replaced Universal Analytics outright, and measuring what a page returns became something a business can own rather than rent.
  • We rebuild onto a modern WordPress framework with speed, crawlable structure and schema handled at the start, not bolted on after launch.
  • Four data sources do the work: Google Search Console, GA4, a weekly inventory of the site’s own content, and Google Ads Keyword Planner.
  • Joined, they answer the question most reporting cannot answer: not how did we do, but how much of the available demand did we capture and what is still unclaimed.

Some of the sites we run today we first built six years ago. They worked then and they work now. What changed is everything else in the ecosystem around them.

Core Web Vitals became a ranking input. GA4 replaced Universal Analytics outright. Ad platforms shifted control from settings to algorithms based on data. Fragmented platform reporting and spreadsheets evolved to centralized datatables. Structured data became how a search engine reads a page rather than a nice-to-have. And the tooling for measuring what a page actually returns stopped being enterprise software and became something a business can own.

So we rebuild them. Same business, same customers, a framework that can carry what is available now: automation in the places that were manual, AI where a defined job exists for it, and measurement wired in from the first commit.

What we rebuild onto

A modern WordPress build on a framework our own team builds and maintains, with the parts that usually get deferred handled at the start: page speed and Core Web Vitals, crawlable structure, schema markup, redirect architecture, data layer variables, and analytics that records what happened rather than approximating it.

The rebuild is the foundation. What makes it worth doing is what we can measure once it is in place, which is where a conversion-focused website redesign and creative and UX design stop being matters of taste and start being decisions with evidence behind them.

The measurement stack, and what each part is for

Four data sources, each answering a question the others cannot.

Four data sources joined into one view of demand available against demand captured Diagram. Four sources feed a single joined view. Google Search Console supplies queries, impressions and position, which is the demand the site is already eligible for. GA4 supplies sessions, key events and channel, which is what visitors did after arriving. A weekly content inventory supplies word count, headings, internal links, anchor text and schema, which is what each page actually says. Google Ads Keyword Planner supplies monthly search volume and competition locally and nationally, which is the demand that exists at all. Joined on the same term, they show the demand available, the share of it captured, and what the page says, in one row. Search Console GA4 Content inventory Keyword Planner queries, impressions, position sessions, key events, channel words, headings, links, anchor text, schema monthly volume and competition what we are eligible for what visitors did what the page says what exists to win Joined on the same term The demand available, the share of it we captured, and what the page actually says. One row, one term, one decision. Impressions alone say how we did. Only the join says how much was there to get.
The four sources and what each one can answer. Joined on the same term, they give the demand available, the share captured, and what the page says.

Google Search Console

What people searched on Google, what the site was shown for, and where it ranked. It is the source for search demand a site is already eligible for. We pull it weekly into a warehouse rather than reading it in the interface, because the interface lacks the ease of per page query and time analysis for global site SEO rankings.

GA4

What visitors did once they arrived: sessions, users, landing pages, key events, bounce rate, scroll depth, and which channel delivered them. Dropped into BigQuery to manage the scale of data and ease the access and speed to access.

Site content inventory

What is actually on each page: word count inside the main content, page copy content, the heading outline, internal links and their anchor text, schema, meta descriptions, and page speed. Crawled weekly from the live site, because what a page says is half of why it ranks and nothing else records it.

Search volume and competition for the terms a business does target and could target, nationally and locally. This is the one source that can provide data-driven targets of what market opportunity exists. The other three are data-reactive and can only report on what the site already reaches.

Market share and volume benchmarks

Joining those four sources answers the question most reporting cannot. Not how did we do, but how much of the available demand did we capture, what that capture represents in market share, and which search terms are open opportunity we are not reaching at all.

For a given term we hold the local and national monthly search volume, the impressions the site earned, the position it held, and what the page actually says. That is a market share view rather than a traffic view. It is also how we tell the difference between a page losing ground and a page that was never in the running for the term it was written for.

Our own site is the simplest example we can give, because we can publish it without asking anyone. The term “seo agency” carries 27,100 searches a month in the United States. We earn zero impressions on it. A traffic report would never raise that, because nothing happened. A market share view raises it immediately, and then makes us answer the harder question of whether that demand was ever reachable for a site our size. We published the full read of our own search data, including the parts that do not flatter us, in what our own search data says.

The same joined data separates the two places growth comes from. There is efficiency in what already runs: a page ranking at 25 that could rank at 12, a budget returning less than it should. And there is the demand nobody at the business has claimed yet, which is usually the larger share and is invisible to any report built on current traffic alone. That is the work behind search engine optimization, Google Ads management and growth strategy.

What we do with it

Insights without action are just reports. The stack exists to produce a decision each week: which page to change, which term to target, which test to run next, and if changes made the week prior should be held or adjusted.

One example, from a client in an industrial category. The industry searches for the machine by its name and technical solutions for use case. The category language the site was originally built to speak to was not what buyers searched for. The content was rebuilt and expanded around the terms the data showed aligned with the client’s product market, and those pages now rank for them.

That is the whole argument for measuring the site’s own content alongside its search data. Without the inventory, a page that ranks for nothing looks the same as a page nobody has linked to, and both look the same as a page aimed at a term with no demand behind it. They need different fixes.

That same client’s site was rebuilt for CRO as well as for content and SEO. Scroll depth and bounce rate come from the analytics data, and isolating them to organic traffic is what turns a general read of site performance into a specific one: it shows where the SEO opportunity and the CRO opportunity actually sit, which are rarely the same pages. Our modernized framework then reduces the friction of changing a page and testing it. Because the GA4 data arrives on a standing schedule with traffic source attached, CRO tests can be judged per source rather than against a sitewide average.

The same loop drives conversion rate optimization, marketing automation and sales automation, where the data decides what to build next rather than what to report on.

Automation and AI, where there is a defined job

AI works in specific, well-defined jobs rather than everywhere at once. In these builds that means process automation, moving data between systems without anybody rekeying it, and agents trained on our rubric for analysis based on years of experience auditing the data to provide weekly insight and ready for our prompts for deeper dives. Our use of AI doesn’t replace our expertise or use AI as expertise, we built agents to free up our time and leverage our expertise.

The same principle of our AI applications runs on the advertising side, where two systems read an account every week: one judges every Google Ads search term not included in ads account negative keywords against a written description of the business and returns the negative keywords to add and why. The other audits Google Ads search ad copy performance to identify the weakest ad in each ad group and writes replacements for it. Both stop short of the account. They produce recommendations, and a person reviews and approves or declines.

My opinion, and I will label it as one: this is where the largest return on AI sits for most businesses right now. Agents that absorb manual and redundant tasks, agents that audit data at a scale nobody can read by hand, and the operational detail it takes to aim them at one specific business, which is a deeper exercise than writing an SOP. I am not claiming it is the only application worth having, and something else may well overtake it. It is the one we can show results for today.

The shiny object of AI communicating via voice or text is cool, we build that where necessary also, wired into the CRM and the phone system rather than sitting beside them, which is the difference between an assistant that can answer a question and one that can book the appointment. See AI chat and voice agents and SMS and email marketing.

The rebuild is not a prerequisite

Worth saying plainly, because the order of this piece implies otherwise: the ads, analytics, and search stack plugs into any site. It does not need to be one we built, or one built on WordPress. The measurement goes onto a site as it stands, and it will usually tell you whether the site is the constraint or something else is, which is a better way to decide on a rebuild than a proposal is.

What a site we build adds is that it arrives already wired in. Analytics, events, structured data and the content inventory are integrated by default rather than retrofitted, which is what makes the optimization work proactive instead of reactive: search, conversion and paid media are read against the same data, so a change in one is measured in the others rather than reported separately. That is a difference in speed and in what can be seen, not a gate on entry.

When a rebuild is worth it

A site that works and is showing its age, a sales process worth measuring end to end, and nobody in-house whose job is to hold the technical and the analytical together. That is the shape where a rebuild pays for itself.

Where it does not, the stack still runs. We work with clients nationally and with businesses across Central Ohio through our Columbus digital marketing services. Whether you are weighing a rebuild or want the measurement pointed at the site you already have, tell us what you are working with.

What Would You Ask Us?

Fifteen minutes, your business, your question. We’ll spend it answering instead of pitching.