Research

What we measured

Most published work on ad measurement is survey work: marketers are asked what they believe is installed on their websites, and the answers are totalled up. That produces a picture of what an industry thinks about itself. We wanted the other thing - what is actually running on the page when a customer arrives on it.

So the question we set out to answer was narrow and checkable. For a home-services business that is spending money on ads right now, is there any mechanism on its website capable of recording that a lead happened? Not whether the mechanism is configured well, not whether the numbers it produces are trusted internally - simply whether one exists at all. That is the floor beneath every other measurement question, and it turns out to be missing far more often than the industry assumes.

How we measured it

As of the September 6, 2026 report, we had loaded 7,331 home-services business websites in a real headless browser, let each page execute its JavaScript, and inspected the rendered DOM for tracking signals. The rendered part is the whole point. A raw HTML fetch cannot see anything a tag manager injects at runtime, so a fetch-based scan would report a far higher share of untracked sites than really exists. Detection is deterministic pattern-matching rather than a model, which means the same site returns the same verdict every time it is scanned and the dataset can be regrown without the earlier rows drifting.

A site is counted as untracked only when the browser session finds none of the following: a Google Tag Manager container, GA4, a Google Ads conversion tag, a Meta pixel, Microsoft UET, or a call-tracking script. A single one of those is enough to move a site out of the untracked group. Separately, advertising status was checked against Google's Ads Transparency Center, so "confirmed advertiser" means live ads were visible there at the time of the scan rather than inferred from the site itself.

What we found

Of the 7,331 businesses loaded, 1,619 were confirmed to be running ads. Within that group, 462 - 29% - had no conversion tracking of any kind. Not misconfigured, not partially firing: nothing on the page capable of recording a lead. Across the full sample, tracked or advertising or not, 4,831 of 7,331 sites - 66% - had nothing installed.

The individual platform numbers say the same thing from another direction. GA4 was present on 26% of all sites, a Google Ads conversion tag on 21%, and Google Tag Manager on 18%. Those are the tools the entire optimisation apparatus of paid search assumes are in place.

These figures are from the September 6, 2026 edition of the report. The scan runs continuously, so the current totals are higher - see the full report for the latest counts.

The rate is also not evenly spread. Broken out by trade, the top of the table sits near a coin flip and the bottom sits in the low teens, which is a wider spread than most operators would guess about their own vertical. The report publishes every trade with at least 15 confirmed advertisers, alongside the advertiser count for each row, so a reader can see which rates rest on a thin sample before quoting one.

What it means for an advertiser

An advertiser with no conversion tracking is not measuring badly. They are not measuring at all. Every decision in that account - which keyword to raise, which campaign to cut, whether the channel is worth running - is being made on clicks. Clicks and jobs are not the same thing, and the gap between them is where budgets disappear. A campaign producing expensive clicks and cheap jobs reads as a failure. A campaign producing cheap clicks and no jobs reads as a success, and can run that way for years before anyone catches it.

The practical implication is not that a business should install more tags. It is that nothing downstream - bid strategy, budget allocation, creative testing, agency reporting - carries any weight until this layer exists and is verified to fire. The cheapest available improvement in most accounts we inspect is not a new campaign; it is making the existing measurement true.

Limitations of the method

The 29% is a ceiling rather than an exact figure, and it is almost certainly slightly too high. Server-side tagging is largely invisible to a page-level scan - a business running a server-side container can be fully tracked while a scan sees nothing. Consent-gated tags may never fire for an automated visitor that does not click accept. And a business can record leads perfectly well with a phone log and a notepad; that is a measurement system, just not one a campaign can be optimised against.

The sample is also not a national random sample. It is a large, systematically collected set of home-services businesses drawn from metropolitan areas in Arizona and California, and it should be read as that. Absence of ads in the Ads Transparency Center is not proof a business is not advertising, so businesses we could not confirm are excluded from the advertiser group entirely rather than counted as non-advertisers - which keeps the denominator honest at the cost of making it smaller.

All of the above is published on the report itself, with the full trade table and the exact detection criteria. The page is updated in place as the dataset grows, so cite the URL rather than a figure in isolation.

Questions about the method