Flying blind: 59.8% of established UK firms selling online run no analytics

Firmfeed research, July 2026. We took every active UK company incorporated twenty or more years ago under a digital, creative or IT SIC code — 23,542 of them — resolved their websites, and looked at what those sites actually run. This page is the whole finding: the numbers, how they were produced, and where the method is weak. No email required, no form, nothing gated.

Cohort and method

Companies House publishes no website field, so domains were resolved from registered company names, then each live site was fetched and parsed once in early July 2026 against the 1 July 2026 bulk register. Every stage of that funnel loses companies, and the losses are not random — so here is the funnel before anything else.

Why we threw away half the scan

The easy version of this report uses all 7,390 scanned sites and leads with “26.3% of UK digital firms run analytics”. That number is wrong in an interesting way. Detection rates move sharply with how confident we are that the site belongs to the company:

A metric that swings from 19.3% to 38.8% with match quality is measuring our matching as much as the market's behaviour, so everything below is counted on confirmed and probable matches only. The headline survives the tightening in both directions — on the full scanned set 64.1% of e-commerce sites lack GA4, on the tightened cohort 59.8% — which is what makes it a finding rather than an artefact of where we drew the line.

What the verified cohort runs

The Meta pixel figure is the one that surprised us — roughly 37 to 1 against. Paid-social retargeting is effectively absent from established UK B2B digital firms, which quietly kills any campaign premised on finding them there.

Measurement maturity tracks the CMS, hard

A hand-built site is a little over half as likely to carry analytics as a WordPress one, and a Wix site is 5.3x less likely than WordPress and 3.0x less likely than a custom build. The plausible reading is that platforms with a plugin ecosystem make measurement a checkbox, while custom builds make it a ticket nobody raises after launch. Whatever the cause, the CMS is a usable proxy for whether a company measures anything at all — and it is a segmentation axis almost nobody sells on.

Hiring scales with size, steeply

Measured across the full scanned set (n = 7,390) for sample size. Two thirds of large established firms are hiring at any given moment against one in eight micro firms — a 5.3x spread, and the sharpest size discriminator among the signals we collect.

The e-commerce stack is a WordPress market

Among the 1,548 sites with e-commerce signals in the full scanned set, the startup world's Shopify default barely registers:

In a twenty-year-old UK company cohort this is a WordPress market with a shop bolted on, not a Shopify one. Named payment providers appear on the surface of almost none of them — which says more about what is detectable from a homepage than about how these firms take money.

Geography barely matters

GA4 adoption sits in a 20–30% band across every English county with a usable sample — Hertfordshire 30.2% at the top, Buckinghamshire 20.0% at the bottom, on n=126 and n=95. 58% of the register's county field is blank. Location is a weak targeting variable in this sector. If you are segmenting UK digital firms, platform and size will separate them; postcode will not.

Caveats, stated plainly

• Discovery bias is the big one. Domains were resolved from registered company names, so the cohort skews toward firms whose trading name matches their registered name. A twenty-year-old agency that rebranded is under-represented. This is a free-tier method and it shows.
• GA4 detection is HTML-based. Server-side tagging or a proxied container reads as absent, so 33.6% is a floor, not a ceiling — which makes the flying-blind segment a conservative estimate rather than an inflated one.
• One scan window, early July 2026. No trend, no seasonality, no re-check.
• Company-level data only. No personal names, no personal email addresses, no scraped contacts — public registry fields and public website URLs, by design.
• Absence of a signal is evidence about a homepage, not about a business. A firm may run analytics we cannot see from the outside, and no company is named here for exactly that reason.

How to check any of this

Every figure on this page is counted from a single committed export of the scan and reproduced by one script — scripts/sector-signals-report.py — rather than typed in from a draft. Nothing is sampled, weighted or estimated. If you want to argue with a number, the useful question is whether the 3,197-company cohort represents the sector, not whether the arithmetic holds, and the caveats above are our own best case against it.

Related reading: how to get a list of UK companies by the technology their site runs, by SIC code or industry, or with websites attached. The full UK company data providers guide compares every route to this data end to end.