Become a partner
Insights 3 August 2026

This Week in UK Filings: The Most Common SIC Code Tells You the Least

82990 is the single most common SIC code across our 650 tracked UK companies. Here's why that's a data-quality problem, not a sector story.

Every week we pull the SIC codes attached to the 650 UK companies Archive Partners tracks, and every week the same pattern holds: construction and holding-company activity dominate the list. This week is no exception — combined, the four core construction codes (development of building projects, domestic and commercial construction, and other specialised construction activities) account for 77 tagged instances across our tracked population, and “activities of other holding companies” (SIC 64209) adds another 28. Construction and corporate-structure filings remain, as ever, the backbone of what shows up in Companies House data for firms of this size.

But this week’s most interesting number isn’t the construction total. It’s the single most frequently occurring SIC code in the entire dataset: 82990, “other business support service activities not elsewhere classified.” Thirty of our 650 tracked companies carry it — more than any construction code, more than holding companies, more than IT services. And 24 of those 30 carry it as their only SIC code.

That second number is the real story. A SIC code is supposed to tell you what a company does. When nearly a fifth of the companies wearing this particular label have no second code to add nuance — no software development tag, no professional-services tag, no sector-specific classification alongside it — the code isn’t describing a business activity so much as recording the absence of a better one. “Other… not elsewhere classified” is Companies House’s catch-all, and for 24 firms in our tracked population, it’s doing all the classification work by itself.

This matters more than it might look like on the surface. Procurement teams, credit analysts, and compliance functions increasingly use SIC codes as a first-pass filter — to scope a supplier risk review to “construction firms” or “IT services vendors,” for instance. A code that’s genuinely uninformative breaks that filter quietly: the company doesn’t fail the screen, it just returns a null result dressed up as data. You can’t distinguish, from the code alone, a genuine multi-service business-support firm from a company that was simply never re-classified after its actual activity changed. Both show up identically as “82990, single code.”

The financial numbers underline the gap between label and substance. Companies tagged 82990 average £22.6m in turnover — solidly mid-market, not shell entities. Holding companies (64209) average £57.9m, more than double, despite being the second-most common code rather than the first. That inversion is worth sitting with: the code that appears most often across our tracked population is not the one attached to the most economically significant firms. Frequency of classification and weight of economic activity are two different rankings, and treating a SIC-code count as a proxy for “how much of the economy this sector represents” — a mistake it’s easy to make when skimming a frequency table — get the story backwards if you don’t also weight by turnover.

IT-related codes tell a similar story from a different angle. Split across “other IT service activities” (62090, 13 firms, £37.1m average turnover), “business and domestic software development” (62012, 11 firms, £26.4m), and “IT consultancy activities” (62020, 10 firms, but only £9.9m average) — the sector is real and reasonably well-represented, but it’s fragmented across three adjacent codes rather than concentrated in one. A procurement screen that filters on a single IT SIC code will systematically undercount the sector; a screen that pools all three gets closer to the true picture, which is itself a small lesson in why single-code filtering is a blunt instrument.

None of this means SIC data is useless — the construction and holding-company totals are directionally real and consistent week to week, and they track sensibly against what we’d expect from a UK small-to-mid-cap population skewed toward property, infrastructure and corporate-group structures. But this week’s filings are a good reminder that the most common code in a dataset is sometimes the least informative one in it, and that “not elsewhere classified” is doing more quiet work in UK company data than most users of that data realise.