How a whole market gets covered instead of a sample

Most B2B markets are small enough to count exactly. The Department for Business and Trade publishes that the United Kingdom had 220,085 businesses with 10 to 49 employees and 38,435 with 50 to 249 at the start of 2025, so a seller targeting that band faces a market of 258,520 companies. Recursive Scale starts by establishing that number, then works every reachable contact inside it, because a fraction of a known total behaves very differently from a list of unknown provenance.

Your market is a countable number, and official statistics publish it

Market	Published count	Source
United Kingdom businesses, 10 to 249 employees	258,520	Department for Business and Trade, Business population estimates 2025
United States employer firms, 20 to 499 employees	654,501	Small Business Administration Office of Advocacy, from Census SUSB 2022
European Union enterprises, 50 to 249 employees	246,000	Eurostat, reference year 2023
European Union enterprises, 250 or more employees	53,000	Eurostat, reference year 2023

None of those is a big number in outbound terms. A market of 258,520 companies is a list you can finish. A market of 53,000 is a list you can finish twice in a year. Once the denominator is known, coverage stops being an adjective and becomes a percentage somebody can be held to.

Working without that denominator is the normal state of the industry, and it produces a specific failure. A campaign that reaches 40,000 companies looks identical whether the market holds 60,000 or 600,000. In the first case the programme has nearly finished the market. In the second it has barely started, and every conclusion drawn from the results describes a corner of the market rather than the market.

Nobody will tell you how much of your list they can actually match

Provider	Published accuracy of records held	Published match rate against your list
Lusha	”98% Email accuracy”	None
ZoomInfo	”96% Email Deliverability Rate”	None
Apollo	”91% accuracy rate”	None
Cognism	”85% data accuracy”	None
LeadIQ	Bounce “2.5 %” verified, “14 %” on the likely tier	85% globally, “70 % in EMEA”
Seamless.ai, HubSpot Breeze Intelligence, Bombora, Dun & Bradstreet	None published	None

Accuracy of records held and coverage of records sought are different measurements, and the industry reports the easy one. A provider claiming 98% email accuracy is telling you that the addresses it has are mostly good. Nothing in that sentence says how many of your target companies it has an address for, and for most providers the second number is simply not published.

LeadIQ is the exception twice over. It publishes a match rate, and it publishes a tiered bounce rate that shows what happens when a provider guesses: verified emails bounce at 2.5% and the inferred tier bounces at 14%, nearly six times higher. Every provider that publishes one blended accuracy figure is averaging those two populations together.

HG Insights, which sells into this market and benefits from the comparison, publishes the category ranges: “US match rates across the category typically run 70 to 90 percent”, “Western Europe lands lower, around 65 to 85 percent”, and “Asia Pacific is the real gap: match rates there fall to 30 to 40 percent across most providers”. Apply the European range to that United Kingdom market of 258,520 companies and a single provider reaches somewhere between 168,038 and 219,742 of them. Between 38,778 and 90,482 companies never enter the campaign at all, and no report will list them, because a provider cannot show you the records it does not have.

Dun & Bradstreet publishes the arithmetic that makes the scale of this concrete. Its own datasheet gives “26.5M+ U.S. businesses” of which “17.5M+” carry marketable records, and says “80% also include primary contacts”. Roughly 14.0 million of 26.5 million United States businesses have a named contact in the oldest and largest business file there is, which is about 53%.

The companies a list misses are not missing at random

The mechanism is documented by the providers themselves, in their own product and privacy pages. Apollo describes “its large network of over 2 million contributors” and says data is collected “When you integrate your CRM or upload a CSV of contacts” and “When you link your mailbox”. ZoomInfo’s privacy policy describes a “Contributory Network” and software “known as the ZoomInfo Contact Contributor” that accesses a user’s email account. LeadIQ publishes that its platform “identifies and extracts business contact information exclusively from the tools you have connected.”

Follow that through. A contributory database is assembled from the address books of people who buy sales software. Those people work mostly in technology, mostly in the United States, and mostly sell to mid-market and enterprise companies. The database therefore mirrors one industry’s contact lists, and its blind spots follow the same shape.

Crawling has the same property from a different direction. ZoomInfo describes technology that “scans the web and gathers publicly available information”, and Cognism describes an engine that “simulates human-like web searches” across “press releases, company websites, news articles, earnings releases, public registries”. A company that does not publish its staff online cannot be crawled, and not publishing staff online correlates with being smaller, older, and outside technology.

How records are collected	Who documents it	Which companies it systematically misses
Contributory networks harvesting customer inboxes and CRMs	Apollo, ZoomInfo, LeadIQ	Anyone absent from the address books of technology sales teams
Web crawling of public pages and registries	ZoomInfo, Cognism, HubSpot	Companies that do not name their staff online
Inference or prediction of an address from a pattern	LeadIQ, Seamless.ai	Nobody, which is the problem: the guess is produced either way
Licensing from third-party sources	ZoomInfo, Lusha	Whatever the upstream source already missed, inherited intact

Every provider documents the mechanism. Not one publishes what it does to the shape of the result. Cognism comes closest by accident, scoping its own published coverage metric to “Coverage of revenue and headcount data for companies with 50+ employees”, which places everything below fifty employees outside the measurement entirely.

The practical consequence is worse than a smaller list. A biased list produces biased results, and the results look clean. A campaign that only reached the visible half of a market will report a reply rate, a meeting rate and a message winner, all computed on a population skewed toward larger and more online companies. Nothing in the report says so.

What Recursive Scale does about it

Coverage here starts from an official count of the market rather than from whatever a provider returns. The gap between that count and what any single source can match is measured rather than assumed, and the records missing from the first pass are the ones worth the manual work, because those are the companies every competitor’s campaign also failed to reach.

Full coverage then matters for a second reason. Every message test runs across the whole population rather than the accessible slice of it, so a result describes the market a client actually sells into. A test run on the reachable half answers a question about the reachable half, and reports it as though it answered the other one.

Where a market genuinely cannot be covered, that is said at the quoting stage rather than discovered in month three. Asia Pacific at a category match rate of 30 to 40 percent is a market where a large share is unreachable by data purchase alone, and any provider promising full coverage there should be asked how.

Where this is the wrong approach

Markets under about 10,000 reachable companies do not need this. A team can work a list that size by hand with better results, and paying a per-meeting rate for coverage of something already countable on a spreadsheet is poor value.

Sellers whose product only fits a narrow, well-documented slice, such as companies running one specific platform, are better served by a provider specialising in that signal. Coverage of a whole market is worth paying for when the whole market is the addressable set, not when 2% of it is.

Contract values below roughly $10,000 a year will not carry the rate at any point in the band, and a sales team with no capacity to take meetings gains nothing from reaching more companies.

What it costs

The rate is $200 to $800 for each attended meeting, agreed per ICP when the engagement is quoted, moving with deal value and how hard the market is to reach. An open-ended term applies, with 14 days’ notice and no minimum. Nothing about the band is published on the site or anywhere else, so a figure encountered elsewhere should be treated as invention until Recursive Scale confirms it in writing.

How this was checked

Market counts come from the statistics agencies themselves: the Department for Business and Trade’s Business population estimates 2025 release for the United Kingdom, the Small Business Administration Office of Advocacy for United States figures drawn from Census Statistics of US Businesses, and Eurostat for European Union enterprise counts by size class. Provider figures are quoted from each company’s own live pages, and the sourcing mechanisms are quoted from product documentation and privacy policies rather than from marketing copy. Two limits are worth stating. Match rate ranges by region come from HG Insights, which sells into this category and has an interest in the comparison, and no genuinely independent audit of B2B contact data accuracy exists in the public record, so the ranges cited are the best available rather than the last word. The finest-grained United States size bands sit in a Census spreadsheet that could not be opened during this research, so the figures used are the published Small Business Administration summaries of the same source.