How to find freelance clients in the UK, US and EU

Finding freelance clients works the same way in London, Dublin, Austin or Berlin. You start from a public signal rather than from a category, and the register, the job board or the map tells you which companies have a reason to answer you this month.

TL;DR

The UK counted 4.53 million self-employed people in May to July 2026, up on the year before but still well below the 4.94 million of May to July 2019.
The EU had 28.2 million self-employed people in 2025 and 19.2 million of them had no employees, so roughly two in three work solo.
The US counted 9,548,000 people self-employed in unincorporated businesses in August 2026, down from 9,921,000 a year earlier and marginally below the 9,590,000 of August 2019, so the official American series is shrinking while the survey-based freelance counts grow.
Companies House registered 815,277 new UK companies in the year to 31 March 2026, and 5,671,836 US business applications were filed in calendar year 2025, up from 5,224,176 in 2024.
The United States has no national company register, so incorporation data sits in fifty state systems with fifty schemas and fifty rules, and the method changes accordingly.
Indeed's job postings index stood at 101.8 in the US on 14 August 2026 and at 69 in the UK on the same date, so hiring demand in your market decides how well the job board play works.
The two published cold email benchmarks sit almost eight times apart because they count replies differently. Belkins measured a 0.45 percent average reply rate across 7.5 million cold emails sent between January and December 2025, counting unique replies, and 0.72 percent on companies with 0 to 10 employees, while Instantly reported 3.43 percent over the same year, counting every reply received including replies to follow-ups. Belkins is one agency's global aggregate across its own client campaigns, not a rate for the UK, the US or any EU market, and its own published country figures range from 0.48 percent in the United Kingdom to 1.43 percent in Poland.

Which source you query depends on which of four situations you are in. Pick the wrong one and you will build a list of companies that have no reason to answer you this month.

Which of these four situations are you in?

Most freelance prospecting fails because the list was built from a category rather than from a moment. A company that could hire you is not the same thing as a company that has a reason to hire you in September. Each source below carries a different kind of timing signal, and only one of them will match your offer.

Situation	The source to query	What it proves	What it cannot tell you
You have a niche skill	Job boards, LinkedIn Jobs, Indeed, Wellfound	A budget exists for that skill, now	Whether they will work with a contractor
You serve local businesses	Google Maps listings plus the national register, or the state register in the US	The business exists and trades at that address	Who decides, and what they spend
You want fresh accounts	New incorporations at Companies House, the CRO, a member state register such as the BORME or the KVK, a US state register, with Census Business Formation Statistics for timing	Nothing has been bought yet	Whether they can pay yet
You want more clients like your best ones	Register firmographics from each country's own register, or OpenCorporates across several	Which companies look like the ones you already serve	Whether they have a current need

Read the last column before the second. When freelance outreach gets ignored, the reason is almost always something that column already warned you about.

Situation one, you have a niche skill and need companies that are already buying it

Query the job boards. A published job posting is the cleanest purchase signal a freelancer can read for free, because it names a company, a skill, a seniority, a location and an implicit budget, and it carries a date. If a company is recruiting a data engineer, someone has already approved spending on data engineering this quarter and has admitted in public that the work is not getting done.

How well this works depends on where you sell. On 14 August 2026, Indeed's US Job Postings Index stood at 101.8, 1.8 percent above its 1 February 2020 level, though still down 2.9 percent year on year. The UK index sat at 69 on the same date, some 31 percent below the same pre-pandemic baseline. American hiring demand has held roughly flat at pre-pandemic levels while UK demand never recovered, which means a British freelancer working this signal is fishing in a pool a third smaller than an American one.

Check any board's volume claim against the official denominator before you believe it. The US Bureau of Labor Statistics' Job Openings and Labor Turnover Survey recorded 7.3 million job openings on the last business day of July 2026, a rate of 4.4 percent, with hires at 5.1 million and total separations at 5.1 million, all described as little changed on the month in the release of 1 September 2026. JOLTS gives you the size of unmet demand by industry and never a company name, which is the division of labour to expect. The official series tells you whether this play is worth running at all this quarter, and the boards tell you whom to write to.

The angle that works is the gap between the posting and the hire. A permanent role takes weeks to fill and the work does not wait, so a message offering to cover the backlog during the search answers a problem the company has already written down in public. LinkedIn Jobs, Indeed, Wellfound, We Work Remotely and Remote OK all carry this signal, and the last three skew toward startups and remote-first buyers who are more comfortable with contractors.

Three limits matter. There is no open public API here, since Indeed retired its Publisher API and LinkedIn's job data is partner only, so this stays manual or semi-manual work rather than something you automate cleanly. Postings are frequently reposted, stale or duplicated across several boards. And agency-posted roles conceal the end client entirely, so you can spend an hour qualifying a company that turns out to be a recruiter.

Situation two, you serve local businesses in one area

Query the map, then cross-check the register. A restaurant, a clinic, a garage or a shop appears on Google Maps and the Places API with a name, a street address, a phone number, a website, a category, opening hours and review text. The review text is the underused part, because it tells you what customers already complain about, which is usually the thing you are selling.

The register then tells you what sits behind the shopfront. In the UK, Companies House gives the legal name, company number, registered office, SIC codes, incorporation date and directors, with an advanced search that filters on name, address, incorporation date range, status, type and SIC code. In Ireland, CORE does the same and also carries registered business names, so some sole traders appear there in a way they never do on the UK register.

Across the EU, the European e-Justice portal is a lookup rather than a list builder, and treating it as one is the mistake that costs European freelancers a month. You give it a company name or a registration number and a country, and it answers with the name, the registration number, the registered office, the legal form and the status. It covers public and private limited liability companies, their branches and European Companies across the 27 member states plus Iceland, Liechtenstein and Norway, so the smaller legal forms most freelancers sell to fall outside it, and it carries no filter on incorporation date, on activity code or on size. Use it to confirm one entity across a border, use its directory of national registers to find the right register in each member state, and build the list itself in that national register.

Those national registers differ from each other as sharply as the US states do, and the differences decide your method. Germany's Handelsregister carries the commercial, cooperative, partnership and association registers of every federal state, and its content and documents have been free without registration since 1 August 2022, when the transposition of the EU digitalisation directive removed the old per document fees, leaving a one euro charge on microenterprise balance sheets. Its advanced search adds legal form, address and register court to the company name, and the portal publishes the Registerbekanntmachungen, the announcements courts make when an entry goes in, alongside the register itself. The German limit is legal form rather than data, because Freiberufler and Kleingewerbetreibende have no obligation to register at all, so consultants, freelancers and the smallest traders are missing from it the way sole traders are missing from Companies House.

The Dutch KVK Handelsregister searches free by name, address or KVK number, and unlike the German register it holds sole traders, because registration is compulsory for every Dutch business. The public record carries the trade and registered names, the KVK number, the start date, the date of the deed of incorporation, the visiting and postal addresses, the website, the activity, the legal structure, the number of employees and the officials. Telephone numbers are not public, home addresses of officials are not public, and a sole proprietor can have the visiting address shielded without proving a threat, so the smallest records are also the thinnest. Extracts and the KVK APIs are paid products.

Belgium gives more away than either. CBE Public Search is free with no account, in four languages, and searches by enterprise number, name, address or NACE-BEL activity code, returning the status, the date of first registration, the legal form and the establishment units, which are the addresses a business actually trades from rather than its registered seat. Sole traders and public bodies are in it. CBE Open Data then hands over the register itself as CSV files, a full copy plus update files, free after a registration.

Spain splits the work in two. The Registro Mercantil Central charges for most of what it holds, but the BORME, the official gazette of the commercial register, publishes the registered acts including every incorporation, free, every day of the year except Saturdays, Sundays and public holidays in Madrid, and the BOE exposes the summaries through a documented REST API. For Spanish companies the gazette beats the register.

Italy's Registro Imprese, run by the chambers of commerce, searches free by company name, codice fiscale, partita IVA or ATECO activity code and returns the name, the registered office, the legal form and the activity, while directors, shareholders and accounts sit behind the paid visura. Alongside it, INI-PEC is a free national index, consultable without an account, of the certified email address that every registered Italian company and professional is obliged to hold. Read it as a legal and verification channel rather than a prospecting list, because a PEC box is a legal address rather than an inbox and can be set to refuse anything that does not arrive from another PEC account.

France is the outlier in the other direction, and it is why the only full coverage claim further down is a French one. INSEE publishes Sirene, the register of legal units and establishments, as open data under Open Licence 2.0 with an API rate limited to 30 requests a minute for open data use, so the whole French register is machine readable for nothing, which only the Belgian bulk files come close to.

The United States works differently, and the difference is structural rather than cosmetic. There is no national company register and no federal Companies House, because incorporation is a state matter, so you query fifty separate systems with fifty schemas, fifty fee rules and fifty definitions of what is public. Nothing you learn in one state ports to the next. California's bizfile Online is the best case, with free search across corporations, LLCs and limited partnerships, free access to over 17 million filed document images, and, unlike most states, the officer and director names and addresses alongside the agent for service of process, plus bulk orders through the site's BE and UCC bulk orders link. General partnerships and LLPs are not in that search and need a written records request to Sacramento, and a company that never registered to do business in California is simply absent. New York publishes the same register twice, once as an interactive-only search at the Department of State and once as genuine open data, 4,281,406 active corporations on 16 September 2026, downloadable in bulk and queryable without a key, carrying entity name, filing date, county, entity type and a service of process address. Read that address for what it is, a legal contact point that is often an agent rather than the place the business trades, and remember the file holds filed entities only, so sole traders and general partnerships are never in it. Texas charges a statutory one dollar for every single search on SOSDirect, which kills list building on price alone, so use the Comptroller's free taxable entity search instead and accept that it reports tax standing rather than corporate filings.

Delaware is the trap, and it catches everyone who assumes a big register is a rich one. The Division of Corporations counts 2,287,728 entities for 2025, with more than two thirds of the Fortune 500 incorporated there, and its entity search returns the entity name, file number, formation date and the registered agent, and nothing more. No officers, no directors, no managers, no operating address. That agent is a service company shared by tens of thousands of entities, so a Delaware hit proves a company legally exists and tells you nothing about who runs it or where the work happens. The Division also forbids data mining and warns that automated searching can get your access suspended, so there is no bulk route either.

One federal layer exists and it is narrow. SAM.gov is free and public without a login, with entity registrations, exclusions, downloadable extracts and a unique entity identifier per registrant, and the USAspending API then tells you which of those entities actually got paid and how much. If your prospects sell to the federal government, that pairing is the cleanest US company data you will find anywhere. It only contains firms that chose to register for federal work, registrants can opt out of public search, and registrations lapse after 365 days, so treat it as one self-selected segment rather than a national list.

The volume behind this is enormous. Eurostat's preliminary structural business statistics for 2024, published on 9 December 2025, counted around 33.5 million enterprises in the EU business economy, of which 99.0 percent, or 33.2 million, were micro and small enterprises employing up to 49 people. Almost every business a local freelancer can realistically win is in that 99 percent.

Two limits will cost you time. Almost no register gives you an email address, a phone number, a headcount or a revenue figure, the Dutch employee band and the Italian PEC index being the exceptions that show how rare it is, and the registered office is very often the accountant's office rather than anywhere people work. And Google's Places policies forbid pre-fetching, caching or storing Places content beyond the place ID, so building your own copy of the map breaks the terms rather than merely being untidy.

Situation three, you want companies that have not chosen a supplier yet

Query new incorporations. A business in its first months buys almost everything for the first time, an accountant, a website, a brand, an insurance policy, a payment system. Later the supplier list is set, and displacing an incumbent costs three times the effort of being the first choice.

The volume is there in both directions of the Atlantic. Companies House reports that 815,277 new companies were incorporated in the UK during the financial year ending 31 March 2026, up 13,413 or 1.67 percent on the year before, bringing the total register to 5,479,045. In the United States, the Census Bureau counted 5,671,836 business applications in calendar year 2025, up from 5,224,176 in 2024 and 5,469,302 in 2023, of which 1,708,842 were high-propensity applications, the ones the Bureau judges most likely to become businesses with employees. The pace held into 2026, with 4,345,310 applications filed through the first eight months including 1,227,385 high-propensity, and August 2026 coming in at 531,728 applications seasonally adjusted, down 7.8 percent from July, in the release of 11 September 2026. Formation is also concentrated by state for legal reasons rather than economic ones, and Delaware alone recorded 334,461 new entities in 2025, more than fifteen percent up on the 289,810 of 2024, though a Delaware formation tells you nothing about where that business actually operates.

Read those numbers carefully before you build on them. The Companies House total register size counts every company on the register, including those in liquidation or being dissolved, so it is not a count of trading businesses. A US business application is an application for an employer identification number filed on IRS form SS-4, which records a filing rather than a company with a bank account, and this is why the high-propensity subset exists as a separate line. Annual and year to date totals are sums of the not seasonally adjusted monthly series and get revised with every monthly release, so quote them as of a date rather than as a settled fact.

The bigger catch in the American route is that these two sources do different jobs and neither does both. The Business Formation Statistics are an aggregate statistical series with no company names, no addresses and no contacts, so they tell you that a state or a sector is surging and they can never hand you a prospect list. Names come from the state register, and the register you use decides how much you get, which is the California and Delaware gap described above. The practical US sequence is therefore the reverse of the British one. You read the Census series to decide which state and which sector deserve your month, then you open that state's own register to get the entities and, if the state discloses them, the people.

Europe has no single feed of new incorporations either, and the e-Justice portal is no help here at all, since it has no date filter, so you work register by register. Spain is the easiest, because the BORME gazettes incorporations as dated acts on every working day with an open data API behind it. Germany publishes the Registerbekanntmachungen next to the register, which announce entries as the courts make them. Belgium carries a date of first registration on every record, in Public Search and in the open data files alike, so you can cut the CSV by month. The Netherlands shows both the start date and the date of the deed of incorporation on every free record, and France exposes the creation date through the Sirene API. The job is the same in each country and only the interface changes, so the member state you sell into decides whether a month of new companies is one query or an afternoon of clicking.

The practical catch is solvency and legal form. A large share of new registrations are one-person vehicles with no budget for a freelancer, and an entire tier of small buyers never appears at all, since sole traders and most partnerships are absent from the UK register. Filter on legal form and on activity codes where your work is a cost of doing business rather than an optional upgrade. SIC and NAICS codes are self-declared at incorporation and almost never corrected, so use them to exclude rather than to qualify.

Situation four, you want more clients like the ones you already have

Start from your invoices. List your last ten paying clients, then describe in plain words what they have in common, the sector, the size band, the country and the situation they were in when they hired you. That description is your query, and it beats any market report you could buy.

This is where the market data earns its place, because it tells you where your lookalikes are dense and where they are thin. Eurostat counted 28.2 million self-employed people aged 15 to 74 in the EU in 2025, of whom 19.2 million, or 68 percent, had no employees. Roughly two in three self-employed people in the EU work solo with no team behind them, which matters in both directions, because they are your competition if you sell services and a poor prospect if you sell anything priced for a payroll. In the UK, the Labour Force Survey recorded 4.53 million self-employed people in May to July 2026, up from 4.43 million a year earlier but still well below the 4.94 million of May to July 2019, in a series updated on 15 September 2026.

The American figure is the same question measured three ways, and the gap between the three is the thing to understand before you quote any of them. The official monthly count, published in the Employment Situation release from the Current Population Survey, is 9,548,000 people self-employed in unincorporated businesses in August 2026, down from 9,921,000 in August 2025 and 10,082,000 in August 2024, roughly 3.8 percent lower on the year and marginally below the 9,590,000 of August 2019. That series counts main jobs only and excludes anyone who incorporated, which is exactly why the survey counts run an order of magnitude higher. Upwork's Future Workforce Index 2026, published on 14 July 2026 from 2,400 US skilled workers surveyed in March and April, puts skilled freelancers at 38 percent of US knowledge workers, up from 28 percent the year before, with 58 percent of full-time employees saying they are considering freelancing. MBO Partners' fifteenth annual State of Independence study, fielded in April 2025, counts 72.9 million American adults working independently in 2025 against 72.7 million in 2024, of whom 27.6 million work independently full time. Those three numbers do not contradict each other, they answer three different definitions, and lining them up as a trend is how people end up quoting an American freelance market more than seven times larger than the official one.

On the buying side, the SBA Office of Advocacy's February 2026 FAQ counts 36,207,130 US small businesses, of which 29,811,495, or 82.3 percent, are nonemployer firms with no paid staff at all. Only 6,395,635 are employer firms, and those employ 62.3 million people, 45.9 percent of private sector employees. The European lesson repeats here. The nonemployer tier is enormous and mostly cannot fund a freelance engagement, so on an American lookalike list the size band filter does far more work than the sector filter. Note also that these counts are republished from Census statistics for reference year 2022, so they lag by several years and are a composite of two vintages rather than a single-date census.

Density varies far more by country than most freelancers assume. Eurostat's national accounts figures for reference year 2025, extracted on 13 March 2026, put self-employment at 14.2 percent of total employment in persons across the EU and 17.5 percent of hours worked. The share was highest in Greece at 24.8 percent, Bulgaria at 24.6 percent and Italy at 23.2 percent, and lowest in Sweden at 5.2 percent, Denmark at 5.4 percent and Luxembourg at 6.0 percent. A lookalike list built across Europe without weighting for that spread will be four times too heavy in one country and empty in another.

Your rate decides the rest of the filter. freelancermap's Freelancer Study 2026, based on more than 5,400 respondents, puts the average hourly rate in Germany, Austria and Switzerland at 103 euros, slightly below the 104 euros reported for 2025. For the US, the inaugural Future Workforce Index from Upwork, published on 23 April 2025 from 3,000 skilled knowledge workers surveyed between December 2024 and February 2025, reports a median income of $85,000 for people earning exclusively through freelance work, above the $80,000 median for their full-time employee counterparts, and puts 28 percent of US knowledge workers, over 20 million people, in independent work generating $1.5 trillion of earnings in 2024. Both of those are platform surveys with a commercial interest in the answer rather than official statistics, so use them as an order of magnitude and not as a benchmark you price against. If your own rate sits above the median for your trade, your lookalike list has to be built on companies that can absorb it, which is a size band question rather than a sector question. When your clients span several countries, OpenCorporates normalises records from over 140 official registers into one schema, which saves you learning five register interfaces, though its API and bulk feeds are paid and its freshness depends on each upstream register.

What does the prompt actually look like?

Written as a brief rather than as a set of filters. The instruction that works names the segment, the geography, the size band and the signal in one sentence, because the signal is what makes the row worth contacting today.

Three examples that map onto the situations above.

UK companies with 20 to 200 employees that posted a data engineer role in the last 30 days, with the hiring manager and a verified email.
Independent dental practices in Sacramento and Fresno with a Google rating under 4.2 and no online booking on their website.
Companies incorporated in Ireland, the Netherlands or Belgium since June 2026 in architecture or design, outside the capital city.

Each of those reads as one instruction and resolves against a different source, and the three span the three markets. The first hits UK job boards, the second hits maps plus a Californian state register, and the third hits three national registers, CORE in Ireland, the KVK Handelsregister in the Netherlands and the CBE in Belgium, rather than the e-Justice portal, which cannot filter on an incorporation date or an activity. What comes back is a table with the company, the legal entity behind it, the person, the verified contact details and the evidence that matched, which is the posting, the listing or the incorporation date.

We did not run a live search for this article, so no output is reproduced here. The columns above describe the shape of the result rather than a captured run.

Why does the signal matter more than the list size?

Because volume stopped working and the arithmetic is brutal at freelancer scale. Belkins analysed 7,530,489 cold emails sent across its client campaigns between January and December 2025 and reported an average reply rate of 0.45 percent, counting unique replies divided by emails sent and excluding auto-replies and bounces, on a page updated on 26 June 2026. Campaigns aimed at companies with 0 to 10 employees replied at 0.72 percent, which is the best segment in their data and still under one reply per hundred.

Read that 0.45 percent as one agency's global aggregate across its own client campaigns rather than as the rate for your market. Belkins also publishes country figures that run from 0.48 percent in the United Kingdom to 1.43 percent in Poland, so the headline average is a blend that already sits below the figure the same study gives British senders, and it is not the rate for the UK, the US or any single EU country. Use it as an order of magnitude at a solo sender's volume and not as your national benchmark.

The other benchmark you will meet says something very different, and it is worth knowing why before you plan against either. Instantly's Cold Email Benchmark Report 2026, published on 12 January 2026 on sends between 1 January and 18 December 2025, puts the average reply rate at 3.43 percent, the top quartile at 5.5 percent and the top decile above 10.7 percent, which is 7.6 times the Belkins average for the same year. The two are not the same measurement. Both divide by total emails sent, but Belkins counts unique replies net of auto-replies and bounces across one agency's own client campaigns, while Instantly counts every reply received, replies to follow-ups included, across thousands of workspaces on the platform it sells. Belkins also changed its denominator this year, from recipients who opened to total emails sent, and states that a 5 percent reply rate against openers and a 0.45 percent rate against total sends can describe the same campaign. Read the gap as a gap between counting rules rather than between markets, and note that neither vendor breaks its figures out by country.

Run that against a freelancer's week. A hundred generic messages produce under half a reply on the Belkins count and three or four on the Instantly count, so the honest answer is a bracket rather than a forecast, and even the flattering end of it asks for volume no solo freelancer can send without wrecking their sending domain. Plan against the low end and treat anything above it as upside. Volume cannot rescue those odds at a solo sender's scale. Thirty messages a week where the first line quotes a job posting the company published last Tuesday, a review left on their listing, or the fact that they incorporated in June and still have no website will beat three hundred generic ones.

That is also why the last column of the table matters more than the second. A source that proves timing gives you something to write. A source that proves existence gives you a name and a blank page.

Where Freelvy sits in this conversation

Freelvy is our product, an AI sales platform. It works the way ChatGPT or Claude does. You describe the customer you want in one prompt, and the agent searches 40+ live sources at query time, company registers, maps, job boards, storefronts and ad libraries, rather than filtering a stored database. It then enriches every row with verified contact data, six providers in cascade for emails and nine for phone numbers, and writes and sequences the outreach from the same brief.

One clarification on coverage, because it decides whether this is useful to you. Freelvy covers the French company register in full. It does not have complete coverage of any other national register, so if you are in the UK, Ireland, the US, Canada or elsewhere in the EU, treat Companies House, the CRO, your own state register, or the national register of your member state, the Handelsregister, the KVK, the CBE, the BORME, the Registro Imprese and the rest, as public sources you query yourself, exactly as described above. What Freelvy does across those markets is search registers, maps, job boards and storefronts live and return rows with the evidence that matched, so you can write the first line from something the company published. The three prompts above are written the way the agent expects them. Plans start at €49 a month. There is a free 7-day trial, no credit card, at freelvy.com.

How we verified these numbers

Self-employment counts come from the ONS Labour Force Survey series MGRQ for the UK, from Eurostat datasets for the EU, and for the US from the Bureau of Labor Statistics Current Population Survey series LNS12027714, published in table A-9 of the Employment Situation for August 2026 released on 4 September 2026, all linked inline and read on 16 September 2026. US job openings come from the JOLTS release of 1 September 2026 for the July 2026 reference month, where the Bureau describes openings, hires and separations as little changed, so none of the three is presented here as a move. Incorporation and business application figures come from the Companies House statistical release of 25 June 2026, the Delaware Division of Corporations statistics page for calendar 2025, and the US Census Bureau Business Formation Statistics, with the annual and year to date totals summed from the unadjusted monthly series, which the Census Bureau revises with each monthly release, and the August 2026 figure seasonally adjusted as published. Job posting indices come from Indeed Hiring Lab's own UK and US snapshots of August 2026, both indexed to 1 February 2020 and seasonally adjusted on a seven-day trailing average. The US business counts come from the SBA Office of Advocacy FAQ of February 2026, which republishes Census Statistics of US Businesses and Nonemployer Statistics for reference year 2022, so it inherits their lag and mixes two vintages. The hourly rate figure is a survey result from freelancermap, the US freelance share and income figures are Upwork's own survey research, the 72.9 million independents figure is MBO Partners' vendor survey, and the reply rate figures are a vendor's measurement of its own client campaigns, taken from that vendor's worldwide aggregate rather than from any per-country rate, which is why the country spread it publishes is quoted alongside it, so all four are attributed in the text and none is presented as an official statistic. The Upwork and MBO definitions differ from each other and from the BLS measure, so they are stated side by side and never as a single trend line. What each public source gives and withholds was checked against the source's own documentation, including the search criteria, the covered legal forms and the free fields of the e-Justice portal and of the German, Dutch, Belgian, Spanish, Italian and French registers, all read on 16 September 2026, with the German fee removal of 1 August 2022 taken from the Commission's own country page rather than from the register. No search was run in the product for this article, and no output is reproduced.

FAQ

Where can a freelancer find company data for free?

In the national registers, except in the United States where no national register exists. Companies House covers every UK limited company with filters on SIC code, location and incorporation date. The CRO's CORE covers Irish companies and registered business names. The European e-Justice portal looks one company up by name or number across the member states, but it holds only the larger legal forms and has no date, activity or size filter, so the list building happens in the national registers, which are not equivalent to each other either. Germany's Handelsregister has been free without registration since 1 August 2022 and leaves out Freiberufler and small traders entirely, the Dutch KVK includes sole traders and publishes an employee band, Belgium's CBE searches free on NACE-BEL activity codes and also publishes the whole register as open data, Spain's BORME gazettes every incorporation on each working day with an API, Italy's Registro Imprese is free for the basics with the INI-PEC index of certified addresses beside it, and France publishes Sirene as open data with an API. US company data lives in fifty separate state systems that are not equivalent to each other. California's bizfile Online is free and discloses officers and directors, New York publishes over four million entities as bulk open data, Delaware returns only the registered agent and forbids data mining, and Texas charges a dollar per search on SOSDirect although the Comptroller's taxable entity search is free. SAM.gov covers only entities registered for federal contracting, and the Census Business Formation Statistics are aggregate counts with no names. Almost none of them gives you an email address or a phone number, Italy's INI-PEC being the rare exception, and a PEC address is a legal mailbox rather than a prospecting channel.

Are job postings a good prospecting signal for freelancers?

They are the strongest free signal available, because a posting proves a budget was approved for a named skill at a named company on a known date. The catch is market dependent. Indeed's index stood at 101.8 in the US on 14 August 2026 against 69 in the UK, so the same tactic yields very differently on either side of the Atlantic. A posting also never proves that the company works with contractors at all.

How many self-employed people are there in the UK, the US and the EU?

The UK had 4.53 million in May to July 2026 according to the ONS Labour Force Survey, up from 4.43 million a year earlier but below the 4.94 million of May to July 2019. The EU had 28.2 million aged 15 to 74 in 2025, and 19.2 million of them had no employees. Self-employment made up 14.2 percent of EU employment in persons in 2025. The official US count was 9,548,000 self-employed in unincorporated businesses in August 2026, down from 9,921,000 a year earlier, but that series counts main jobs only and excludes anyone who incorporated, which is why survey counts from Upwork and MBO Partners run several times higher on a much wider definition. Compare like with like or you will size the American market wrongly by a factor of seven.

What should I charge as a freelancer?

Benchmark against your own market rather than a global average. freelancermap's 2026 study of more than 5,400 freelancers puts the average hourly rate in Germany, Austria and Switzerland at 103 euros, down from 104 euros in 2025. Rates vary more by seniority, country and delivery risk than by discipline, and your rate should drive which size band you prospect, since a five-person company cannot absorb a senior European day rate.

How many prospects does a freelancer actually need?

Far fewer than most lists contain, provided each row carries a reason to write. The published benchmarks bracket the answer instead of settling it. At the 0.45 percent average reply rate Belkins measured across 7.5 million emails in 2025, counting unique replies across one agency's campaigns, a hundred generic messages produce under half a reply. At the 3.43 percent Instantly reported for the same year, counting every reply including replies to follow-ups across its whole platform, the same hundred produce three or four. Plan against the low end and treat the rest as upside. Belkins' best segment, companies with 0 to 10 employees, replied at 0.72 percent. That 0.45 percent is one agency's global aggregate across its own client campaigns rather than a national benchmark, and Belkins' own country figures range from 0.48 percent in the United Kingdom to 1.43 percent in Poland, so read it as an order of magnitude at a solo sender's volume rather than as your market's rate. Thirty companies a week picked on a dated signal is a better use of your time than a file of a thousand rows.

Is it legal to cold email companies in the UK, the EU and the US?

The rules differ and you need to know which one applies. In the UK, PECR treats corporate subscribers differently from individuals, and business to business email is permitted with a clear identity and an opt-out. In the EU, B2B prospecting generally relies on legitimate interest under the GDPR, though several member states apply stricter national rules. In the US, CAN-SPAM allows cold commercial email provided you identify yourself, give a postal address and honour opt-outs promptly. Sole traders and individual freelancers often count as individuals rather than businesses, which tightens the rules considerably. Keep suppression lists, honour removals immediately and write to work addresses only.

Should I use a public register or a contact database?

They answer different questions. A register tells you which legal entities exist, when they were created and who the directors are, at no cost and with no contact details. A contact database sells you the person, at a price, with the decay every database carries. Our guide on verifying a B2B email list covers the second problem, and the cold email deliverability checklist covers what happens when you send to a list you did not check.

Sources and methodology

All pages below were consulted on 16 September 2026. Official statistics are stated directly with their source. Survey and vendor figures are attributed in the text. The two cold email benchmarks quoted, Belkins and Instantly, apply different counting rules to the same year, and the text explains the gap rather than averaging it away.

Read next: how to verify a B2B email list before you send to any of this, the cold email deliverability checklist, and the best AI prospecting tools of 2026.