Why do closed companies stay in B2B databases?

Because loading data and re-checking it are two separate jobs, and typically only the first one gets built. A record is written with the status active and nothing ever goes back to ask the register whether that company still exists. Closed companies keep their name, domain, industry code and often a phone number, so they pass automated quality filters and look identical to real leads. In one audit 50,968 of 289,225 closed companies had a working domain attached. The only reliable test is the register's own status field.

What does a closed company in a prospecting list actually cost?

In this audit one cold outreach batch consisted of 54.5% closed companies. Those addresses bounce, and bounce rate is the fastest way to damage a sending domain's reputation, which then suppresses deliverability for the messages that would have landed. A caller ID index built from the same data attached 268,242 phone numbers to companies that no longer exist, on numbers frequently reassigned to other businesses or private people.

How does AtlasForgeX handle closed companies now?

A nightly job re-reads each country's current official register, marks every stored record that is no longer present as terminated, and restores any record that reappears. Company search, industry segment counts, the caller ID index and outreach lists all read that single status field, so a correction propagates everywhere at once. If a register download returns implausibly few companies the job aborts without writing, so a truncated file cannot mark an entire country as closed. The first run across seven countries marked 844,549 companies as closed.

// Data study · July 2026

We Found 289,225 Dead Companies in Our Own Database

Q: How many companies in a business register are closed?

Measured on 22 July 2026 directly from the official open data files: 70.6% of the 2,102,483 entities in Singapore's ACRA datasets are not live, 60.5% of the 2,857,430 organisations in Sweden's Bolagsverket register carry a deregistration date, 43.9% of the 29,803,585 legal units in France's INSEE Sirene file have the administrative state cessee, and 43.9% of the 821,699 companies returned by Finland's PRH company endpoint are no longer on the register. National company registers are cumulative archives rather than lists of trading companies, so every company that ever closed stays in them permanently, marked by a status field.

Q: How can I check whether my B2B list is full of closed companies?

Take 100 random companies from the list, look each one up in its national register by company number, and count how many come back closed. The free primary sources are PRH for Finland, Bolagsverket for Sweden, Bronnoysund for Norway, Companies House for the United Kingdom, INSEE Sirene for France, Ariregister for Estonia and data.gov.sg for Singapore. Match on the company registration number rather than the domain name: domain matching produces false positives, because a struck-off company and a trading one can share a domain.

Q: Does this affect Apollo, ZoomInfo or other B2B databases?

This audit covered one vendor's own database and the national registers, not competitors' records, so no claim is made about their specific numbers. The mechanism, however, is not vendor-specific: any pipeline that loads a national company register without continuously re-reading its status field will accumulate closed companies over time, and the registers themselves show that between 44% and 71% of the records they contain are closed. The way to find out about any vendor is to test a sample of their list against the register directly.

Then we audited six more countries, and the national registers themselves. In France's official register, 43.9% of all legal units are closed. In Sweden's, 60.5%. This is the whole audit, including what it cost us.

Published 22 July 2026Free to read · free to cite

289,225

companies in our own Finnish database that had already been struck off the register, and were still stored as active

38.6% of 749,497 records, measured 22 July 2026

This is not a competitor teardown. It is our own data-quality audit, published with the numbers that make us look bad, because the mechanism behind them is industry-wide and almost nobody measures it.

Here is what happened. Our Finnish company data was built from the national register, PRH, which is the correct primary source. The ingest wrote every record with the status active, and nothing in the pipeline ever went back to ask whether a company had since closed. Records only ever went in. Nothing ever came out. After a year of that, more than a third of the database was companies that no longer exist.

50,968

of those dead companies had a working domain name attached, so they passed every enrichment and quality filter we had, and looked exactly like real leads

this is why the problem survives quality checks

That second number is the one worth sitting with. A dead company with a domain does not look dead. It has a name, a website, an industry code, often a phone number that still rings somewhere. Every automated quality check we ran said it was fine. The only thing that could tell us otherwise was the register itself, and we were not asking it.

01 //Why this happens to everyone

National company registers are cumulative archives, not lists of currently trading companies. They keep every company ever registered, with a status field marking the ones that closed. If you read a register and ignore that field, you get the graveyard along with the living, and the graveyard is usually bigger than people expect.

We measured exactly how much bigger. These are the official registers themselves, downloaded on 22 July 2026 and counted record by record:

Official register	Records it returns	Closed or removed
🇸🇪 Sweden, Bolagsverket	2,857,430	60.5%
🇫🇷 France, INSEE Sirene	29,803,585	43.9%
🇫🇮 Finland, PRH	821,699	43.9%
🇸🇬 Singapore, ACRA	2,102,483	70.6%

In France, 13,080,910 of the 29,803,585 legal units in the Sirene file carry the administrative state cessée. In Sweden, 1,728,033 of 2,857,430 organisations have a deregistration date. In Finland, the register's own company endpoint returns 821,699 companies, while the file of currently registered companies contains 461,107. The gap, 360,592 companies, is the graveyard.

Singapore is the extreme case. ACRA's own entity datasets contain 2,102,483 entities, and only 617,218 of them are Live. Seven out of ten entities in Singapore's register are struck off, terminated, cancelled or dissolved.

Build a list from the obvious endpoint without reading the status field, and roughly four out of ten Finnish or French companies on it have already closed. In Sweden it is six out of ten, in Singapore seven.

// Not every register even shows you the graveyard

The UK's Companies House monthly file and Norway's Brønnøysund download contain only companies currently on the register. Dissolved companies are simply absent. That sounds safer, and in one way it is, but it creates the opposite failure: if you loaded the file once and never reloaded it, your copy silently ages, and nothing in the data tells you which records went stale.

Estonia publishes a third shape again. Its open data file lists 374,810 current entities, of which 9,042 are in liquidation and 662 in bankruptcy. Those companies are legally alive and commercially finished, which is a distinction most data pipelines never make.

02 //What it cost us downstream

The dead records did not stay in the database. They flowed into everything built on top of it, and each of those had its own multiplier:

Where it surfaced	What we found	Share
Company search results	289,225 dead companies served as active	38.6%
Caller ID index	268,242 phone numbers attached to closed companies	not applicable
One cold outreach batch	545 of 1,000 companies no longer existed	54.5%
Industry segment counts	Retail shown as 37,616 companies, actually 24,711	34.3% inflated

The outreach line is the expensive one. A cold email to a company that closed in 2019 does not just fail, it bounces, and bounces are the single fastest way to burn a sending domain's reputation. We had a batch running at 54.5% dead. Every marketing metric would have called that a copywriting problem.

The caller ID line is the one that damages trust. In Finland a disconnected business number gets reassigned, often to a private person. An index built on stale register data will confidently name a company that has not existed for a decade.

03 //How to check your own list in ten minutes

You do not need our tooling to do this, and we would rather you checked than took our word for it. Take 100 random companies from whatever list you are about to use, look each one up in its national register by company number, and count how many come back closed. If the answer is more than two or three, you have the same problem we had.

These are the primary sources, all free and all public:

Country	Source	The field that matters
🇫🇮 Finland	PRH avoindata, YTJ API v3	companyForms endDate
🇸🇪 Sweden	Bolagsverket bulk file	avregistreringsdatum
🇳🇴 Norway	Brønnøysund enhetsregisteret	slettedato
🇬🇧 United Kingdom	Companies House basic data	CompanyStatus
🇫🇷 France	INSEE Sirene, StockUniteLegale	etatAdministratifUniteLegale
🇪🇪 Estonia	Äriregister avaandmed	ettevotja_staatus_tekstina
🇸🇬 Singapore	data.gov.sg, ACRA entities	entity_status_description

One warning from our own audit, because it cost us a wrong answer at first. Matching by domain name alone produces false positives. We flagged one Finnish company as dead because its domain matched a struck-off record, and the company turned out to be trading normally under a different registration. Match on the company number, and if you only have a domain, require a second signal before you delete anything.

04 //What we changed

The fix is not clever, which is rather the point. Every night the pipeline downloads the current register for each country, marks every record that is no longer in it as terminated, and puts back anything that reappears. Search results, segment counts, the caller ID index and the outreach lists all read the same status field, so they all cleared at once.

Across seven countries the first run marked 844,549 companies as closed: 289,225 in Finland, 506,015 in France, 25,921 in the United Kingdom, 7,965 in Singapore, 6,968 in Norway, 6,792 in Sweden and 1,683 in Estonia. Finland was the disaster because that pipeline was oldest. The others were mostly accumulated drift, which is what every database gets if nobody checks.

There is a safety rule worth copying if you build this yourself: if the register download returns implausibly few companies, the job aborts without touching the database. A truncated download would otherwise mark an entire country as dead in one transaction.

05 //Methodology

Register figures were downloaded on 22 July 2026 from each authority's own open data and counted record by record, not sampled: France INSEE StockUniteLegale (1 July 2026 edition), Sweden Bolagsverket bulk file, Finland PRH YTJ API v3 and the all_companies file, Estonia Äriregister lihtandmed, Singapore data.gov.sg ACRA entity datasets, United Kingdom Companies House monthly basic data, Norway Brønnøysund enhetsregisteret. Swedish and Finnish figures are deduplicated to unique organisations, with a company counted as live if any current record shows it live. Our own database figures are counts of our stored records before and after the 22 July 2026 correction. A random sample of the companies we marked closed was re-checked one by one against the Finnish register's per-company endpoint, and every company in the sample was confirmed closed. Point-in-time figures, they will move as registers change. Free to cite with attribution to AtlasForgeX.

06 //FAQ

How many companies in a business register are closed? +

Measured on 22 July 2026 from the official files: 70.6% of the 2,102,483 entities in Singapore's ACRA datasets are no longer live, 60.5% of the 2,857,430 organisations in Sweden's Bolagsverket register are deregistered, 43.9% of the 29,803,585 legal units in France's INSEE Sirene file are cessée, and 43.9% of the 821,699 companies returned by Finland's PRH company endpoint are no longer registered. Registers are cumulative archives, so the closed companies stay in them permanently.

Why do dead companies stay in B2B databases? +

Because ingesting data and re-checking it are two different jobs, and only the first one is ever built. A record enters with the status active and nothing goes back to ask the register whether the company still exists. Dead companies keep their name, domain, industry code and phone number, so they pass automated quality filters. The only reliable test is the register's own status field.

How can I check whether my B2B list is full of closed companies? +

Take 100 random companies from the list, look each up in its national register by company number, and count how many come back closed. Use PRH for Finland, Bolagsverket for Sweden, Brønnøysund for Norway, Companies House for the UK, INSEE Sirene for France, Äriregister for Estonia and data.gov.sg for Singapore. All are free. Match on company number, not domain: domain matching produces false positives.

What does a dead company in your list actually cost? +

In our audit one cold outreach batch was 54.5% closed companies. Those addresses bounce, and bounce rate is the fastest way to damage a sending domain's reputation, which then suppresses the deliverability of the mail that would have worked. A caller ID index built on the same data named companies that had not existed for a decade, on numbers since reassigned to other people.

Does this affect Apollo, ZoomInfo or other databases? +

We did not audit their records, so we make no claim about their numbers. What we can say is that the mechanism is not specific to us: any pipeline that loads a national register without continuously re-reading its status field accumulates closed companies, and the registers themselves show that between 44% and 71% of what they contain has closed. The way to know is to test a sample of any vendor's list against the register yourself.

How does AtlasForgeX handle this now? +

A nightly job re-reads each country's current register, marks anything missing from it as terminated, and restores anything that reappears. Search results, industry counts, the caller ID index and outreach lists all read that one status field. If a register download returns implausibly few companies the job aborts rather than marking a whole country dead. The first run across seven countries marked 844,549 companies as closed.