Veridion has raised a $20m Series A led by Hoxton Ventures, with its existing backers Underline Ventures, OTB Ventures, GapMinder, Day One Capital and LAUNCHub all returning. The company sells a continuously updated graph of the world’s businesses to risk, insurance, procurement and market intelligence buyers.
Its own boilerplate puts the graph at 642 million companies across 249 countries, with 461 attributes each. The release rounds that to 640 million.
Hussein Kanji, Hoxton’s founder, said in the announcement that Veridion could become “one of the most important data companies in the world”. That is the bet, and the number underneath it has been moving quickly.
The argument Veridion is making
Stefan Gergely, the company’s head of growth, put the thesis plainly in the announcement.
“Information updated quarterly or annually is no longer intelligence. It’s history.”
He also asked why organisations are “still making billion-dollar decisions using data that’s months old”. It is a fair question, and one the incumbents have not answered well.
The number has moved a long way
When Veridion raised $6m in 2023, under its former name Soleadify, it described a weekly-updated database of more than 80 million company profiles. Three years later the figure is roughly eight times larger.
Coverage expansion is the product, so growth is the point rather than a problem. It does make the definition of the unit the most important thing in the announcement, and the definition is missing.
Commonly cited estimates put the number of businesses worldwide in the low hundreds of millions, with formally registered companies a good deal fewer than that. TNW could not trace those figures to a single authoritative count, which is part of the difficulty.
What might be inside the count
The boilerplate offers a clue without settling it. Veridion says it resolves each company into legal registrations, operating locations and corporate hierarchy, which are three different things that can each be counted.
A graph assembled from websites, registries, filings, product catalogues and social profiles will also meet the same business more than once. Deduplication across those signals is the hard engineering problem in this category, and it is the one nobody markets.
None of that means the number is wrong. It means a reader cannot tell whether 642 million counts firms, entities, sites or records, and the difference matters to everyone downstream.
The claim Veridion does make
The company says every data point traces back to its source. That is a real commitment and more than many data vendors offer.
Traceability is not the same as verification. Knowing which website a fact came from tells you the provenance of the claim, not whether the claim is true.
The speed pitch rests on that distinction. Registries and filings are slow because they are authoritative; web signals are fast because they are not, and the release presents the speed without the trade.
Where it lands
Experian is the named partner and its statement is the substantive part of the announcement. Jon Roughley, its director of data strategy and innovation, said the signals let Experian capture “risk that traditional credit data simply couldn’t see” and score businesses previously invisible to clients.
Scoring the previously invisible is the whole commercial case, and it is also where the regulatory question sits. The entities missing from conventional credit files are the smallest ones, and a large share of those are sole traders.
It also places Veridion next to a category that already attracts scrutiny. The data broker industry spends heavily lobbying against restrictions on what it collects and sells.
Why that matters in Europe
Sole traders are natural persons in law, and the EU AI Act lists systems evaluating the creditworthiness of natural persons among its high-risk categories. Business data is mostly outside that scope, and the long tail is exactly where the boundary blurs.
The deadline moved recently. Brussels agreed to thin out the AI Act this year, pushing standalone high-risk obligations to 2 December 2027.
That is roughly fifteen months away, and the models being trained on this data now are the ones that will be in scope. The Commission already has powers to inspect models, restrict market access and fine 3% of global turnover.
The crisis framing is doing a lot of work
The release opens on the Strait of Hormuz, describing it as disruption and a single geopolitical event. It has been a war since late February, with tanker traffic collapsing and oil above $100.
Veridion says organisations using its technology could immediately identify businesses and suppliers exposed to the crisis. No customer is quoted saying they did.
The insolvency citation, read in full
The release cites Allianz’s global insolvency outlook for record business failures, and that is fair. Allianz does forecast 2026 as a fifth consecutive year of increases and a record high.
The same report says two other things. Western European insolvencies are forecast to fall 2% in 2026, and global insolvencies to decline in 2027.
So the trend the product is sold against is forecast to turn, in the region the announcement is datelined from. That does not undermine the business, and it does complicate the pitch.
The fact the release leaves out
Veridion is Romanian. It was founded as Soleadify by Florin Tufan with Sorina Vlăsceanu and Mihai Vinaga, and its earlier funding came from the Bucharest and Sofia venture ecosystem.
Tufan, still the chief executive, gets one short line in the announcement, saying the money will let the company “expand internationally”. The head of growth gets four.
The word Romania does not appear in the announcement, which is datelined London. One of Europe’s larger company-data businesses coming out of Bucharest is the most interesting thing here, and it has been edited out.
What to watch
Watch for a published methodology. A count of entities, a deduplication rate and a definition of company would settle most of the questions above, and competitors in this category do publish them.
Watch how the signals are labelled inside credit models. Opacity about where scoring inputs come from is a long-running problem, and web-signal collection without notice has already caused trouble elsewhere.
Be the first to comment