Greek company information in GEMI: what it shows and what is missing

The company information that GEMI, the General Commercial Registry of Greece, shows comes in five kinds: registry data such as the name and the registered address, activity codes (KAD, the Greek extension of NACE), the management and the partners, the documents the business has filed, and its trade names. What is missing is the numbers: the financial statements are among the documents as PDF files, and a company's revenue, profit and debt can only be read by opening the file.

For each of the 1,148,876 Greek businesses in BizList, I checked 23 more kinds of data, from financial statements as numbers to public contracts. 310,546, or 27.0%, have data from at least one source outside GEMI. Among active companies, leaving out sole proprietorships, the share rises to 44.0%.

What GEMI shows

There are more records than businesses, because a business usually has several activity codes, several people and several documents. GEMI publishes the documents as PDF files, as they were filed.

Financial statements as numbers

From the PDF financial statements, 539,359 financial years have been extracted for 134,336 businesses, with 8,671,538 values: revenue, profit, assets, liabilities, cash, depreciation and more. Each financial year has on average 16 of the 24 fields filled in. Cash flows exist in only 1.7% of financial years, because only large companies must publish them.

A language model reads the PDF files. Every change to its instructions first runs against a set of financial statements checked by hand, and it does not go through if accuracy drops in any category. One of the errors the check catches is units: a statement that says "in thousands of euros" while the amounts are already in euros gives revenue a thousand times too large. There are also 676 financial years that looked wrong, were checked with a second reading and turned out correct.

How well these financial statements cover the Greek economy, I have measured separately: for the 2023 financial year they cover 67.9% of the revenue that Eurostat measures, the statistical office of the EU.

The same documents also give the signatories, who signed each financial statement or decision, for 50,881 businesses.

Shared management and similar businesses

GEMI shows the people on the page of each business. The VAT number of each person links the pages together, and 457,134 businesses have at least one other business with a person in common in the management or among the partners. How many companies the average Greek takes part in I have measured here.

For 99,084 businesses there are also similar businesses, based on the description of their activity.

Sources outside GEMI

Six of these sources have their own analysis: how many Greek businesses advertise on Google and Facebook, how many have a website, which ones win public contracts, where the NSRF money goes, which ones produce their own electricity, which sectors are hiring.

How much extra data a business has

538,554 businesses, or 46.9%, have none of the 23 extra kinds. Between them and the 27.0% with a source outside GEMI there is a 26.1%, 299,776 businesses, that only have data drawn from GEMI itself: financial statements as numbers, signatories, shared management or similar businesses. 432,058 of them are active sole proprietorships, which do not file financial statements with GEMI: 64.7% of the 667,782 active sole proprietorships. Of the 396,564 active companies, 83.7% have at least one extra kind of data. At the other end, 16,410 businesses have ten or more.

Methodology

The data comes from the BizList.gr database on 7 October 2026: 1,148,876 Greek businesses, companies and sole proprietorships, of which 1,064,346 are active. The unit is the kind of data per business, present or absent. A business with seven activity codes counts once for activity codes, and one with fifty contracts once for contracts. In records, the 7.8 million activity code rows would weigh more than all outside sources together.

The share with sources outside GEMI counts only the 19 sources in the table. Financial statements as numbers, signatories, shared management and similar businesses count among the 23 kinds, but they come from GEMI's own data. For ads and website traffic, only businesses where something was found count: of the 139,159 that were checked, 72,059.

Sources joined by VAT number, that is contracts and bankruptcies, leave out the VAT number EL000000000, which 16 thousand businesses in GEMI have instead of their real one. The first count without this filter joined each contract with all those businesses at once, and gave 31.6 million contract and business pairs instead of 1.76 million. Shared management counts only people with a full VAT number, and sanctions only matches that were confirmed.

Three kinds that are computed for almost every business were not counted: businesses with the same manager or the same address, rankings and sector indicators. With them, the number of extra kinds per business would be higher. In the opposite direction leans the VAT number with several records in GEMI, which counts in all of them: contracts give 72,901 businesses, while the unique VAT numbers are 65,311.