Methodology

Financial metrics come from annual statements filed with the Tax Administration (IRMS). A missing value is shown as “—” and is never counted as zero: an unparsed statement is not 0€ revenue, and a year without a filing is not a 100% drop.

• Total revenue (€): operating income from the income statement (AOP 201–204: sales, change in inventories, capitalised work and other operating income) plus financial income (AOP 222, 226 and 230).
• Profit (€): net comprehensive income (AOP 260), or net result (AOP 287) when AOP 260 is empty. A statement whose profit exceeds 5× its revenue and 1,000,000€, or whose profit plus payroll costs exceeds revenue by more than 1,000,000€, is left out of profit totals, margins and profit rankings.
• Employees: the average headcount on the form (row 001). An implausible figure (for example one equal to the filing year) is not shown.
• Avg payroll cost per employee (€ per month) = AOP 211 (net wages, taxes and contributions on wages including the employer's, and other personnel expenses) / employees / 12. Values below 300€ a month are not shown. It is not an employee's gross or net salary; the net pay estimate applies the filing year's deductions and is shown from 2022 only.
• Derived metrics: revenue per employee = revenue / employees; margin = profit / revenue (only with positive revenue); YoY change = (current − previous) / |previous|, with no percentage against a zero base or when profit turns into a loss (or back); CAGR = (latest / first)^(1/years) − 1, only when both values are positive.
• Economic indicators keep their source frequency; annual GDP is summed only from all four quarters and incomplete years are clearly marked.

BizStats covers legal entities that file annual statements with the Tax Administration, linked by their eight-digit PIB. Company data comes from the IRMS snapshot of 1 Nov 2026.

• Sole proprietors (natural persons registered for a business) are never shown individually: there are no pages with their names, and company counts, rankings, search and multi-year trends leave them out. Their statements, without names or identification numbers and under the same quality rules, are included in the year, sector, activity and municipality totals with the note “Totals include N sole proprietors”. Where that note is absent, a total covers legal entities only: in activities dominated by sole proprietors (such as legal services or renting) it is then well below the whole activity.
• Banks report to the Central Bank (CBCG) and insurers to the Insurance Supervision Agency (ASO), not to the Tax Administration. Bank revenue and profit come from their CBCG statements, without employees or pay; insurers are missing. The finance & insurance sector is therefore left out of pay and margin comparisons.
• Companies are primarily joined by eight-digit PIB. Tenders retain CEJN and internal identifiers. Links without a verified key are marked uncertain and never presented as confirmed.
• An IRMS listing entry without a parsed statement (status no_metrics, every amount 0) is a missing statement, not a company with 0€. Every 2014 entry is one, so 2014 is not an available year.

A company is assigned to the municipality of its current seat for every year: a move changes the municipality in earlier years too, and new municipalities have series from before they existed — Tuzi (a municipality since 2018) and Zeta (since 2022). Municipal totals for earlier years therefore describe today's territory and seats, not the municipality of the time.

Totals (revenue, profit, employees) add up every statement of the year in scope; profit leaves out implausible statements.

• Average payroll cost for the country, a municipality or a sector = sum of AOP 211 / sum of employees / 12, over companies that report pay only; it is hidden when they cover less than 80% of employees. For 2015–2019 there is therefore no national average and none for most municipalities and sectors: about a quarter of employees work at companies whose reported pay is below 300€ a month.
• The latest year comes from an IRMS snapshot that does not yet hold every late filing, so it is marked preliminary while more than 10% of last year's filers are missing. Changes against the previous year are computed on the same companies (those that filed in both years, excluding empty filings); the change across all published filings is shown in the tooltip.
• Employees, pay and profit from 2020 come from a different source than 2015–2019, so the series have a break, marked on the charts.

Rankings are never paid for and follow the rules below. The floors come from one configuration shared by the API and the pages, and every ranking carries a “Criteria” line.

• Average payroll cost and revenue per employee: only companies with at least 10 employees — in rankings, in the municipal top 10 and as the default filter when sorting the company explorer (where it can be switched off).
• Biggest YoY gainers and decliners: in the base year at least 100,000€ revenue and 5 employees; ranked by absolute change in euros; a swing between loss and profit gets no percentage and is left out of the median.
• For payroll cost per employee also: at least 5 employees in the current year and a headcount change of at most 50%. Values above 3× the sector's 99th percentile or that changed more than 2× are flagged for checking.
• Revenue CAGR: fixed 3/5/10-year windows ending at the latest year; at least 100,000€ revenue and 5 employees at both the start and the end; the start year is never a company's first filing year. Companies with revenue below 1,000€ in the latest year are listed separately as “Ceased activity”, ranked by absolute decline.
• Company explorer: the margin sort ranks companies with at least 100,000€ revenue first, and the revenue-change sort those with at least 1,000,000€ revenue and 5 employees in the previous year; the others follow. A change always compares with the previous year's filing, never across a gap: a zero base is “n/m”, and employees, payroll cost and profit are not comparable between 2019 and 2020.
• Profit rankings leave out implausible statements; the finance and insurance sector is not ranked on payroll cost or margin while banks and insurers are missing; sector medians on company pages use companies with at least 5 employees. Municipal pay and revenue-per-employee figures are shown only where the municipality's companies employ at least 10 people in total.

The verdict applies fixed rules (version verdict-v1) to public sources and shows the worst signal: “Serious warning signs”, “Caution advised”, “Worth watching” or “Looks solid”. Every claim carries its source date. The rules, worst first:

• Bankruptcy or a forced-liquidation risk → “Serious warning signs”. Source: the Tax Administration registry (status and court orders) and the latest edition of its bankruptcy list.
• Account blockade → “Serious warning signs” from 180 days of blockade, otherwise “Caution advised”. Source: the Central Bank's (CBCG) latest blocked-accounts snapshot.
• Tax-debtor list → “Caution advised” from a debt of 50,000€, otherwise “Worth watching”. A list whose latest edition is older than 12 months is shown only as a historical record (“Worth watching”), and amounts from different lists are never added up. Source: Tax Administration lists, with their edition dates.
• Missing annual statements, from the registration year to the latest year the IRMS snapshot expects → “Caution advised” for 2 years or more in a row at the end, otherwise “Worth watching”. Source: IRMS and the registry (CRPS).
• An inactivity warning, liquidation or removal from the registry → “Caution advised”. Source: the Tax Administration and the registry (CRPS).
• Financial trend, over consecutive years only (a year without a filing breaks the run) → “Worth watching”: a loss in the latest year or 2 years in a row, or falling revenue 2 years in a row. Source: IRMS.
• No negative signal → “Looks solid”, with notes for profit (3 years or more in a row) and revenue growth on the previous year; no financial data → “Not enough data yet”. “No blockades, tax debts or missing statements” is only claimed for an active company whose sources are current.

The peer-group position (peer-position-v3) is a descriptive measure: where the company's latest statement sits among comparable companies of the same year. It is not a credit rating or a risk assessment: there is no letter grade or composite score, and it does not replace the at-a-glance verdict.

• Peer group: companies in the same KD 2010 class and the same size by employees (1–9, 10–49, 50–249, and 250+) in the same year, the company itself excluded. With fewer than 20 of them the KD 2010 division (the first two digits of the code) of the same size is used; with fewer than 20 there too, it reads “too few comparable companies”.
• For revenue, employees, revenue per employee, net margin and average gross pay the page shows the share of peers below the company (0–100; equal values count half) and the group median. A margin needs positive revenue and is not computed beyond ±1000% or on an implausible profit, and pay only when it is reliable; a measure without data is left out and the others are shown.
• Consistency: profitable years among the statements of the last 5 years, the average yearly revenue growth over 3 years and the largest fall in revenue from an earlier peak. A year without a statement is skipped, not counted as zero.
• The official signals (bankruptcy, account blockade, tax debtor lists, latest statement, forced liquidation) come from the same corrected data as the at-a-glance verdict, each with its source and snapshot date.
• It is not shown for companies in bankruptcy or deregistered, with forced-liquidation or inactivity signals, in the finance and insurance sector, with fewer than 2 years of revenue, or when the latest statement is more than 2 years older than the latest data year.
• Backtest on historical data (21 Oct 2026): for every statement year from 2016 to 2023 the position was recomputed with the same rules and compared with what happened later. Later trouble means the company filed no statement for either of the next two years, is in bankruptcy, liquidation or deregistered, is on a tax debtor list, had a blocked account or meets the forced-liquidation conditions. The registry, lists, blockades and IRMS listing are 2024–2026 snapshots, so the follow-up period is longer for older years.
• Result: the AUC of the individual measures is between 0.53 and 0.65 (0.5 = chance, 1 = perfect separation). That is a weak signal: the position does not predict which company will run into trouble, and the net margin without a peer group (0.63–0.65) and the share of loss years (0.64–0.67) separate as well or better. So the position is not used as a risk assessment, only as a description of where a company sits among similar ones.

Annual statements with data per year (without unparsed entries) and the processing status of every entry. The latest year is preliminary until late filings arrive.

Each dataset shows two dates: “Refreshed” (when BizStats last fetched the source) and “Data through” (the newest date or period the records cover). Datasets without an automatic refresh are marked “not scheduled”. The freshness rating takes the worse of the source's age and the age of the derived views on company pages. Source revisions enter the next successful ingestion.