Journalists
01Where is business activity concentrated?
Journalists can compare cantons and sectors, distinguish operating companies from pure holdings, and identify questions worth investigating.
Swiss open data · company universe
535,494 active Swiss companies, classified by what they do.
This analysis turns official company records into a practical view of the Swiss business landscape. It shows where selected companies are based, which activities dominate, how the current population is distributed by founding year, and where automated classification still needed human review.
Selected company universeZEFIX / UID / NOGA
535'494
Companies included
Integrated snapshot
2026-08-11
| Canton code | Companies |
|---|---|
| ZH | 88850 |
| VD | 46779 |
| BE | 45904 |
| GE | 39217 |
| ZG | 34339 |
| AG | 31336 |
| TI | 30379 |
| SG | 28506 |
| LU | 25407 |
| VS | 25335 |
| SZ | 17958 |
| FR | 17400 |
| TG | 16319 |
| BL | 13728 |
| GR | 13453 |
| SO | 12384 |
| BS | 12033 |
| NE | 9435 |
| JU | 4568 |
| NW | 4439 |
| SH | 4424 |
| AR | 4042 |
| OW | 3311 |
| GL | 2550 |
| AI | 1728 |
| UR | 1654 |
ZH88'850
Boundaries · swisstopoQuestions before technology
Who can use this study of Swiss companies?
Researchers
02How was the population constructed?
Researchers can inspect every inclusion rule, exclusion count, source role and known limitation before reusing the aggregates.
Public administrations
03What can joined open registers reveal?
Public administrations can see how Zefix, UID and NOGA connect, and where missing or non-open evidence limits interpretation.
SMEs and fiduciaries
04What does the local business mix look like?
SMEs and fiduciaries can compare a canton or activity group without exposing contacts, directors or other personal data.
Population design
From 747,725 decisions to a focused universe of 535,494 companies.
The goal was not to reproduce every register entry. We wanted a stable, small-company-oriented population of active AG/SA, GmbH/Sàrl/Sagl and VAT-registered sole proprietorships. Legal forms outside that scope, inactive records, branch entities and clear large-employer signals were excluded.
Every exclusion is counted
747'725
Source decisions
Companies included535'494
Records excluded212'231
- 125'158Sole proprietorship without active VAT registration
- 67'038Legal form outside the selected scope
- 11'270Inactive register status
- 8'579Explicit employee band of 50 or more
- 183More than 50 registered directors
- 3Official branch relationship after legal-form filtering
Legal-form filtering already removes most branch forms. A further three otherwise eligible records were excluded because the official register relationship identified them as branches. Head offices with registered branches remain included.
The population is deliberately narrow
535'494100%
- GmbH / Sàrl / Sagl258'729
48.3% - AG / SA226'048
42.2% - VAT-registered sole proprietorship50'717
9.5%
Where the selected companies are based
- ZH88'85016.6%
- VD46'7798.7%
- BE45'9048.6%
- GE39'2177.3%
- ZG34'3396.4%
- AG31'3365.9%
- TI30'3795.7%
- SG28'5065.3%
- LU25'4074.7%
- VS25'3354.7%
- SZ17'9583.4%
- FR17'4003.2%
- TG16'3193.0%
- BL13'7282.6%
- GR13'4532.5%
- SO12'3842.3%
- BS12'0332.2%
- NE9'4351.8%
- JU4'5680.85%
- NW4'4390.83%
- SH4'4240.83%
- AR4'0420.75%
- OW3'3110.62%
- GL2'5500.48%
- AI1'7280.32%
- UR1'6540.31%
All 26 cantons are represented. Sixteen records have no usable canton and are excluded from the map only. Counts describe the current selected population, not canton-level formation rates or revenue.
Integrated snapshot: 2026-08-11
Open the separate fiduciary landscapeEconomic activity
Professional services, trade, real estate and construction form the largest blocks.
All 21 represented NOGA sections
- NProfessional, scientific and technical activities93'08717.4%
- GWholesale and retail trade68'29312.8%
- MReal estate activities58'05610.8%
- FConstruction53'1609.9%
- LFinancial and insurance activities49'0399.2%
- CManufacturing41'7247.8%
- IAccommodation and food service activities30'0685.6%
- KTelecommunication, programming and information services28'1975.3%
- OAdministrative and support service activities25'6244.8%
- TOther service activities22'9954.3%
- RHuman health and social work activities17'5493.3%
- HTransportation and storage11'4092.1%
- SArts, sports and recreation9'2801.7%
- QEducation8'9511.7%
- JPublishing, broadcasting and content production8'4581.6%
- AAgriculture, forestry and fishing5'7591.1%
- DElectricity, gas, steam and air conditioning supply1'7750.33%
- EWater supply, sewerage and waste management1'6530.31%
- BMining and quarrying3990.07%
- PPublic administration and defence170.00%
- UActivities of households as employers10.00%
The 15 most frequent NOGA groups
- 642#1
Activities of holding companies and financing conduits
34'9336.5% - 702#2
Business and other management consultancy activities
32'1556.0% - 682#3
Rental and operating of own or leased real estate
24'2274.5% - 711#4
Architectural and engineering activities and related technical consultancy
23'4254.4% - 681#5
Real estate activities with own property and development of building projects
22'5874.2% - 561#6
Restaurants and mobile food service activities
21'8484.1% - 433#7
Building completion and finishing
17'6663.3% - 432#8
Electrical, plumbing and other construction installation activities
16'6183.1% - 621#9
Computer programming activities
15'8203.0% - 692#10
Accounting, bookkeeping, auditing and tax consultancy
12'9762.4% - 953#11
Repair and maintenance of motor vehicles and motorcycles
12'4792.3% - 477#12
Retail sale of other goods
12'2722.3% - 683#13
Real estate activities on a fee or contract basis
11'2422.1% - 464#14
Wholesale of household goods
10'7592.0% - 410#15
Construction of residential and non-residential buildings
10'3011.9%
Codes follow NOGA 2025. Official English labels are retained here so the downloadable aggregate stays language-neutral and reproducible.
Current population by founding year
Companies active today by founding year, from 2016 to 2025.
Classification and review
One first-pass classification for every company. A separate review for ambiguous cases.
Analytical operating profile
These five profiles are Navor inferences. They are not official Zefix or FSO attributes.
535'494100%
- Operating462'905
86.4% - Mixed activity35'109
6.6% - Pure holding33'246
6.2% - Uncertain3'914
0.73% - Financing conduit320
0.06%
Final confidence
535'494100%
- High404'428
75.5% - Medium123'027
23.0% - Low8'039
1.5%
Which result is final?
481,562 classifications retain the Gemini result. For 53,932 flagged companies, the reviewed value overrides Gemini in the final merged dataset while the original model output remains preserved.
- Gemini result retained89.9%
- Reviewed result overrides Gemini10.1%
What happened in the review queue
Reviewers confirmed or corrected both resolved and still-flagged cases. A retained flag means the evidence remained genuinely mixed or insufficient after review.
- Corrected, review flag retained31'46458.3%
- Confirmed, review flag retained15'63129.0%
- Confirmed and resolved5'38710.0%
- Corrected and resolved1'4502.7%
Method and provenance
From the official register to the final result in six traceable stages.
01 · Register base
Start from Zefix company identities.
We normalized UID, name, status, legal form, registered purpose, canton, founding date, branch relationships and available register links.
02 · VAT status
Check the official UID and VAT service.
Every source UID was checked through the federal UID service. VAT status is required only for sole proprietorships in the selected population.
03 · Population rules
Apply the same filters nationwide.
We kept active AG/SA, GmbH/Sàrl/Sagl and VAT-registered sole proprietorships. We excluded branch entities, CRIF employee bands of 50 or more, and—through a separate rule—records with more than 50 registered directors. The employee-band field is a private screening signal, not a complete official headcount.
04 · Website discovery
Search for and verify the company website.
An automated Serper workflow queried Google Search and Google Maps. The ten Google results returned for each query were passed to Gemini 3.1 Flash Lite, which identified a likely company domain when one was present. Only domains attributable to the exact UID passed to the next stage; doubtful or conflicting matches were rejected.
Search templates used for website discovery
{
"google_search": [
"{company_name} ({director_1} OR {director_2} OR {director_3}) email"
],
"google_maps_if_crif_website_missing":
"{company_name} {city} {canton} Switzerland"
}05 · NOGA classification
Classify every company and flag uncertainty.
Gemini 3.5 Flash classified each company from its registered purpose and, when available, its verified website. It returned a NOGA group, operating profile, confidence level and a flag for uncertain results.
View the exact company payload sent to Gemini
Company payload in the Gemini classification prompt
UID: {uid}
Official name: {name}
Legal form: {legal_form}
Registered seat: {registered_seat}
Founding date: {founding_date}
Official legal purpose: {purpose}
CRIF industry: {industry}
CRIF employee band: {employee_band}
CRIF turnover band: {turnover_band}
Website URL: {url}
Website domain: {domain}
Website trust status: {trust_status}
Website retrieval status: {retrieval_status}
Homepage title: {title}
Homepage meta description: {meta_description}
Cleaned homepage text (max 8,000 chars): {cleaned_text}06 · Review and merge
Review the uncertain cases with ChatGPT Luna at max effort.
ChatGPT Luna reviewed the cases flagged by Gemini. Only for those companies does Luna’s result replace Gemini in the merged dataset.
How uncertain classifications were reviewed
First pass
Gemini 3.5 Flash
535,494 companies
Gemini classified the full population and flagged the results it considered uncertain.
Review
ChatGPT Luna
53,932 flagged cases
ChatGPT Luna at max effort is a stronger reasoning model. We used it to review Gemini’s uncertain cases, and its result takes priority for those companies.
Source roles, not one undifferentiated data lake
Official open data anchors the analysis. Official services, register interfaces and private sources play narrower supporting roles.
Official open dataZefix on opendata.swissCore active-entity identities and links to register information, updated daily by the Federal Office of Justice.Official APIZefix REST APIMachine-readable company details, branch relationships and SOGC publication references.Official register serviceUID and VAT registerFederal UID lookup and VAT-register status used to test sole-proprietorship eligibility.Official classificationNOGA 2025 KUBBFSO classification used for the 287 three-digit activity groups and the section hierarchy.Register evidenceCantonal register excerptsFounding dates and director records were parsed from cantonal excerpts and register interfaces linked from Zefix. Director counts support one selection rule; names are not published.
Public web evidenceCompany websites and search evidenceTrusted exact-company websites and bounded stored search results helped interpret activity. Unattributed or conflicting domains were rejected or flagged.
Official geodataswissBOUNDARIES3Dswisstopo canton boundaries used for the aggregate map.Supporting private sourceCRIFSupplemental activity and employee-band evidence. CRIF is not open government data and is not redistributed here.Download aggregate CSV
Use the aggregate, check the method, and challenge the assumptions.
The downloadable file contains every chart value on this page. If you find a better public-data interpretation, we want to hear it.