Swiss startup survival, part 9 of 12
Five headlines that are not what they seem
Record foundings, record bankruptcies, 60% AI adoption: each headline is true, and each hides a definition that changes the story. How I audited the sources behind my study.
In consulting, the most expensive mistakes often start with a number nobody questioned. Before building my analysis, I audited the sources behind the claims I planned to use. Five headlines turned out to mean something different from what they seem to say.
The first is “record company foundings”. Commercial register counts include holding companies, which are among the largest groups of new entries.1 The Federal Statistical Office counts only companies with real economic activity, and its figures showed foundings flat in 2023.2 The register measures paperwork; BFS measures businesses.
The second is “record bankruptcies”. Since 2025, tax and social insurance authorities must pursue unpaid claims through bankruptcy proceedings rather than the old seizure route, and Dun & Bradstreet notes that this makes comparisons with earlier years misleading.3 Part of the jump is a change in the law, not in the economy, and the same effect will flatter the figures once it fades.
The third is “60% of Swiss firms use AI”. Depending on the survey, Swiss AI adoption ranges from about 8% of small firms4 to around 60% of all firms, with only 2% of micro-firms using AI systematically.5 “Using AI” can mean one employee trying a chatbot once, or a company redesigning a process. The definition decides the headline.
The fourth is “AI projects succeed 74% of the time”, or 5%, depending on who was asked and what counted as success. Part 8 takes that apart in detail.
The fifth is a funding total. EY puts Swiss startup funding in 2025 at more than CHF 3.3 billion,6 while the Swiss Venture Capital Report puts it at CHF 2.95 billion.7 Different databases, different definitions; neither is wrong, but mixing them in one argument would be.
Always say which source a number comes from. In my study I used the Swiss Venture Capital Report for all funding figures, because it is the only source with the full list of rounds by canton.
How I rated my sources
| Claim | Source | Reliability | Weak point |
|---|---|---|---|
| Survival by canton, sector, size | BFS company demography | High | Two-year lag |
| Venture capital by canton and sector | Swiss Venture Capital Report | High | Disclosed rounds only |
| AI share of venture capital | EY Startup Barometer | Medium to high | EY’s own definition of AI |
| AI adoption in firms | AXA/Sotomo, UBS, KOF | Medium | Each defines “using AI” differently |
| AI project returns | MIT, IBM, McKinsey, S&P | Medium | Selective samples, self-reporting |
Who was counted? What counted? Over what time? Who published it? If you cannot answer all four, do not build a decision on the number.
The full source audit is in section 18 of the study.
-
Federal Statistical Office (BFS), Company demography. ↩
-
KOF ETH Zurich, Digital technology: are small firms being left behind? ↩
-
startupticker.ch and SECA, Swiss Venture Capital Report 2026. ↩
