Healthcare startup failure patterns

13 structural failure patterns emerge from 241 documented healthcare startup deaths. Each company maps to exactly one primary failure node. The most common: Outspent by bigger competitors (27% of failures).

All 13 failure patterns

Each pattern below represents a distinct structural failure mechanism. Counts show how many confirmed failures primarily exhibit that mechanism. Percentages are of the confirmed failure corpus. The top sub-vertical shows where each pattern concentrates most.

1. Outspent by bigger competitors

Top sub-vertical: Care Navigation

66

27%

2. Couldn't get paid by insurance

Top sub-vertical: Care Coordination

50

21%

3. Customer acquisition costs too high

Top sub-vertical: Revenue Cycle Management

29

12%

4. Multi-stakeholder approval bottleneck

Top sub-vertical: Telehealth Infrastructure

19

8%

5. Regulation moved faster than the company

Top sub-vertical: Compliance & Regulatory

18

7%

6. Couldn't prove the product worked

Top sub-vertical: Healthcare B2B (Other)

18

7%

7. Sold through the wrong channel

Top sub-vertical: Healthcare B2B (Other)

10

4%

8. Ran out of money before clinical trials

Top sub-vertical: Healthcare B2B (Other)

9

4%

9. Preclinical science did not translate

Top sub-vertical: Healthcare B2B (Other)

8

3%

10. Product disrupted existing workflows

Top sub-vertical: Credentialing

5

2%

11. Ran out of pivots before finding fit

Top sub-vertical: Revenue Cycle Management

4

2%

12. Couldn't connect to existing systems

Top sub-vertical: Care Coordination

3

1%

13. Product claims exceeded actual capabilities

Top sub-vertical: Compliance & Regulatory

2

1%

How patterns are coded

Every corpus entry is coded using the Gioia three-level qualitative methodology. First-order codes capture raw observations from primary sources — post-mortems, investor write-ups, press coverage, and SEC filings. Second-order themes group those codes into structural patterns through constant comparison. Third-level aggregate dimensions map themes to one of the 13 structural failure nodes.

Each company receives exactly one primary failure node — the mechanism most directly responsible for the shutdown. A company may exhibit signals from multiple nodes, but the primary assignment reflects the dominant structural cause. This constraint forces specificity and prevents the corpus from over-attributing failures to diffuse multi-causal explanations.

A consensus threshold governs node assignment: at least two independent sources must support the primary node before an entry is finalized. Entries with ambiguous evidence receive a confidence flag and may be reviewed in subsequent audit cycles.

Concentration limits

No single failure node may exceed 25% of all confirmed failures. This concentration ceiling is enforced by a pre-commit hook that computes the distribution before every corpus modification. A commit that would push any node above the ceiling is blocked at the command line until the offending entries are reviewed.

Shannon entropy is monitored at each corpus update. A healthy corpus produces entropy at or above 70% of the theoretical maximum for its node count. When entropy drops below this threshold, it indicates that a small number of failure patterns are dominating the corpus — which reduces the engine's ability to surface nuanced structural risks for less common failure modes.

The concentration ceiling and entropy floor work together to ensure the corpus reflects the actual diversity of structural failure mechanisms in healthcare B2B, rather than the patterns that happen to be easiest to document or most prominent in press coverage.

Corpus basis: 13 structural failure nodes across 241 confirmed failures. Node taxonomy is append-only — new nodes require a closed decision, 3 anchor entries, and dual validation. The 25% concentration ceiling is enforced at commit time. Shannon entropy monitored at each corpus update.

Related methodology

How the corpus is built →How scoring works →Sub-verticals →
Run a stress test →