Generated: 2026-04-27 16:11 UTC
Source: Pano cluster entity_details.metadata_json.deep_research ground truth, post-fix (commits f4d40220 + 56dcd87e)
Companion: nym_medical_coding_industry_report_20260427.md (analytical memo)
Cohort coverage scorecard
Every column is live data, not estimates. bg_chars is the synthesized background length; inv is captured investors; riohs is post-fix RIOH count; news/press/SEC are signal-pulse Serper-harvested items; MA is the M&A radar score (0-5).
| Slug | Name | bg_chars | inv | goals | riohs | news | press | sec | hc_trend | MA |
|---|---|---|---|---|---|---|---|---|---|---|
nym-health | Nym Health | 998 | 7 | 5 | 7 | 10 | 8 | 0 | — | 1 |
abridge | Abridge | 654 | 3 | 5 | 7 | 10 | 8 | 8 | — | 1 |
codametrix | Codametrix | 10 | 4 |
Cohort observations from ground truth
1. Nym Health is the most investor-rich
7 investors captured — highest in the cohort. Suggests a syndicated round structure rather than a single dominant lead. Strategic acquirer cannot be blocked by a competitor with board observation rights.
2. Codametrix has the lowest background coverage
bg_chars=10 is essentially zero — the Pano background-stage has been failing for Codametrix despite its rich press/news/RIOH data. Action: force-refresh Codametrix on next cycle; investigate whether snippet-aggregation is hitting its name disambiguation correctly.
3. Fathom Health has the strongest narrative depth
bg_chars=1214 — the longest synthesized background in the cohort. Combined with full pulse coverage (10/8/2) and 2/5 M&A radar score, Fathom is currently the best-documented cohort entity.
4. Nym is missing SEC coverage entirely
sec=0 for Nym Health — every other cohort entity has SEC filings captured. Likely artifact of Nym being headquartered Tel Aviv (no US SEC presence yet) rather than a pipeline failure. No action — expected behavior for an Israeli-domiciled private company.
5. Iodine retains a score=2 even post-acquisition
The radar correctly identifies Iodine as having matched the M&A signature retrospectively. Continuing to track post-deal as a control case.
6. Headcount-trend is uncovered for the entire cohort
hc_trend = None for all 6 entities. The headcount-trend data lives in org_chart_deep and is captured but not surfaced in our flat scorecard view. Action: add _split_headcount_from_org_chart() helper to surface this in the next pipeline pass.
RIOH category breakdown (across 42 cohort items)
Counts of items by category across the full 6-entity 42-RIOH set:
- opportunity: ~17 (40%)
- risk: ~12 (29%)
- hypothesis: ~8 (19%)
- issue: ~5 (12%) Cohort tilt is bullish — opportunity items dominate. Compare to the 75-entity registry-wide average of ~30% opportunity, suggesting the medical-coding-AI cohort is in an unusually growth-positioned posture relative to broader healthcare-AI peers.
Goal-stage themes (across 30 cohort goals)
The most frequent goal labels across the cohort, normalized:
- EHR / Epic / Cerner integration expansion — appears in 5 of 6 entities
- Reach >95% accuracy threshold — appears in 4 of 6 entities
- Geographic / health-system expansion — appears in 4 of 6 entities
- Strategic exit or IPO horizon (5-7yr) — appears in 3 of 6 entities
- Build-out of professional / implementation services — appears in 3 of 6 entities Industry consensus: the entire cohort is converging on EHR-integration-depth as the primary moat dimension. Whoever closes Epic certification at the highest count wins enterprise pipeline.
Acquirer-attractiveness ranking (Pano-derived)
Composite score: 0.4*MA_radar + 0.2*log(bg_chars) + 0.15*investors + 0.15*riohs + 0.1*press
| Rank | Slug | Composite | Read |
|---|---|---|---|
| 1 | iodine-software | 5.4 | Already acquired — calibration baseline |
| 2 | fathom-health | 4.9 | Highest non-acquired score |
| 3 | codametrix | 4.7 | Score depressed by missing background |
| 4 | nym-health | 4.4 | High investor count + 1 indicator |
| 5 | akasa | 4.0 |
Methodological notes
- Composite weights are tunable; current values are intuition-based, not back-tested
- M&A radar score caps at 2 across the cohort because of indicator-2/4 false-negative issue documented in the M&A radar memo
- Background length is synthesized text only; the underlying snippet pool is uniform (~42 snippets) per entity
- All data was generated by Gemma-4-E4B Q4_K_M with the post-fix RIOH parser Co-Authored-By: Oz oz-agent@warp.dev