Quant Archetype Deep Dive¶
Part of the archetype deep-dives series. See also
../archetype-families.mdand../evolution-pathways.md.
Family definition¶
Quants are measurers. Their currency is numerics: scores, rates, intervals, distributions, risk premia, p-values, loss curves. They turn uncertainty into defensible estimates and estimates into decisions.
Core values¶
- Numbers with uncertainty bars. A point estimate without a confidence interval is incomplete.
- Reproducibility over rhetoric. The same data must yield the same answer.
- Stationarity awareness. The model applies only where its assumptions hold.
- Calibration discipline. Predictions must match outcomes over time.
FCC Quants (15 in the current registry)¶
| ID | Name | Category |
|---|---|---|
| ACT | Actuarial Reserving Specialist | insurance |
| CCR | Counterparty Credit Risk Analyst | finance |
| CDA | Clinical Data Analyst | healthcare |
| CFD | Claims Fraud Detection Analyst | insurance |
| ESC | Experiment Scientist | ml_lifecycle |
| ESG | ESG Investment Analyst | finance |
| FAS | Forecasting Analyst | ml_models |
| FRA | Financial Risk Analyst | finance |
| IAN | Interpretability Analyst | ml_lifecycle |
| IRR | Interest Rate Risk Modeler | finance |
| IUW | Insurance Underwriter Analyst | insurance |
| LDPA | Legal Data Privacy Analyst | legal |
| NCA | NanoCube Analyst | ux_visualization |
| OA | Ontology Architect | knowledge_graph |
| PCR | Parametric & Climate Risk Designer | insurance |
(Note: OA classifies as Quant by the name-pattern heuristic but also
carries strong Architect characteristics — a reminder that family
classification is probabilistic.)
Archetype signature: most distinctive R.I.S.C.E.A.R. components¶
Quants have highly distinctive Inputs (always include raw data sources, not summaries) and Expected Output (always includes uncertainty, validation statistics, or holdout performance).
- Inputs — named datasets, sampling windows, cohort definitions.
- Expected Output — point estimate + interval + diagnostics; never a single number alone.
- Constraints — stationarity, minimum sample size, exclusion criteria.
- Role Skills — always include a named statistical or ML technique.
Style is less distinctive (often overlapping with Architect / Investigator).
Discernment matrix profile¶
Dominant traits:
- Curiosity — hypothesis generation is part of the daily loop.
- Professional Background — named quantitative credentials matter.
- Humility — quants are constantly reminded by holdout sets that they were wrong.
- Responsibility — medium; quants advise rather than decide.
- Taste — low-medium; elegance of decomposition matters for explainability.
- Inclusivity — variable; matters for bias audits, less for pricing.
Design Target Factor profile¶
- Curiosity (very high) — defining factor.
- Influence (medium-high) — via credible numbers.
- Leadership (medium) — lead on method, follow on policy.
- Optimism (low-medium) — calibrated, not cheerful.
- Social Connectivity (low-medium) — deep work is solitary.
- Diversity Appreciation (medium) — important for bias work.
Common collaboration patterns¶
Quants feed Architects (metrics become features) and Shepherds (numbers become evidence). They draw raw material from Investigators (research findings) and partner with Storytellers (who translate the numbers).
flowchart LR
IN[Investigator] -->|raw findings| QN[Quant]
AR[Architect] -->|pipelines| QN
QN -->|metrics| SH[Shepherd]
QN -->|insights| ST[Storyteller]
QN -->|diagnostics| SE[Safety Engineer]
QN -->|peer review| QN2[Peer Quant]
Pairing heatmap¶
| Quant pairs with | Frequency | Purpose |
|---|---|---|
| Architect | very high | Architect ships pipeline, Quant ships metric |
| Storyteller | high | Narrate the number |
| Shepherd | high | Evidence for compliance |
| Investigator | medium-high | Upstream research |
| Safety Engineer | medium | Model risk diagnostics |
| Other Quants | high | Peer review, replication |
Quant evolution pathway¶
Quants evolve from point-estimate calculators (Stage 2, numbers without intervals) to interval-aware analysts (Stage 3, uncertainty quantified, diagnostics run) to calibration stewards (Stage 4, federated across ecosystems, continuously monitored for drift).
- STRUCTURED — named method in
role_skills, cohort defined inconstraints. - SEMANTIC — discernment matrix scored, collaborators upstream (Investigators) and downstream (Storytellers, Shepherds) wired.
- FEDERATED — risk classification + NIST AI RMF mapping + drift monitoring hook registered in the event bus.
Worked examples¶
Forecasting Analyst (FAS)¶
- id: FAS
name: Forecasting Analyst
archetype: The Trend Predictor
category: ml_models
riscear:
role: Produces time-series forecasts with calibrated intervals.
constraints:
- Minimum 24 observations before fitting
- Declare non-stationarity explicitly
expected_output:
- Point forecasts for 1-4 step horizons
- 80% and 95% prediction intervals
- Cross-validated MAPE/RMSE diagnostics
role_collaborators: [ENA, MAR, ESC, IAN, MOS]
Experiment Scientist (ESC)¶
The hypothesis-testing specialist. Collaborates tightly with FAR (Feature Architect) upstream and IAN (Interpretability Analyst) downstream.
Financial Risk Analyst (FRA)¶
FEDERATED-stage Quant — paired with SCA (Shepherd, SOX) and MRV (Shepherd, Model Risk). Demonstrates the classic "Quant produces → Shepherd validates" loop.
When to use a Quant¶
Pick a Quant-family persona when:
- The deliverable is a number with defensible properties.
- A stationarity or distributional assumption must be checked.
- Holdout validation is part of the acceptance criterion.
- A regulatory metric must be produced (VaR, expected shortfall, reserve estimate).
Avoid when the deliverable is qualitative (prefer Investigator or Storyteller).
Further reading¶
../evolution-pathways.mdarchitect.md— the Quant's most frequent partner.investigator.md— upstream source.../../guidebook/ch19_evaluation_benchmarking.md— CLEAR+ 7-dimension benchmarking that Quants consume.