DBbun/Policy_and_Grants_Decision_Simulator_v1.0

Policy & Grants Decision Simulator (Synthetic)

Overview

This repository provides fully synthetic datasets and analysis outputs generated by the Policy & Grants Decision Simulator (v1.0) — a Python-based simulation engine that models end-to-end government funding, grants, and acquisition workflows.

The simulator captures:

• vendor participation and self-selection,
• review and compliance phases,
• policy and regulatory “knobs,”
• administrative burden and delays,
• award and non-award outcomes.

The datasets are designed for comparative policy analysis, not prediction.

All data is entirely synthetic. No real vendors, proposals, agencies, or individuals are represented.

GitHub: https://github.com/DBbun/policy-grants-decision-simulator

---

How the Data Is Generated (Conceptual)

Each run simulates many funding episodes. For each episode:

1. A policy configuration is sampled.
2. A population of vendors is sampled.
3. Vendors decide whether to bid.
4. Submitted proposals flow through multiple review phases.
5. Time, attrition, and outcomes are recorded.

Monte-Carlo repetition produces distributions suitable for stress testing and sensitivity analysis.

---

Configuration Inputs (from config_snapshot.json)

Every dataset includes a machine-readable snapshot of the configuration used to generate it.

Top-level simulation parameters

• runs – number of Monte Carlo runs
• episodes_per_run – funding opportunities per run
• vendors_per_episode – min/max vendors sampled
• random_seed – reproducibility seed

Policy knobs (examples)

• documentation_burden – relative paperwork / compliance load
• sb_preference – weighting favoring small businesses
• oem_direct_adoption – degree of direct OEM engagement
• export_cui_strictness – data/control restrictions
• research_security_strictness – security review intensity

Knobs are sampled within configured ranges and may be perturbed for sensitivity analysis.

---

Dataset Contents

1. Core Simulation Outputs (output/)

policy_runs.csv

One row per Monte-Carlo run

Column	Description
run_id	Unique run identifier
policy_tag	Baseline or perturbed policy label
avg_award_days	Mean time-to-award (days)
sb_win_rate	Fraction of awards to small businesses
avg_admin_load	Synthetic administrative burden proxy
awards	Number of awards
no_awards	Episodes ending without award

---

decision_traces.csv

Event-level log (multiple rows per episode)

Column	Description
run_id	Run identifier
episode_id	Funding episode ID
vendor_id	Vendor identifier (synthetic)
actor	vendor or system
phase	Process phase (see below)
action	Action taken (e.g., bid, no_bid, advance)
reason_code	Attrition / disqualification reason
duration_days	Time spent in phase
is_small_business	Boolean
p_value	Probability used for stochastic decision

Phases include:
BID_DECISION, SUBMIT, COMPLIANCE_SCREEN, TECH_REVIEW,
COST_REALISM, RISK_REVIEW, AWARD, POST

---

phase_bottlenecks.csv

Aggregated phase-level timing statistics

Column	Description
phase	Process phase
share_of_total_days	Fraction of total process time
mean_days	Mean duration
p90_days	90th percentile duration

---

run_summary_by_knob.csv

Sensitivity analysis results

Column	Description
knob	Policy knob name
direction	plus / minus
delta_avg_award_days	Change vs baseline
delta_sb_win_rate	Change vs baseline
ci95_low_*	95% CI lower bound
ci95_high_*	95% CI upper bound

Negative deltas indicate improvement (e.g., faster awards).

---

assumptions.csv

Static modeling assumptions

Column	Description
assumption	Assumption name
value	Configured value
notes	Interpretation

---

config_snapshot.json

Full configuration used to generate the dataset, enabling exact reproduction.

---

Process Phases (Detailed Semantics)

Phases represent abstracted functional steps commonly found in government grants and acquisition workflows.
They are conceptual, not tied to any specific agency or regulation.

BID_DECISION

Who: Vendor
Purpose: Vendor self-selection

Models whether a vendor chooses to engage at all.

Drivers:

• expected payoff,
• documentation burden,
• capacity constraints,
• policy preferences,
• stochastic noise.

Outcomes:

• bid
• no_bid → vendor dropout

This phase is the primary driver of early attrition.

---

SUBMIT

Who: Vendor
Purpose: Proposal preparation and submission

Duration increases with:

• documentation burden,
• compliance complexity,
• limited vendor capacity.

Outcomes:

• successful submission,
• dropout during preparation.

---

COMPLIANCE_SCREEN

Who: System
Purpose: Administrative and eligibility screening

Models early checks such as completeness and eligibility.
Introduces short but variable delays and early disqualification risk.

---

TECH_REVIEW

Who: System
Purpose: Technical merit evaluation

Represents expert or panel review of technical quality.

Often contributes the largest share of total processing time.

---

COST_REALISM

Who: System
Purpose: Cost and budget evaluation

Models assessment of budget realism and cost sharing.
Higher scrutiny can increase delays and disproportionately affect smaller vendors.

---

RISK_REVIEW

Who: System
Purpose: Programmatic and compliance risk assessment

Abstracts security, data-control, and broader programmatic risks.
Strongly influenced by policy strictness.

---

AWARD

Who: System
Purpose: Selection and award processing

Final selection and award preparation.
Critical for final outcome and small-business win rate.

---

POST

Who: System
Purpose: Post-award administrative processing

Optional phase representing notifications, setup, and closeout.

---

Phase Timing Interpretation

• Phase durations are relative, not absolute.
• Absolute day counts should not be interpreted as real timelines.
• Comparative statements (e.g., “TECH_REVIEW dominates total time”) are meaningful.

---

Phase Interaction with Policy Knobs

Policy Knob	Primary Affected Phases
documentation_burden	BID_DECISION, SUBMIT, COMPLIANCE_SCREEN
sb_preference	BID_DECISION, AWARD
oem_direct_adoption	BID_DECISION, COST_REALISM
export_cui_strictness	RISK_REVIEW
research_security_strictness	TECH_REVIEW, RISK_REVIEW

---

Analysis Outputs (figures/)

Generated by analyze_results_v1.0.py.

Includes:

• distribution histograms,
• tradeoff plots,
• bottleneck and Pareto diagnostics,
• attrition analysis with explicit denominators,
• sensitivity plots for policy knobs.

A human-readable REPORT.md summarizes the main findings.

---

Interpretation Notes

• Rates use explicit denominators (per vendor decision, per episode).
• Dropout buckets are inferred, not observed.
• Metrics are comparative, not predictive.
• Results should not be interpreted as real-world performance estimates.

---

Intended Uses

Allowed without additional permission

• academic research,
• policy exploration,
• internal evaluation and benchmarking,
• education and training,
• visualization and methods development.

Requires a separate license

• operational or production use,
• integration into live decision-support systems,
• use in real policy or procurement decisions,
• commercial products or consulting,
• government agency deployment beyond exploratory analysis.

---

License

Research & Evaluation Use (Proprietary)

This dataset and all derived figures are fully downloadable for research and evaluation.

Commercial use, operational deployment, or use by government agencies for real decision-making requires a separate paid license.

Licensing inquiries:
[email protected]

---

Disclaimer

This dataset is entirely synthetic and does not represent any real agency, program, vendor, or procurement process.
Any resemblance to real systems is conceptual and coincidental.