THEUNPARTYRUNWAY | ASSUMPTIONS

CORE

Quantitative data quality assessment system that generates statistical reports on business assumptions with transparent, modifiable scoring methodology. Uses the theunpartyrunway foundation labels schema for consistent terminology across all outputs.

Foundation Labels Schema

The system uses foundation labels from theunpartyrunway-foundation-labels.json:

Label	Definition	Maps To	Weight	Visual
invitation	A container that when score threshold is met grants access to process	`evidence_level`	40%	Blue row (invitation_row)
party	A single-CORE gathering of logic	`data_source`	30%	Purple row (party_row)
guest	A data-source required by an invitation to execute properly	`completeness`	30%	Orange row (guest_row)
bug	A data entity that has been manipulated from its original state	`overall_score`	-	Title row score

Scoring Formula

code

score = (invitation × 0.4) + (party × 0.3) + (guest × 0.3)

Where:

invitation = Evidence level factor (MEASURED=1.0, CALCULATED=0.8, ESTIMATED=0.5, ASSUMED=0.3, etc.)

party = Data source quality (API=1.0, DATABASE=0.9, CALCULATED=0.8, MANUAL=0.5, etc.)

guest = Completeness factor (0.0 to 1.0)

bug = Final overall data quality score (0-10 scale)

Files

This directory contains exactly 5 files:

1. assumptions.py - Generator script (run this to regenerate all outputs) 2. assumptions.json - Machine-readable data with full methodology and foundation labels 3. assumptions.csv - Spreadsheet format with foundation label column headers 4. assumptions.md - Human-readable report with foundation labels documentation 5. assumptions.png - Statistical visualizations using invitation_row/party_row/guest_row layout

Usage

bash

python3 assumptions.py

Regenerates all 4 output files (.json, .csv, .md, .png) from source data and weights defined in assumptions.py.

Visualization Layout

The PNG visualization (assumptions.png) uses a 3-row grid structure:

1. invitation_row (Blue) - Evidence levels distribution and contribution 2. party_row (Purple) - Source quality distribution and contribution 3. guest_row (Orange) - Category weights and scores

Each row contains: [Table | Pie Chart | Bar Chart]

The bug score appears in the title row (top right). 'budget': 0.15, # Financial planning 'grow': 0.10 # Growth metrics }

code

### Evidence Levels
- `MEASURED` (1.0) - Direct measurement from tools/APIs
- `CALCULATED` (0.8) - Computed from available data
- `DATA_DRIVEN` (0.7) - Based on actual data but derived
- `ESTIMATED` (0.5) - Educated guess with some data
- `ASSUMED` (0.3) - Pure assumption
- `MISSING` (0.0) - No data available

## Modifying Weights

Edit the configuration dictionaries at the top of `assumptions.py`:

1. **Category Weights** - `CATEGORY_WEIGHTS` dict (relative importance)
2. **Scoring Components** - `SCORE_WEIGHTS` dict (formula components)
3. **Evidence Levels** - `EVIDENCE_LEVELS` dict (0.0-1.0 scale)
4. **Source Quality** - `SOURCE_QUALITY` dict (0.0-1.0 scale)

All weights have inline documentation explaining rationale and manipulation instructions.

## Current Results

**Overall Data Quality Score:** 6.2/10

**Category Breakdown:**
- 🟢 core: 9.2/10
- 🟡 product: 4.4/10
- 🟡 brand: 5.0/10
- 🟡 risk: 5.0/10
- 🟢 budget: 8.6/10
- 🟡 grow: 5.1/10

## Data Points

27 total data points tracked, all with purely numerical values:

**Examples:**
- `codebase_lines`: 106965
- `test_coverage_percentage`: 5
- `claude_api_spend`: 412.84
- `commits_per_day`: 10.2
- `vendor_concentration_percentage`: 55

## Output Formats

### JSON (assumptions.json)
- Full methodology documentation
- Category scores
- Individual data point scores
- Machine-readable for integrations

### CSV (assumptions.csv)
- Tabular data for spreadsheet analysis
- All numerical values preserved
- Import into Excel/Sheets/etc

### Markdown (assumptions.md)
- Human-readable report
- Category breakdowns
- Priority recommendations
- Improvement roadmap

### PNG (assumptions.png)
- Statistical visualizations
- Distribution charts
- Score comparisons
- Evidence level breakdown

## Transparency

All outputs include complete methodology documentation. Scoring is deterministic and reproducible. Weights are explicit and documented. No hidden calculations.