Skip to content

Design: Phase 1 - The Productivity-Spending Correlation Audit

Overview

This audit aims to establish a "Source of Truth" by correlating financial spending patterns with professional productivity (GitHub activity). The goal is to identify the relationship between AI/Impulse spending and coding output to inform future optimization strategies.

1. Data Ingestion & Cleaning

1.1 Financial Data

  • Sources: download-transactions(2).csv and provided text for pending transactions.
  • Normalization: All transactions will be converted to CAD using the exchange rates provided in the CSV.
  • Cleaning: Remove noise from merchant descriptions to facilitate pattern matching.

1.2 Productivity Data

  • Source: GitHub CLI (gh).
  • Metrics: Commits, push events, and pull request activity for the period Jan 2026 - June 2026.

2. Targeted Categorization

Transactions will be mapped into four primary segments:

  1. AI_SPEND: Subscriptions and API usage for AI services (OpenAI, Claude, OpenRouter, DeepSeek, etc.).
  2. WEED_IMPULSE_SPEND: Transactions identified as cannabis-related or high-frequency convenience spending (e.g., UberEats, fast food, bars).
  3. ESSENTIAL: Necessary living expenses (Rent, basic groceries, utilities, essential transport).
  4. NON_ESSENTIAL_OTHER: All other discretionary spending.

3. Correlation & Analysis

The analysis will compute:

  • AI ROI: Total AI Spend / Total GitHub Activity.
  • Behavioral Correlation: Statistical correlation between Daily Spending (AI + Weed/Impulse) and Daily GitHub Activity.
  • The "Cost of Inactivity": Daily burn rate of subscriptions during periods of zero GitHub activity.

4. Visualizations

  • The Correlation Plot: A dual-axis time-series graph (Spending vs. Commits).
  • The Segmented Sunburst: A radial chart showing the distribution of spending across the four segments.
  • The Leakage Heatmap: A calendar view of high-spending days.

5. Constraints & Privacy

  • Privacy: No data will be transmitted outside the local environment.
  • Integrity: All calculations must use the normalized CAD values.

Skaft Software · MIT License