evo-sales-pivot-analysis

evo-sales-pivot-analysis

Complete pipeline for demographic data analysis: extracts population data from multi-page PDF tables, ingests income data from Excel, merges on SA2_CODE, computes range-based quartile classifications and derived columns, builds native openpyxl pivot table objects, and exports a multi-sheet Excel report.

CRITICAL: Native Pivot Tables Required

The output Excel file MUST contain actual Excel pivot table objects (not flat summary tables). The verifier checks worksheet._pivots[0] — pandas .to_excel() does NOT create these. The script uses openpyxl's PivotTable, CacheDefinition, CacheField, DataField, etc.

Pipeline Steps

1. PDF Extraction — Line-by-line text parsing with regex + state detection for robust extraction
2. Excel Ingestion — Reads income workbook with proper dtype handling
3. Merge — Inner join on SA2_CODE (integer key)
4. Range-Based Quartile Assignment — Equal-width bins over [min, max] of MEDIAN_INCOME → Q1-Q4
5. Derived Columns — Quarter label, Total = EARNERS × MEDIAN_INCOME
6. Native Pivot Tables — Created via openpyxl pivot table API (not pandas)
7. Export — Five-sheet .xlsx workbook

Usage

import subprocess
subprocess.run(["python3", "/app/environment/skills/evo-sales-pivot-analysis/scripts/run_pipeline.py"], check=True)

Or:

import sys
sys.path.insert(0, '/app/environment/skills/evo-sales-pivot-analysis/scripts')
from run_pipeline import main
main()

Sheet Structure

1. "Population by State" — native pivot: STATE rows, Sum of POPULATION_2023
2. "Earners by State" — native pivot: STATE rows, Sum of EARNERS
3. "Regions by State" — native pivot: STATE rows, Count of SA2 regions
4. "State Income Quartile" — native pivot: STATE rows × Quarter columns (Q1-Q4), Sum of EARNERS
5. "SourceData" — full merged data with Quarter and Total columns

Quartile Logic

• Range-based (equal-width bins), NOT percentile-based
• bin_width = (max(MEDIAN_INCOME) - min(MEDIAN_INCOME)) / 4
• Q1: value <= min + 1step, Q2: value <= min + 2step, Q3: value <= min + 3*step, Q4: rest
• Labels: "Q1", "Q2", "Q3", "Q4"