Purpose

Promotional Penetration: 52% of orders (5,196) received promotions with an average discount of 30%, indicating aggressive promotional activity across the business
Revenue Paradox: Promotions generate +2.6% revenue lift ($232.74 vs $226.74 average order value), yet this gain is not statistically significant (p-value=0.632, negligible effect size=0.01)
Profit Destruction: Promoted orders show -110% profit impact, averaging -$6.66 profit versus +$66.90 for non-promoted orders\u2014a $73.56 swing
Loss Prevalence: 18.7% of all orders (1,871) are unprofitable, with loss-making orders averaging 48.1% discounts versus 8.1% for profitable orders
Breakeven Threshold: Discounts at 30-50% and above consistently generate negative

This section documents the data preprocessing pipeline for a promotional discount analysis spanning 9,994 orders across three product categories and customer segments. Perfect data retention (100%) indicates no rows were removed during cleaning, suggesting either exceptionally clean source data or minimal validation criteria applied. Understanding preprocessing decisions is critical because data quality directly impacts the reliability of the profitability conclusions, particularly given the analysis reveals significant profit erosion from discounting.

Key Findings

Retention Rate: 100% (9,994 rows preserved) - No observations were excluded during preprocessing
Rows Removed: 0 - Complete dataset remained intact through the pipeline
Train/Test Split: Not documented - No explicit model validation split is recorded
Data Transformations: Not specified - Preprocessing steps applied remain undocumented

Interpretation

The perfect retention rate suggests the dataset arrived in usable condition with no missing values, duplicates, or outliers requiring removal. However, the absence of documented transformations and train/test methodology raises questions about data validation rigor. Given that the analysis identifies 1,871 loss-making orders (18.7%) and a -110% profit impact from promotions, the lack of explicit data quality checks or anomaly detection could mask underlying data issues that might explain extreme profit swings in certain discount buckets.

Context

The undocumented preprocessing approach limits transparency