How pandas drop_duplicates decides which duplicate row survives

Data analyst Michael Nocito published a technical guide on August 8, 2026, explaining how pandas handles duplicate rows using the drop_duplicates function. The article draws a key distinction between exact row duplicates and rows that share a key column but contain conflicting data, demonstrating that the two cases require different analytical approaches. Using an eight-row sample dataset, Nocito shows that the default keep='first' setting retains whichever row appears higher in the dataframe, meaning the survivor is determined by row order rather than data quality. He illustrates this with order ID 1003, where identical input data yields different results depending solely on which conflicting row was loaded first. The guide also notes that pandas and spreadsheet tools can produce different de-duplication outcomes on the same dataset, a discrepancy the article addresses in detail.