Business Data Cleansing

Problem Addressed

After LoadData produces a raw business table, two common data issues can break downstream analysis: rows where sales are null, and rows that are exact duplicates. Feeding these directly into statistical skills can interrupt aggregations or inflate totals, making results harder to validate.

How It Works

The skill expects raw_data as a two-dimensional business data list. It performs two low-risk filtering operations:

• Filter invalid rows: removes records with empty sales values so downstream calculations do not hit nulls.
• Remove duplicates: drops fully duplicated business rows to reduce statistical bias from repeated records.

It returns clean_data, a standardized structured business dataset. The skill does not add new calculation definitions, rewrite valid records, modify business logic, or perform file reads/writes, external API calls, or data egress.

Boundaries

It is intended for internal, authorized structured business data pipelines, especially as a step between LoadData and statistical skills. Do not treat it as a general-purpose transformation tool: it is not for personal privacy data, sensitive or classified data, external distribution, export, or commercial use; it also does not infer fields, impute missing values, or recalculate metrics.