If you torture the data enough, it will confess.

A proper analysis must be objective, rigorous, transparent, and reproducible. Analysis plans should be detailed in advance; analysts are prone to adjusting their analyses to fit a specific conclusion, sometimes inadvertently. These plans should detail all steps taken at the collection, processing, and analysis stages. In cases of prescriptive analyses, where analytic results direct specific actions, the analyst should also have a plan for quantitatively measuring the impact of the action.





Analysis plans should include details about:

• The data sources used in the analysis
• Processing steps on the raw data, for example
  • Selecting – which observations and features are included in the analysis?
  • Formatting – are the data modified in any way?
  • Filtering – are specific observations excluded due to values in their features?
• The category of analysis (descriptive, predictive, prescriptive)
• The type of analysis (regression, categorization, clustering, optimization, etc.)
• Assessing the analysis performance