Methodology Memo
Analytical Question
What is the client asking? What type of question is this (descriptive, inferential, predictive, causal)? What would a good answer look like?
Data Sources and Quality
What datasets are you working with? What is the date range, granularity, and size? Any quality issues discovered during profiling?
Preparation Decisions
Feature Engineering
What features did you create? What features did you exclude and why? How did you handle categorical variables? What transformations did you apply?
Temporal Splitting
How did you split training and test data? What date boundary did you use? Why is a temporal split necessary instead of a random split?
Leakage Assessment
Which features were evaluated for leakage? What did you find? Which features were removed or lagged and why?
Data Quality Issues
What issues did you discover in the data (missing values, censored data, system changes)? How did you handle each one?
Methodology
What model did you choose? Why this model over alternatives? What evaluation metrics did you use and why?
Limitations
What can this analysis NOT tell the client? What assumptions does the model rely on? Under what conditions would the results be unreliable?
Findings Summary
What are the key findings, organized around the client's requirements? What should the client do with this information?