Methodology Memo

Fill in each section as you work through the analysis. This is a living document -- update it as decisions are made and findings emerge.

Data Source

Describe the dataset: where it came from, how many rows, how many columns, and what the class distribution looks like. Note any data quality issues found during profiling.

Data Preparation

Document the cleaning and encoding decisions: which columns were removed and why, how categorical features were encoded, how the binary target was created, and any features excluded from the model with justification.

Classification Approach

Which models were tried and why. What features were included. How was the model selected -- what criteria mattered beyond raw metric performance?

Evaluation Strategy

Which metrics were used to evaluate the models and why. If accuracy was not the primary metric, explain what was used instead and why it better serves the client's needs. Document any threshold adjustments and the rationale.

Limitations

What the model cannot predict or does not account for. Any data quality issues that affect reliability. Conditions under which the model's predictions should not be trusted.

Key Findings

Summary of results in the client's language. What production factors drive quality? How reliable is the model? What should the client do with these results?