Context

ShopLens is building a binary classification model to predict whether a visitor will purchase within the current session so the marketing team can trigger a limited discount. A product manager is concerned because the model scored well on one test split, but performance varied when the data science team re-ran evaluation on different samples.

Current Performance

The Problem

The CMO asks: "Why do we need cross-validation instead of just reporting the best test score?" You need to explain the importance of cross-validation to a non-technical stakeholder and connect it to business risk.

Requirements

1. Explain what cross-validation is in plain language.
2. Use the metrics above to show why one split can be misleading.
3. Describe what the mean and standard deviation across folds imply about model reliability.
4. Explain how cross-validation helps with model selection and stakeholder confidence.
5. Recommend how ShopLens should report performance to executives.

Constraints

• Discount offers reduce margin by $4 per converted order.
• False positives create unnecessary discount spend.
• The team needs an evaluation approach that can be explained in under 2 minutes to business leaders.