Skip to main content

Design Phase

Updated May 12, 2023 ·

Overview​

The design phase defines the project's goals, impact, and key performance metrics.

Added Value​

Before building a model, it's important to estimate its potential value.

  • Ensures resources are used efficiently.
  • Reduces uncertainty in ML projects.
  • Supports responsible deployment.

Business Requirements​

A machine learning model must meet business needs beyond just accuracy.

  • Prediction needs

    • Frequency of predictions (e.g., daily, hourly).
    • Speed of predictions (e.g., real-time or batch).

  • Transparency & explainability

    • Important for understanding errors and improving models.
    • Some industries require explanations for decisions.

  • Regulatory compliance

    • Example: Financial models must provide decision explanations.

  • Budget & team size

    • Determines project feasibility and complexity.

Key Metrics​

Aligning these metrics ensures that all stakeholders understand and support the ML project.

From different perspectives:

  • Data Scientist

    • Focuses on accuracy and correctness.
    • Example: Mean Squared Error (MSE) for regression models.

  • Subject Matter Expert

    • Looks at business impact, like improved workflow.
    • Example: A hospital model reducing patient wait times.

  • Business Stakeholder

    • Measures financial impact (profit, cost savings).
    • Example: Increased sales due to better demand forecasting.