To meet this certification criterion, candidates must demonstrate that they can connect their technical work to the real-world business problem.
This is about showing not just how well a model fits data, but also how well it serves the business goal. That means identifying a meaningful Key Performance Indicator (KPI), explaining why it matters, and using it to compare approaches.
To meet the expectation, your report should:
- Choose a KPI that reflects what the business ultimately cares about. For example: prediction accuracy against a business target (e.g., average call-center wait time)
- Explain why this KPI was chosen and show how it directly relates to the problem. For instance, “Accuracy is appropriate because the business goal is to correctly predict call wait time 90% of the time.”
- Use the given data to demonstrate how both of your models/approaches perform relative to the KPI.
- Highlight which model better supports the business objective, and explain how you determined that from the results.
Submissions fall short when they:
- use only technical/statistical metrics (e.g., RMSE, R², standard error) without tying them to the business goal. These are useful for evaluation, but they don’t show business impact.
- don’t define a KPI at all, or they present multiple metrics without clarifying which one matters to the business.
- don’t explain why the KPI was chosen, leaving unclear how it measures success.
- fail both models’ performance on the same KPI, which prevents a meaningful comparison.
Ultimately, this criterion is about demonstrating that you can translate technical results into business insight. A strong submission will not only show that you can build and evaluate models, but also that you understand how to measure their success in terms that the business cares about.
By clearly defining, justifying, and applying a KPI, you show that your data science work is actionable, comparable across approaches, and aligned with driving real-world outcomes.