Is it permissible to loop back to earlier steps in CRISP-DM?

• A. No
• B. Only during data preparation
• C. Only during deployment
• D. Yes

Answer: (D)

CRISP-DM is an iterative, non-linear approach. In real data projects you often learn something new after you start modeling or after you evaluate results, which prompts you to revisit earlier steps. For example, while building a model you may realize the data lacks important features or has quality issues you didn’t spot initially, so you return to data understanding or data preparation to clean and engineer new features. During evaluation, if the model doesn’t meet the business goals, you might reframe the business understanding or adjust the model and data accordingly. Deployment can also reveal drift or new requirements, leading you to revisit data understanding or modeling again. Because of this, looping back to previous steps is not only permissible but expected; the process is designed to be flexible to improve outcomes.