This post is a part of the DP-600: Implementing Analytics Solutions Using Microsoft Fabric Exam Prep Hub; and this topic falls under these sections: 
Implement and manage semantic models 
    --> Design and build semantic models 
        --> Choose a storage mode
This is supplemental information to what is included in the "Choose a storage mode" post.

---

DP-600 Cheat Sheet: Choosing a Storage Mode in Microsoft Fabric

Storage Mode Decision Matrix

Requirement / Scenario	Import	DirectQuery	Direct Lake	Composite
Best query performance	✅ Excellent	❌ Depends on source	✅ Excellent	✅ Very good
Near real-time data	❌ No	✅ Yes	✅ Yes	✅ Yes
Large datasets (TB-scale)	❌ Limited	✅ Yes	✅ Yes	✅ Yes
Minimal refresh overhead	❌ Requires refresh	✅ No refresh	✅ No refresh	⚠ Partial
Uses OneLake Delta tables	❌ Not required	❌ Not required	✅ Required	✅ Optional
Full DAX & modeling features	✅ Full support	⚠ Limited	⚠ Limited	✅ Full
Calculated tables supported	✅ Yes	❌ No	❌ No	✅ Yes (Import tables only)
Lowest data duplication	❌ High	✅ None	✅ None	⚠ Mixed
Simple to manage	✅ Yes	⚠ Depends on source	⚠ Fabric-specific	❌ More complex

---

When to Choose Each Storage Mode

✅ Import Mode — Choose when:

• Dataset fits comfortably in memory
• You need complex DAX, calculated tables, or calculated columns
• Performance is the top priority
• Data freshness can be managed via scheduled refresh

Exam clue words: fastest, complex calculations, small to medium data

---

✅ DirectQuery — Choose when:

• Data must always be current
• Source system is highly optimized (SQL, Synapse, etc.)
• Data volume is very large
• You want zero data duplication

Exam clue words: real-time, source system, no refresh

---

✅ Direct Lake — Choose when:

• Data is stored as Delta tables in OneLake
• Dataset is large and frequently updated
• You want Import-like performance without refresh
• You’re working fully within Fabric

Exam clue words: OneLake, Delta, no refresh, Fabric-optimized

---

✅ Composite Model — Choose when:

• You need flexibility across different tables
• Fact tables are large and live (Direct Lake / DirectQuery)
• Dimension tables are small and stable (Import)
• You want performance and modeling flexibility

Exam clue words: hybrid, mix storage modes, dimension vs fact

---

Fast Exam Inclusion/Elimination Tips

• Calculated tables required? → Import or Composite
• OneLake + Delta tables? → Direct Lake
• Real-time + external source? → DirectQuery
• Best balance of flexibility and scale? → Composite

---

One-Sentence Exam Rule

If it’s in OneLake and too big to refresh, Direct Lake is usually the right answer.