Ravindra BagaleCourses & study guides Track your progress

Chapter 50

50.1 Composite Models: Import, DirectQuery, Dual

Chala mitrano — mix Import and DirectQuery safely; know Dual mode and aggregation hit/miss. Last chapter of this arc — finish strong.

Fictional sample data for learning. All company names and brands belong to their owners. We use the names only so students can understand. Numbers on this page are made up.
After this lesson you can
  • Compare Import, DirectQuery, Dual, and Live connection.
  • State composite model limits at a practical level.
  • Explain Dual mode relationship behaviour.
  • Describe aggregation hit vs miss.
  • Know when Live connection still wins.
  • Sketch a hybrid SQL fact + Import dimension model.

Start with a problem (fictional). Huge SQL fact cannot fully Import overnight. Dimensions are small. Priya builds a composite: DirectQuery fact + Import City/Date, Dual where needed, aggs for City-Day cards. Live connection to a central dataset still wins for some self-serve teams — she documents both paths for Cloud.

Sample layout — fictional data: Import dim, DQ fact, Dual, agg hit
Sample layout — fictional data (not a real Service screenshot)

1. Storage modes side by side

Why: One word Import is not enough when facts are huge.

What: Import caches data in the model. DirectQuery sends queries to the source. Dual can act as either. Live connects to an already published model/AS.

How — pick with eyes open:

Vertical steps (one under another)

  1. Small/medium + need speed: Import.
  2. Huge fact + need freshness: consider DirectQuery.
  3. Mixed: composite with Dual dimensions often.
  4. Org already has a certified dataset: Live may win.
  5. Write the choice on the architecture one-pager.

Tiny example: City and Date Import; Sales fact DirectQuery (fictional MH).

Ravindra Bagale's Tip

Friends, DirectQuery freshness is not free — source must answer quickly.

2. Composite model limits (practical)

Why: Not every feature loves composite.

What: Some interactions, calculations, and source types have limits — verify current docs for your sources.

How — stay practical:

Vertical steps (one under another)

  1. List sources in the composite.
  2. Check Training/Cloud for known blocked patterns.
  3. Test RLS behaviour with DQ.
  4. Watch query reduction and timeouts.
  5. Keep a fallback Import extract for critical morning cards if needed.

Tiny example: Timeout on a wild cross-join visual — redesign visual, do not only raise timeout (fictional).

Ravindra Bagale's Tip

Friends, read current Microsoft limits — this lesson teaches caution, not a frozen SKU sheet.

3. Dual mode and relationship behaviour

Why: Dimensions often set to Dual so they can satisfy both Import aggs and DQ facts.

What: Dual tables can be treated as Import or DirectQuery depending on the query path.

How — place Dual wisely:

Vertical steps (one under another)

  1. Common dims (Date, City), then Dual.
  2. Large fact, then DirectQuery.
  3. Understand relationships still define filter flow.
  4. Test a card (Import path) and a detail table (DQ path).
  5. Document Dual choices for the next author.

Tiny example: Date Dual lets City-Day agg Import-hit while detail stays DQ (concept).

Ravindra Bagale's Tip

Friends, Dual is not automatic magic — wrong relationships still hurt.

4. Aggregation tables with DirectQuery sources

Why: Morning cards should not scan the entire SQL fact.

What: Aggregations can answer summarised queries from Import agg tables; detail misses fall back to DQ.

How — plan hit/miss:

Vertical steps (one under another)

  1. Define grains users need on home pages.
  2. Build matching agg tables.
  3. Manage aggregations mapping.
  4. Prove hits with tools/Training.
  5. Accept miss for rare drills to detail.

Tiny example: Home City-Day card = hit; invoice-level matrix = miss to SQL (fictional).

Ravindra Bagale's Tip

Friends, a miss is OK for rare drills — a miss on every home card is a design bug.

5. When Live connection still wins

Why: Not every team should own a composite.

What: Live to a certified enterprise dataset gives one version of truth with less local modelling power.

How — choose Live when:

Vertical steps (one under another)

  1. A central model already has RLS and measures.
  2. Your team should only build reports.
  3. Governance wants fewer composite snowflakes.
  4. You still can use some thin report patterns.
  5. Document why you did not composite.

Tiny example: Finance self-serve Live to certified Margin model; Ops keeps hybrid composite for SQL fact (fictional split).

Ravindra Bagale's Tip

Friends, Live is governance-friendly — composite is power with responsibility.

6. Worked hybrid model + course close

Why: You need one architecture sketch you can draw in interviews.

What: SQL fact DQ + Import/Dual dimensions + City-Day agg + App on shared workspace — then stop and revise chapters 37-50.

How — draw and revise:

Vertical steps (one under another)

  1. Sketch boxes: SQL, then DQ Sales; Import City/Date Dual; Agg CityDay.
  2. Note refresh: dims schedule + source latency for fact.
  3. Note RLS still tested as role.
  4. Link back to capstone checklist (ch 44).
  5. Write three interview bullets for composite vs Live.

Tiny example: Course arc close: Service governance (37-44) + DAX/model power (45-50).

Ravindra Bagale's Tip

Friends, end with a drawing — interviewers remember sketches more than jargon.

Quiz (5 questions)

1. Import mode mainly:

2. DirectQuery mainly:

3. Dual mode is often used for:

4. Aggregation hit means:

5. Live connection is attractive when:

Practice tasks (12) — hints only, no full answers

  1. Task. Table Import vs DQ vs Dual vs Live — one line each.
    Hint: section 1.
  2. Task. Why source speed matters for DQ.
    Hint: section 1 tip.
  3. Task. List two practical composite risks to test.
    Hint: section 2.
  4. Task. Which tables often Dual in a star?
    Hint: Date City.
  5. Task. Home card hit vs invoice miss — explain.
    Hint: section 4.
  6. Task. When do you choose Live over composite?
    Hint: section 5.
  7. Task. Draw hybrid MH model boxes (text sketch).
    Hint: section 6.
  8. Task. Refresh story for Import dims + DQ fact.
    Hint: latency.
  9. Task. Does composite remove need for RLS Test as role?
    Hint: no.
  10. Task. Interview three bullets: composite vs Live.
    Hint: section 6.
  11. Task. Why agg miss on every home card is bad.
    Hint: section 4 tip.
  12. Task. Course close: name two themes from 37-44 and 45-50.
    Hint: governance / DAX.

Ravi Sir's cards (Excel card style)

Standalone cards — no tip or photo sections. QR points to this lesson. Download PNG to revise offline.

Interview Q&A: What is a composite model and when do you use it?
Download interview-card PNG
Comparison Card: Live connection vs composite model
Download comparison-card PNG
Revision Card: Hybrid model checklist
Download revision-card PNG

Think like a professional

In a real tenant, rights and sharing are controlled by Fabric / Power BI admins. Always follow your company policy. This lesson uses fictional people and numbers so you can practise the ideas safely.