Exam Strategy and Cross-Domain Review

You made it — let’s tie it all together

Domain weight recap

Domain	Weight	Modules	Key Focus
D1: Design & Implement Data Models	35-40%	M1-M11	Partition keys, data modelling, SDK, queries, server-side
D2: Data Distribution	5-10%	M12-M14	Consistency levels, global replication, multi-region writes
D3: Integrate & Move Data	5-10%	M15-M17	Change feed, Synapse Link, data movement
D4: Optimize Performance	15-20%	M18-M21	Indexing, RU optimisation, integrated cache, materialized views
D5: Maintain Solution	25-30%	M22-M28	Monitoring, backup, security, cost, DevOps

Study time allocation:

• 60% on D1 + D5 (they’re 60-70% of the exam)
• 25% on D4 (performance optimisation is practical and code-heavy)
• 15% on D2 + D3 (important but lower weight)

Most-tested topics (the “golden list”)

These topics appear most frequently in exam questions. Know them cold:

Rank	Topic	Domain	Why It’s Tested
1	Consistency levels	D2	5 levels, trade-offs, per-request relaxation, 2× RU cost
2	Partition key design	D1	Even distribution, cross-partition queries, synthetic keys
3	Indexing policies	D4	Composite indexes, include/exclude, impact on writes
4	Change feed	D3/D4	Lease container, modes, materialized views, soft-delete
5	RU cost factors	D4	Point reads, consistency impact, query optimisation
6	Backup and restore	D5	Periodic vs continuous, restore to new account
7	Security (RBAC vs keys)	D5	Master keys danger, resource tokens, data plane roles
8	Throughput modes	D5	Serverless vs provisioned vs autoscale, limitations
9	Transactions	D1	Partition-scoped, batch operations, stored procedures
10	Global replication	D2	Failover priority, recovered region = read-only

Common exam traps

💡 Trap 1: 'Recovered region becomes writer again'

After automatic failover, the recovered region does NOT automatically become the write region. It returns as a read-only replica at the lowest priority. You must manually reprioritize. The exam loves this — Option A is almost always “it reclaims its role” and is almost always wrong.

💡 Trap 2: 'Strengthen consistency per-request'

You can only weaken (relax) consistency on a per-request basis. If the account default is Session, you CANNOT request Strong for a specific read. Direction: Strong → Bounded → Session → Prefix → Eventual.

💡 Trap 3: 'Serverless has unlimited throughput'

Serverless is NOT unlimited. It provides 5,000 RU/s per physical partition (total throughput scales with partition count), supports single region only, and has SLA only with availability zones. It can now be converted to provisioned throughput (NoSQL API).

💡 Trap 4: 'Restore always goes to a new account'

For periodic backup, restore always creates a new account (via Azure Support). For continuous backup, you can restore deleted containers/databases into the existing account or create a new account. In-place overwrite of data (not containers) does not exist — you choose between restoring to existing vs new account.

💡 Trap 5: 'Change feed captures deletes'

The default change feed (Latest Version) does NOT capture deletes. Use the soft-delete pattern (set a flag + TTL) or switch to All Versions and Deletes mode (preview — requires continuous backup, NoSQL API only).

Question-answering strategies

For “which consistency level” questions:

1. Does the scenario need read-your-own-writes? → Session
2. Does it need linearizable reads? → Strong (if single-write region)
3. Is cost the priority? → Eventual or Consistent Prefix
4. Are multi-region writes mentioned? → Eliminate Strong and Bounded

For “how to reduce RU cost” questions:

1. Can you use a point read instead of a query? → Always cheaper
2. Is the query cross-partition? → Add partition key filter
3. Is it SELECT *? → Project specific fields
4. Is the indexing policy default? → Optimise to include only queried paths

For “choose the throughput mode” questions:

1. Traffic is sporadic/dev-test? → Serverless
2. Traffic is predictable? → Provisioned manual (+ reserved capacity)
3. Traffic varies but somewhat predictable? → Autoscale
4. Need multi-region or SLA? → Eliminate serverless

The “if in doubt” rules:

• If in doubt on consistency: choose Session (it’s the default and most balanced)
• If in doubt on authentication: choose RBAC (recommended over keys)
• If in doubt on backup: choose Continuous (more granular, self-service)
• If in doubt on network security: choose Private Endpoints (most secure)

Final tips from the team

🚀 Priya (NovaSaaS): “Domain 1 is 35-40% of the exam. If you nail partition key design, data modelling, and the SDK, you’re already nearly halfway to passing. Don’t skimp on the fundamentals.”

🛒 Jake (GlobeCart): “Consistency levels were my nemesis. Make flashcards for all five levels — know the analogies, the RU costs, and which ones work with multi-region writes. I saw at least 6 consistency questions on my exam.”

📡 Amara (SensorFlow): “The change feed appears everywhere — not just in D3. Materialized views (D4), denormalization propagation (D1), and event-driven patterns (D4) all build on it. Understand the lease container and the soft-delete workaround deeply.”

⚙️ Marcus (FinSecure): “Domain 5 is 25-30% and it’s broad — security, monitoring, backup, cost, DevOps. The good news is most questions are scenario-based: ‘what should Marcus do when he sees 429 errors?’ Think about the operational response, not just the theory.”

Exam day checklist

• Review the golden list of most-tested topics
• Know all five consistency levels cold (analogies + technical details)
• Be comfortable reading C# SDK code for CRUD, queries, and change feed
• Understand CompareTable relationships (serverless vs provisioned, periodic vs continuous, etc.)
• Read each question twice — scenario questions hide the real question in the last sentence
• Eliminate obviously wrong answers first — then choose between the remaining
• Flag and return to questions you’re unsure about — don’t get stuck

🎬 Video walkthrough

Flashcards

Question

What are the two heaviest domains on the DP-420 exam?

Click or press Enter to reveal answer

Answer

Domain 1: Design and Implement Data Models (35-40%) and Domain 5: Maintain an Azure Cosmos DB Solution (25-30%). Together they're 60-70% of the exam. Focus your study time accordingly.

Click to flip back

Show hint

Question

What are the three 'if in doubt' defaults for the exam?

Click or press Enter to reveal answer

Answer

1) Session consistency (default, most balanced). 2) RBAC with Entra ID (recommended over master keys). 3) Continuous backup (better granularity, self-service). When a question doesn't give enough detail to be specific, these are the safest answers.

Click to flip back

Show hint

Question

What's the single most important thing to know for DP-420?

Click or press Enter to reveal answer

Answer

Consistency levels — all five levels, their guarantees, their trade-offs (RU cost, latency, availability), per-request relaxation (weaken only), and which ones work with multi-region writes (Session, Consistent Prefix, Eventual). This topic alone can be 10-15% of the exam.

Click to flip back

Show hint

Knowledge Check

Knowledge Check

A Cosmos DB account has multi-region writes enabled with Session consistency. A developer wants to perform a point read with Strong consistency for a critical financial check. What happens?

AThe read succeeds with Strong consistency at 2× RU costBThe read succeeds but silently uses Session consistency insteadCThe read fails — Strong consistency is not available with multi-region writes, and you cannot strengthen consistency per-requestDThe read fails because point reads don't support consistency overrides

Check Answer

Knowledge Check

A cross-partition query costs 500 RU per execution and runs 100 times per minute. Which optimisation would have the biggest impact?

AAdd a composite index for the ORDER BY clauseBRedesign the query to include the partition key in the WHERE clause (single-partition query)CEnable the integrated cache with 5-minute stalenessDSwitch from Session to Eventual consistency

Check Answer

Knowledge Check

Which of these is a common exam trap? (Select all that apply)

ARecovered regions automatically become the write region after failoverBYou can strengthen consistency on a per-request basisCServerless Cosmos DB supports multi-region writesDContinuous backup can restore deleted containers to the existing account

Check Answer

---

🎉 Congratulations! You’ve completed all 28 modules of the DP-420 study guide. You’ve built a deep understanding of Azure Cosmos DB — from data modelling through global distribution, performance optimisation, and operational excellence.

Your next steps:

1. Review the golden list of most-tested topics
2. Re-do all flashcards and quizzes (aim for 90%+)
3. Take a practice exam under timed conditions
4. Schedule your exam — you’re ready!

Good luck from Priya, Jake, Amara, Marcus, and the entire team. 🚀🛒📡⚙️