AzureOctober 14, 202621 min read

AZ-305 Case Study Questions: How to Read and Answer Them (2026)

Most of the AZ-305 is a reading test. Here is how to find the constraint, the qualifier and the noise in a stem, with three examples taken apart option by option.

  • 40-60 typicalQuestions
  • 100-120 minTime
  • 700 / 1000Pass mark
  • ~2 minPer question
  • Min admin effortKey qualifier
AZ-305 case study questions decoded - keywords, qualifiers and traps

Table of Contents

AZ-305 rarely asks what a service does. It hands you a company, a set of requirements and four designs that all sound plausible, then asks which one an architect should sign off. Most wrong answers are real Azure services used in the wrong place: a regional load balancer where a global one was needed, a backup product where a replication product was needed, a virtual machine where a managed database would have met every requirement with less work. Knowing the services gets you to two candidates. Reading the stem picks the winner.

Case studies make this sharper. You get a long brief (overview, existing environment, planned changes, business and technical requirements) followed by several questions, and each question tests only one or two of those requirements. The rest is real, but it belongs to a different question. Rereading the whole brief for every question burns the clock; building one requirements table and matching each question to its rows does not.

The one rule for reading AZ-305 stems

Read every stem as a requirements document: the right answer is the simplest design that meets every stated requirement, not the most capable one.

AZ-305 options are usually all valid Azure designs, so knowing what a service does rarely separates them. What separates them is one requirement word (a region, an RPO, an engine feature) and the qualifier that ranks whatever survives. The exam rewards matching, not maximising.

Read it like this

  • Name the design target before reading options
  • Kill each wrong option with a named requirement
  • Apply the qualifier only to the survivors

Not like this

  • Add requirements the stem never stated
  • Default to the most powerful or newest service
  • Reread the whole case brief for every question

The 5-step decode routine

Run the same five steps on every case study. The time on each step is a target for a practised reader, not a rule; the order is the part that matters.

  1. ~15s

    Name the design target

    Read the last sentence first. Are you recommending a database tier, an entry point, a backup design or an identity control? That target decides which requirements matter and which belong to someone else's question.

  2. ~45s

    Pull the matching requirement rows

    In a case study, spend about four minutes on the first question turning the business and technical requirements into numbered rows tagged data, network, identity, continuity or governance. For every later question, read only the rows tagged for this target. In a standalone item, underline the constraints.

  3. ~15s

    Strike the noise

    Cross out details that do not change the answer: company size, office count, the identity setup in a networking question, tagging rules in a database question. In a case study they are real requirements, just not for this question.

  4. ~30s

    Eliminate on hard constraints

    Remove each option by naming what it breaks: scope (zone, region or global), layer (4 or 7), an engine feature (SQL Server Agent, cross-database queries) or a number (RPO, retention days, size limit). If you cannot name the requirement, the option stays.

  5. ~20s

    Rank with the qualifier

    Only now apply 'minimize administrative effort' (managed service, fewest components) or 'minimize costs' (cheapest tier that still passes). For Choose-two items, check whether each answer is part of the solution or a complete one.

Anatomy of a AZ-305 case study

Every highlighted phrase below is pinned to a note. Four kinds of phrase matter:

  • Constrainta hard requirement that eliminates options
  • Qualifierthe word that ranks the survivors
  • Keywordpoints at a specific service, control or concept
  • Noiselooks important, changes nothing
Scenario 1 · Design data storage solutionsSingle answer

Case study: picking the database tier

Overview: Corvane Logistics books freight for 3,000 shippers from two on-premises datacenters. Existing environment: The booking app uses three SQL Server 2019 databases on one instance. Stored procedures join tables across all three databases by using three-part names1, and 40 SQL Server Agent jobs run nightly invoicing2. Staff sign in with Active Directory accounts that are synced to Microsoft Entra ID3. Planned changes: Corvane will move the booking app to Azure and close one datacenter. Technical requirements: The database tier must not require Corvane staff to patch an operating system or SQL Server4. Existing database code must not be rewritten5. Branch offices will connect to Azure by using site-to-site VPN. Every Azure resource must have a CostCenter tag6. Business requirements: Minimize administrative effort7 for all new Azure services. Question: You need to recommend a database service for the booking app.

Which database service should you recommend? The solution must meet the technical requirements and MINIMIZE administrative effort.

  1. 1Keyword. Three-part-name queries across databases on one instance work in SQL Managed Instance and SQL Server, but not in Azure SQL Database.
  2. 2Keyword. SQL Server Agent runs in Managed Instance and in SQL Server on a VM. Azure SQL Database has no Agent, so the jobs would need rewriting.
  3. 3Noise. A real requirement for this case's identity questions. It has no bearing on which database engine you choose here.
  4. 4Constraint. Eliminates SQL Server on Azure VMs: whatever automation you add, patching the guest operating system remains the customer's job.
  5. 5Constraint. Blocks the workarounds for SQL Database: elastic query and elastic jobs both mean changing code that must stay as it is.
  6. 6Noise. A governance requirement for another question. Every candidate service supports tags, so it eliminates nothing here.
  7. 7Qualifier. Ranks the survivors. A managed service beats a VM, but only among options that already pass both technical constraints.
  1. AAzure SQL Database single databases in an elastic poolOutLeast administration on paper, but Azure SQL Database does not support three-part-name cross-database queries or SQL Server Agent. Making it work means elastic query and elastic jobs, which breaks the no-rewrite constraint.
  2. BAzure SQL Managed InstanceAnswerSupports cross-database queries within the instance and SQL Server Agent, so the code moves as it is. Microsoft patches the platform and the engine, which meets the no-patching constraint and the qualifier.
  3. CSQL Server 2019 on Azure virtual machines registered with the SQL IaaS Agent extensionOutFully compatible, which is why it tempts. But the guest operating system stays Corvane's to patch, so it fails a technical constraint before the qualifier is even applied.
  4. DAzure SQL Database HyperscaleOutHyperscale answers a size and scale question (up to 128 TB per database) that this case never asks. It is still Azure SQL Database, so cross-database queries and SQL Server Agent are missing.

The lesson: In a case study, tie each question to its one or two requirement rows; the identity and tagging rules are real, but they belong to other questions.

Two more: decode them yourself

Read each one as you would on the day: mark the constraints, find the qualifier, cross out what fails, and commit to a letter before you open the decode.

Scenario 2 · Design infrastructure solutionsSingle answer

One entry point for two regions

A retailer runs its storefront on Azure App Service in East US and West Europe, with both regions serving traffic. Users must be sent to the closest healthy region, and traffic must shift away from a failed region automatically. Requests for /api/* must go to a separate App Service app. Every request must pass a web application firewall that uses a Microsoft-managed rule set before it reaches either region. The team deploys both regions from one GitHub Actions workflow. You need to recommend an entry point that minimizes administrative effort.

Which solution should you recommend to meet the requirements while MINIMIZING administrative effort?

  1. AAzure Front Door Premium with a WAF policy and a route for /api/*
  2. BAzure Traffic Manager with performance routing in front of an Application Gateway WAF v2 in each region
  3. COne Application Gateway WAF v2 in East US with both regions' apps in its backend pool
  4. DA global-tier (cross-region) Azure Load Balancer in front of both regions
Decode it and show the answer

A retailer runs its storefront on Azure App Service in East US and West Europe1, with both regions serving traffic. Users must be sent to the closest healthy region2, and traffic must shift away from a failed region automatically3. Requests for /api/* must go to a separate App Service app4. Every request must pass a web application firewall that uses a Microsoft-managed rule set5 before it reaches either region. The team deploys both regions from one GitHub Actions workflow6. You need to recommend an entry point that minimizes administrative effort7.

  1. 1Keyword. Two regions of a PaaS web app call for a global HTTP entry point, not a balancer that lives inside one virtual network.
  2. 2Constraint. Latency-based routing with health checks. Front Door and Traffic Manager do this; a single regional Application Gateway cannot.
  3. 3Constraint. The entry point itself must survive losing either region, which rules out any design anchored in one region.
  4. 4Constraint. URL path routing is a layer-7 job. Azure Load Balancer and Traffic Manager cannot read request paths on their own.
  5. 5Keyword. Managed WAF rules exist on Application Gateway WAF v2 and Front Door Premium. Front Door Standard supports custom rules only.
  6. 6Noise. Deployment tooling has no effect on which traffic service fits. Every option can ship from the same workflow.
  7. 7Qualifier. Several designs pass the constraints. This ranks them: one global service beats a chain of several resources.
  1. AAzure Front Door Premium with a WAF policy and a route for /api/*AnswerOne global layer-7 service covers everything: latency-based routing with health probes, a route that sends /api/* to the API app, and managed WAF rules at the edge. Premium is needed because the stem asks for a managed rule set.
  2. BAzure Traffic Manager with performance routing in front of an Application Gateway WAF v2 in each regionOutIt works: Traffic Manager picks the region, and each gateway routes /api/* and runs the WAF. But that is three resources and two WAF policies to operate, so the qualifier removes it.
  3. COne Application Gateway WAF v2 in East US with both regions' apps in its backend poolOutApplication Gateway is regional. If East US fails, the entry point fails with it, and European users are never routed to their closest region first.
  4. DA global-tier (cross-region) Azure Load Balancer in front of both regionsOutLayer 4 only, so there is no path routing and no WAF. Its backend pool holds regional load balancers, not App Service apps.

The lesson: When more than one design passes every constraint, 'minimize administrative effort' means the fewest services to run, so a working multi-product chain is the trap.

Scenario 3 · Design business continuity solutionsChoose two

Regional failover plus a year of restores

A manufacturer runs an order system on eight Azure VMs with managed disks in West US 2. If West US 2 becomes unavailable, the system must be running in another Azure region within 2 hours with no more than 15 minutes of data loss. Auditors also require the ability to restore any VM as it was on any day in the past 12 months. Users sign in to the order system with Microsoft Entra ID. You need to recommend a solution. Each correct answer presents part of the solution.

Which TWO actions should you recommend? Each correct answer presents PART of the solution.

  1. AEnable Azure Site Recovery replication of the VMs to a second Azure region
  2. BBack up the VMs with Azure Backup using a daily policy that keeps daily recovery points for 365 days
  3. CUse an Azure Backup Enhanced policy every 4 hours in a geo-redundant vault, with Cross Region Restore as the recovery method
  4. DRedeploy the VMs across availability zones in West US 2
  5. ESet the Site Recovery recovery point retention to 365 days
Decode it and show the answer

A manufacturer runs an order system on eight Azure VMs with managed disks in West US 21. If West US 2 becomes unavailable, the system must be running in another Azure region within 2 hours2 with no more than 15 minutes of data loss3. Auditors also require the ability to restore any VM as it was on any day in the past 12 months4. Users sign in to the order system with Microsoft Entra ID5. You need to recommend a solution. Each correct answer presents part of the solution6.

  1. 1Keyword. IaaS VMs put Azure Site Recovery and Azure Backup in play. Managed disks matter: Site Recovery keeps at most 15 days of recovery points.
  2. 2Constraint. A cross-region recovery time objective. It needs a standing replica you fail over to, not a restore started after the outage.
  3. 3Constraint. A 15-minute RPO. Site Recovery writes crash-consistent points every 5 minutes; the Azure Backup Enhanced policy bottoms out at 4 hours.
  4. 4Constraint. A retention requirement, not a recovery one. Site Recovery points expire within 15 days, so this needs a backup policy.
  5. 5Noise. The sign-in method changes nothing about how disks are replicated or backed up. No option depends on it.
  6. 6Qualifier. The two answers combine. Expect each correct option to cover a different requirement, not two complete alternatives.
  1. AEnable Azure Site Recovery replication of the VMs to a second Azure regionAnswerCovers recovery: continuous replication with crash-consistent points every 5 minutes meets the 15-minute RPO, and Site Recovery's RTO SLA is one hour from the moment you trigger failover.
  2. BBack up the VMs with Azure Backup using a daily policy that keeps daily recovery points for 365 daysAnswerCovers retention: one restorable point for every day of the past year. Azure VM backup policies allow daily retention well beyond 365 days.
  3. CUse an Azure Backup Enhanced policy every 4 hours in a geo-redundant vault, with Cross Region Restore as the recovery methodOutFails the 15-minute RPO: the Enhanced policy's shortest interval is 4 hours. Cross Region Restore is a restore you start after the outage, not a replica ready to fail over.
  4. DRedeploy the VMs across availability zones in West US 2OutZones protect against a datacenter failure inside West US 2. The requirement is surviving the loss of the whole region, and zones add nothing for 12-month retention.
  5. ESet the Site Recovery recovery point retention to 365 daysOutNot possible: with managed disks the oldest usable Site Recovery point is 15 days. Replication exists for failover, not for a year of history.

The lesson: Split the requirements into recovery (RTO, RPO) and retention before reading options; one service rarely covers both, which is why the item says part of the solution.

Now do it at exam speed

The AZ-305 app has 1,600+ exam-style questions with an explanation for every option. Read each stem with the routine above and check whether your elimination matches the explanation.

App StoreGoogle PlayFree AZ-305 practice test

Keyword decoder: what the stem is telling you

These phrases recur in AZ-305 stems. Each one narrows the answer before you read the options.

When the stem saysIt points toUsually the wrong pick
Data
cross-database queries, SQL Server Agent jobs, no OS to patchAzure SQL Managed Instance: instance-level SQL Server features, platform patched by MicrosoftAzure SQL Database, which has no Agent or three-part-name queries
needs OS-level access or a third-party agent on the database serverSQL Server on Azure VMs, the only option that exposes the OSSQL Managed Instance, which gives no OS access
single database expected to grow past 32 TBAzure SQL Database Hyperscale, up to 128 TB per databaseSQL Managed Instance, whose storage tops out at 32 TB
many databases with unpredictable, uneven usage; minimize costsAn elastic pool sharing one set of resources across the databasesSizing every single database for its own peak
accept writes in several regions with low latencyAzure Cosmos DB with multi-region writes enabledSQL geo-replication, which keeps one writable primary
rarely read; retrieval within hours is acceptableArchive tier, with a lifecycle management rule moving blobs thereLeaving the blobs in Hot or Cool indefinitely
must not be modified or deleted for N years (WORM)Blob immutability: a locked time-based retention policySoft delete or RBAC, which an administrator can undo
Network
web app in several regions; WAF at the edge; route by pathAzure Front Door; Premium for managed WAF rules or Private Link originsApplication Gateway, which is regional
one region; TLS termination; URL path routing; WAFApplication Gateway WAF v2Azure Load Balancer, which cannot read URLs
non-HTTP endpoints in several regions; DNS-based failoverAzure Traffic ManagerFront Door, which handles HTTP(S) only
reach the PaaS service on a private IP; disable public accessPrivate endpoint (Azure Private Link) in the virtual networkService endpoint, which still targets the service's public endpoint
private connection to on-premises that does not cross the internetExpressRouteSite-to-site VPN, encrypted but over the public internet
Messaging
process in order; detect duplicates; transactionsService Bus queues or topics, with sessions for FIFOEvent Grid, which does not guarantee ordering
react when a blob is created or a resource changesEvent Grid pushing events to a handlerA scheduled job polling for changes
millions of telemetry events per second; replay the streamEvent Hubs, with Capture to keep the raw streamService Bus queues, built for high-value messages
Continuity
run in another region after an outage; RPO in minutesAzure Site Recovery replication to a second regionAzure Backup with Cross Region Restore
restore as it was on any day in the past yearAzure Backup policy with long-term retentionSite Recovery, whose recovery points last 15 days at most
survive a datacenter failure within the regionAvailability zones or zone-redundant servicesA second region, which costs more than the requirement asks
Identity and governance
prevent or audit resource settings across all subscriptionsAzure Policy assigned at a management groupRBAC, which controls who can act, not which configurations are allowed
app reads Key Vault secrets without stored credentialsManaged identity with a Key Vault data-plane roleService principal with a client secret in app settings
just-in-time, time-bound admin access with approvalMicrosoft Entra Privileged Identity Management (PIM)A permanent Owner or Contributor assignment
send platform logs to a third-party SIEMDiagnostic settings streaming to Event HubsA storage account archive the SIEM cannot read as a stream

The qualifier words

The qualifier decides between options that all work. On the AZ-305 these words carry a specific meaning:

MINIMIZE administrative effort

Among options that meet every requirement, pick the managed service with the fewest components to run, patch and script.

SQL Managed Instance over SQL Server on a VM when both support the features the app needs.

MINIMIZE costs

The cheapest design that still passes every stated requirement. Capability nobody asked for is waste, not insurance.

Front Door Standard when custom WAF rules are enough; Premium only for managed rule sets or Private Link origins.

Each correct answer presents PART of the solution

Your selections combine into one design. Each covers a different requirement and neither works alone.

Site Recovery for the cross-region RPO plus Azure Backup for 12-month retention.

Each correct answer presents a COMPLETE solution

Every selection must satisfy all requirements on its own. They are alternatives, not pieces.

A system-assigned or a user-assigned managed identity, either of which lets an app read a secret without stored credentials.

Does this solution meet the goal?

Judge only the one proposal shown against every requirement. A partial fix or just the first step is No.

A proposal that adds availability zones for a region-loss requirement: No.

LEAST privilege

The narrowest built-in role at the narrowest scope that still allows the task.

Key Vault Secrets User on one vault, not Contributor on the resource group.

How the wrong answers are built

Distractors are not random. Most wrong options on the AZ-305 fall into one of these patterns:

  1. The IaaS fallback

    The tellA VM-based option that supports everything because you can install anything on it.

    Why it temptsIt never fails a compatibility check, so it feels safe. It loses on 'minimize administrative effort' or on a no-patching constraint.

  2. The wrong-scope twin

    The tellRight capability, wrong reach: zones for a region outage, Application Gateway for a multi-region app, LRS where geo-redundancy was asked for.

    Why it temptsThe service matches the requirement word, so readers stop before checking zone, region or global scope.

  3. The working chain

    The tellTwo or three services wired together where one service covers every requirement.

    Why it temptsIt is a real reference architecture and it works, so it survives elimination. Only the qualifier removes it.

  4. The keyword echo

    The tellAn option that repeats a word from the stem: Event Grid for 'events', Azure Backup for 'disaster recovery'.

    Why it temptsWord matching feels like evidence. Check the capability the requirement needs (ordering, RPO, retention), not the label.

  5. The other question's answer

    The tellIn a case study, an option that meets a real requirement from the brief, just not the one this question tests.

    Why it temptsYou read that requirement two minutes ago, so it feels relevant. Map the question to its requirement rows first.

Time budget by question type

Question typeBudgetTactic
Standalone single-answer design questionabout 1.5 minRead the ask first, eliminate on constraints, then apply the qualifier. If two options still survive after 2 minutes, mark it and move on.
Multi-part items (hot area, drag and drop, dropdowns)2-3 minMicrosoft scores most multi-part questions one point per correct part, so lock in the parts you know before working on the hard one.
Case study (brief plus several questions)4-5 min for the brief, then about 2 min per questionBuild the requirements table on the first question and reuse it. Check the review screen before leaving: once you move on, you cannot return.
Problem-solution sets (Does this meet the goal?)about 1 min eachIf your exam has them, each answer is final and they do not appear on review. Judge the one proposal against every requirement; partial means No.

How to practise reading, not just content

Knowing the material is half of it. These drills train the reading:

Five-minute requirements table

Take any long case from your practice material, set a 5-minute timer and turn the requirements into numbered rows tagged data, network, identity, continuity or governance. Then answer each question by citing row numbers only, without rereading the brief.

Name the killer

For every practice question, before checking the key, write one line per wrong option naming the exact requirement or qualifier that eliminates it. If you cannot name it, you guessed, even when you were right.

Flip-word cards

For each close set (SQL Database, Managed Instance, SQL on a VM; Front Door, Application Gateway, Traffic Manager, Load Balancer; Event Grid, Event Hubs, Service Bus; Backup, Site Recovery) write the one requirement word that flips the choice. Drill until each takes under five seconds.

The AZ-305 at a glance

QuestionsNot published for AZ-305; Microsoft says most exams have 40-60
TimeNot listed per exam; 100 min for role-based exams without labs, 120 min if labs may appear
Passing score700 on a scaled 1-1,000 score (not 70% correct)
Question typesNot disclosed per exam; Microsoft's sandbox demos case studies, multiple choice, drag and drop, hot area, build list and labs
Going back / flaggingMark and review within a section; once you leave a case study or section, or take a break, seen questions lock
FeeUS$165 typical for expert exams; varies by country or region
DeliveryPearson VUE: test center or online proctored

Practise the reading, not just the facts

Work AZ-305 questions by domain on the app or try the free practice test. After each miss, find the phrase in the stem you read past.

App StoreGoogle PlayFree AZ-305 practice test

FAQ

Are these real AZ-305 exam questions?

No. The three scenarios on this page are original. They were written to the skills outline Microsoft published for the April 17, 2026 version of AZ-305, not copied from the exam, a dump site, a prep vendor or Microsoft's own samples. Real exam content is confidential under Microsoft's candidate agreement. Use these to practise how to read a stem, and check technical details against Microsoft Learn.

How do case studies work on AZ-305?

Microsoft's exam FAQ says you can refer to the case details as often as you like while answering that case's questions, and you can review and change those answers until you move to the next case or section. Once you leave a case study you cannot return to it. Case studies are not timed separately, and the introduction screen tells you how many your exam contains. Taking a break during a case also locks every question you have already seen.

Can I use Microsoft Learn during the exam?

Yes. Microsoft's exam experience page says associate and expert role-based exams give you learn.microsoft.com in a split screen, except Q&A, practice assessments and your profile. No extra time is added and the clock keeps running. Microsoft itself warns that looking up every question will run you out of time, so save it for a hard limit you cannot recall, such as a retention maximum or a tier feature.

How many questions are on AZ-305 and how long do I get?

Microsoft does not publish a question count for AZ-305; its general guidance is that most exams have 40-60 questions. It lists 100 minutes for role-based exams without labs and 120 minutes for those that may contain labs, and your exact exam time is shown when you register. Passing is 700 on a scaled score of 1 to 1,000, which Microsoft notes is not the same as 70 percent correct.

Does 'minimize administrative effort' always mean the PaaS option?

Usually, but only among options that already meet every requirement. If the stem needs OS-level access, a third-party agent on the server or an engine feature the PaaS service lacks, the PaaS option is eliminated before the qualifier is applied, and the VM wins. The qualifier ranks the survivors; it never revives an option that broke a constraint.

Sources

Figures on this page come from Microsoft and the other primary sources below:

Checked October 14, 2026. Fees, outlines and eligibility rules change — confirm with Microsoft before you book.