There's a pattern I've watched play out a dozen times now. Someone passes AZ-104, feels good, sees "Expert" on the AZ-305 badge, and books the exam six weeks later. They fail. Then they post on Reddit that the case studies were "unfair" and the questions were "ambiguous."
The questions weren't ambiguous. They were architectural. And that's a different thing entirely, which is exactly the point most people miss.
So here's my unpopular opinion, stated plainly: AZ-305 is the best-designed Azure exam Microsoft ships, and most of the people rushing into it have no business sitting it yet. Not because they're dumb. Because the exam tests a skill they haven't had the job experience to build.
The exam is not testing whether you know Azure
This is the thing that breaks people's brains. You can know every Azure service cold — every SKU, every tier, every CLI flag — and still fail AZ-305. I've seen it happen to genuinely strong engineers.
Because AZ-305 doesn't ask "what does Azure Front Door do?" It asks: given this company, with this compliance requirement, this budget, this existing on-prem footprint, and this RTO — what do you build? Every option in the answer list is a real Azure service that would technically work. Three of them are wrong because of a constraint buried in paragraph four of the case study.
That's not a knowledge test. That's a judgement test. And judgement is the one thing you cannot cram.
Here's the shape of the thing, for people who've only skimmed the blueprint:
| Attribute | Detail |
|---|---|
| Questions | ~40–60 |
| Time | ~120 minutes |
| Passing score | 700 / 1000 (scaled) |
| Cost | $165 USD |
| Question types | Case studies, MCQ, drag-and-drop, hot area |
| Prerequisite | None formally (AZ-104 strongly assumed) |
| Renewal | Free online assessment, annually |
Note the question count: forty to sixty. That's low. AZ-104 throws more at you. AZ-305 gives you fewer questions but each one carries more weight and more reading. Two full case studies can eat 40 minutes if you're not disciplined about it.
The four domains, and what they actually mean
Microsoft's current outline splits it like this:
- Design identity, governance, and monitoring solutions — 25–30%
- Design data storage solutions — 20–25%
- Design business continuity solutions — 15–20%
- Design infrastructure solutions — 30–35%
Read those verbs. Every single one says design. Not "configure," not "deploy," not "manage." AZ-104 is a configure-and-manage exam. AZ-305 is a decide-and-defend exam. Same services, completely different cognitive mode.
Take the identity domain. AZ-104 asks you to add a user to a group. AZ-305 asks whether you should use Entra ID Conditional Access with named locations or federate to an existing AD FS estate given that the parent company has a merger closing in eight months. Both questions involve "identity." Only one of them requires you to have been in the room when someone made that call badly.
The business continuity domain is where I see the most carnage. Fifteen to twenty percent of the exam, and it's almost entirely about RPO and RTO tradeoffs — geo-redundant storage versus zone-redundant, Azure Site Recovery versus native database failover, active-active versus active-passive. If you've never had to explain to a CFO why the cheaper DR option means eight hours of data loss, these questions feel arbitrary. If you have, they read like Tuesday.
Why the "just take it after AZ-104" advice is wrong
The internet consensus is roughly: pass AZ-104, do a course, take AZ-305. And that path works for a meaningful chunk of people — usually the ones who already do architecture work and just needed the badge.
For everyone else it produces a specific failure mode I've come to recognise. They memorise decision trees. "If compliance requires data residency → use this. If RPO is zero → use that." And it kind of works, until the case study presents two constraints that pull in opposite directions and there's no tree branch for it. Real architecture is constraint arbitration. Flashcards can't teach arbitration.
My actual position: if you haven't spent at least a year making infrastructure decisions that someone else had to live with, you're not studying for AZ-305 — you're studying for a simulation of it. That doesn't mean don't take it. It means know what you're doing, and build the missing experience deliberately instead of pretending you already have it.
So what does "deliberate" look like?
Concretely, three things I'd tell any engineer heading into this exam.
One: read the Azure Architecture Center like it's the actual textbook. Not the Learn modules — the reference architectures and the Well-Architected Framework pillars. AZ-305 questions are essentially WAF pillar tradeoffs in costume. Cost optimisation versus reliability. Security versus operational excellence. The exam authors clearly write from that mental model, and once you notice it, the "ambiguous" questions stop being ambiguous.
Two: practise the constraint-hunting, not the recall. When you drill questions, don't ask "did I get it right." Ask "which sentence in the scenario made the other three options wrong?" There is always one. Sometimes it's a single clause — "the solution must minimise administrative effort" or "the company cannot modify the application code." That clause is doing all the work. Train yourself to find it before you read the answers. This is the single highest-leverage habit for this exam and almost nobody does it.
Three: use practice questions as a diagnostic, not a scorecard. Run a free AZ-305 practice test early — like, before you feel ready — specifically so you can see which domain your judgement is thinnest in. Most people discover it's business continuity or governance, because those are the two areas engineers rarely touch until they're senior. Then spend your study time there instead of re-reading the storage content you already know. I'd rather someone bombed a practice set in week one and fixed the gap than felt comfortable for six weeks and got surprised on exam day.
Realistic timeline, honestly stated: 8 to 12 weeks for most working engineers with AZ-104-level knowledge. Faster if you already architect. Much slower if you're trying to learn architecture and pass the exam simultaneously — and if that's you, that's fine, just budget for it rather than booking the seat in six weeks and calling it motivation.
The part where I say something nice
I'm hard on the rush-into-it crowd, but I want to be clear about why: it's because AZ-305 is worth respecting. Most certs are vocabulary tests with a timer. This one actually approximates the job. The case studies are irritating in the exact way that real requirements documents are irritating — buried constraints, contradictory stakeholders, a budget line that ruins your elegant design.
Passing it means something. Which is precisely why cramming for it is a waste of a genuinely good exam.
Take it when you've earned the judgement it's testing. And if you're not there yet — go get the judgement first. The badge will still be there. Start by finding your weak domain with some free Azure AZ-305 practice questions, then go read the reference architecture that covers it. That loop, repeated, is the whole strategy.
Top comments (0)