Every thread asking how hard is 350-901 gets the same two replies: "it's a professional-level Cisco exam, so hard" and "easy if you're a real developer." Both are useless, and the second one is what got me.
I write software for a living. I assumed DEVCOR would be a formality with some Cisco product names sprinkled on top. I failed, and the post-mortem was uncomfortable enough to be worth writing down.
The exam, factually
- 90–110 questions, roughly 120 minutes
- Multiple choice, multiple response, drag-and-drop, and scenario items
- $400 USD
- Scaled cut score — Cisco does not publish the pass mark, so there's no "I need 72%" target to aim at
- It's a concentration exam for CCNP DevNet / DevNet Professional, paired with the 350-901 core
Check Cisco's current exam topics page before you build a plan — Cisco revises blueprints and the topic list is the contract.
The unpublished cut score matters psychologically. You cannot calculate a safe margin, so the only sane strategy is to be comfortably strong across every topic area rather than banking on carrying a weak domain.
Where I actually lost it
I studied the APIs and ignored the platforms.
Here's the trap for developers. You look at the blueprint, see "software development and design," "using APIs," "application deployment and security," "infrastructure and automation," "network fundamentals," and you think: I know REST, I know OAuth, I know Docker, I know CI/CD, I know Git. This is my home turf.
And it is — for maybe 60% of the exam. The rest is Cisco-platform-specific, and it is not guessable from first principles. Knowing REST does not tell you the shape of a Meraki Dashboard API response, or how a Webex Bot registers a webhook, or what a DNA Center intent API path looks like, or the difference between a NETCONF and a RESTCONF interaction against IOS-XE, or where YANG models fit, or what pyATS is for.
Those questions aren't hard. They're just unlearnable by inference. You either read the docs and touched the sandboxes or you didn't. I hadn't, and I burned a $400 attempt discovering it.
The second thing I got wrong: "design" is a scored skill here
A meaningful chunk of DEVCOR is software design judgment, not code:
- When do you pick a message queue over a synchronous call?
- Which failure mode does a circuit breaker protect against, and which does a retry with exponential backoff protect against?
- How do you make an operation idempotent, and why does that matter for network automation specifically?
- Where do secrets live in a pipeline, and what's wrong with the obvious answer?
- Twelve-factor principles, applied to an app that talks to network devices
- Sequence and architecture diagrams — you'll be asked to read one and identify the flaw
Developers with strong practical habits but no vocabulary for them lose points here, because the questions test the name of the pattern as much as the instinct.
Where it's genuinely hard, ranked
Hardest: breadth. Not depth. No single topic is brutal, but the surface area spans application design, security, deployment models, multiple Cisco platform APIs, network fundamentals, and automation tooling. You can't cram breadth in a week.
Second: unfamiliar-platform questions under time pressure. ~110 questions in 120 minutes is about 65 seconds each. If three consecutive questions are on a platform you skimmed, you lose time and confidence.
Third: the scenario items. Long stems, code snippets, and answers that differ by one detail. Careless reading is expensive.
Not actually hard: the pure programming content. If you ship code, the Python, the Git, the REST, and the containers are the easy marks.
The plan that worked the second time
Weeks 1–2: live in the Cisco DevNet sandboxes. They're free. Get a token, make real calls against Meraki, Webex, DNA Center, IOS-XE RESTCONF, and NX-OS. Write throwaway scripts. The goal isn't mastery — it's that no platform in the exam is a stranger. This single change is most of the difference between my two attempts.
Week 3: design vocabulary. Go through the blueprint's design topics and, for each pattern, write two lines: what problem it solves and what it costs. Circuit breaker, bulkhead, retry/backoff, idempotency, event-driven versus request-response, blue/green versus canary, secrets management. If you can't state the trade-off, you don't know it well enough for the exam.
Week 4: security and deployment. OAuth flows in detail (which grant type, and why — this comes up), token handling, encrypting data in transit and at rest, container security basics, pipeline security. Then CI/CD models and deployment strategies.
Week 5: timed reps. This is where you find the topic areas that quietly rotted. I ran free 350-901 practice questions in timed blocks and kept a single text file of every miss with one line on why. By the end that file had about eleven recurring themes — that's the real study list, and it's much shorter than the blueprint.
Week 6: re-read the blueprint line by line and rate yourself 1–5 on each bullet. Anything under 4 gets an evening. Then book it.
Is it worth $400?
If you're a network engineer moving into automation: yes, unreservedly. DEVCOR forces you to learn software engineering properly rather than accumulating scripts, and that's the transition that changes your salary band.
If you're a developer with no network context: it's worth it only if you work near networking. Otherwise you're paying $400 to prove you can write Python, which your GitHub already does for free.
If you're chasing CCNP DevNet: it's the core exam, so the question is moot — but budget six weeks, not two.
The one-sentence answer
350-901 is hard in proportion to how much Cisco platform surface you've genuinely touched, and easy in proportion to how much software you genuinely write — and most candidates are strong on exactly one of those two axes, which is why the failure stories all sound like mine.
Topic breakdown, current cost, and free practice questions: ExamCert's 350-901 page.
Go make real API calls. That's the whole tip.
Top comments (0)