Published 20 November 2026 · 7 min read
How this plan is built
The plan assumes about eight hours of study a week for six weeks, 48 hours in total. That figure is our assumption, not a vendor recommendation. The vendor recommends about a year of hands-on experience with the tasks in the exam outline, so if you have that, this plan is mostly a way to find gaps and rehearse. If you have less, lengthen it.
Time is divided in proportion to the weights in the 9 October 2026 exam guide: 23, 15, 12, 12, 10, 10, 8, 5 and 5 percent. Your own gaps matter more than the weights, so let the diagnostic overrule the plan wherever they disagree.
The hours, by domain
Applying those weights to 48 hours gives the following rough budget. Round it however suits you.
- Developing Code for Data Processing using Python and SQL (23 percent): about 11 hours.
- Cost & Performance Optimization (15 percent): about 7 hours.
- Data Ingestion & Acquisition (12 percent): about 6 hours.
- Data Manipulation (12 percent): about 6 hours.
- Monitoring and Alerting (10 percent): about 5 hours.
- Debugging and Deploying (10 percent): about 5 hours.
- Ensuring Data Security and Compliance (8 percent): about 4 hours.
- Data Governance (5 percent): about 2 hours.
- Data Modeling (5 percent): about 2 hours.
Week 1: baseline and the biggest domain, part one
Start with a diagnostic before you read anything. It will feel uncomfortable and it is meant to. It tells you which parts of the plan to shorten. Then read the exam guide's objectives for the first domain and begin with the topics where the diagnostic was weakest.
The first domain covers Python project structure for bundles, troubleshooting library installation, user-defined functions, and unit and integration testing. Write code rather than reading about it: structure a small project, break a dependency on purpose, write a function and test it.
- About one hour on the diagnostic and reviewing every explanation.
- About six hours on the first half of the largest domain.
Week 2: the biggest domain, part two
The second half of the largest domain is the heavy end: streaming pipelines, change data capture with slowly changing dimensions, choosing between streaming tables and materialized views, Lakeflow Jobs and control flow, choosing compute, and the stateful semantics of Structured Streaming such as watermarks, output modes and checkpoints.
These are decision topics. The exam will ask which option fits a set of constraints, so for each one practise stating the trade-off in a sentence. If you cannot explain why you would not pick the alternative, you do not know it yet.
- Five hours reading and building.
- Two or three hours on topic practice questions, reviewing every explanation.
Week 3: ingestion and manipulation
Two 12 percent domains share this week. Ingestion covers file formats and sources, change data capture from relational databases, sharing, clean rooms and federation. Manipulation covers window functions, joins, aggregations, semi-structured data, AI functions inside pipelines and data quality expectations.
For manipulation, work from the data. Take a messy dataset and write the transformation twice, once in SQL and once in PySpark, and compare. For ingestion, draw the architecture for each source type on paper, then check it against the documentation.
Week 4: optimization and data modeling
Optimization is the second-heaviest domain and rewards understanding over memory. Be able to say when each table-layout technique fits an access pattern, what change data feed gives you, and how to read a query profile for skew, shuffles and poor data skipping. Data modeling is small, but its table-layout topics overlap with optimization, so study them together.
Do your first set of mixed practice at the end of the week, and spend longer on the explanations than on the questions.
Week 5: monitoring, debugging and deploying
Monitoring and Alerting and Debugging and Deploying together carry 20 percent. Both are practical: system tables, event logs, alerts, the Spark UI, job repair and parameter overrides, bundle deployment and Git-based CI/CD. If you do not use these in your day job, run through them in a workspace you have access to, because the exam describes situations and expects you to know what you would look at first.
Finish the week by retrying every question you have missed so far. Spaced repetition does this for you on Pro and Unlimited, and on other plans you can use the review-incorrect mode.
Week 6: security, governance, then rehearsal
Security and compliance (8 percent) and governance (5 percent) are small and well defined. Cover access control lists, attribute-based policies, anonymisation methods, PII handling, deletion obligations, tags and permission inheritance in two or three sessions.
Keep the rest of the week for rehearsal. Sit one timed mock exam under exam conditions: 120 minutes, no notes, no pauses. Review every result, including the ones you got right by guessing. Spend the remaining days on the objectives where the mock exam exposed gaps, and stop heavy study the day before the real exam.
Adjusting the plan
If you have only three weeks, keep the weights and halve every block, and put more of the time into practice questions. If you work with these tools every day, shorten the reading and lengthen the practice. In either case, we cannot tell you whether the plan will be enough for you. Nobody can, and the vendor publishes no passing score to aim at. What the plan can do is make sure your effort follows the published blueprint.