You are ready for the Databricks Certified Associate Developer for Apache Spark exam when you can write and read PySpark from memory across all seven exam sections, and you score 70 percent or better on timed, full-length mocks with no section far behind. Seventy percent is Certify's own practice target: Databricks publishes no passing score.
Last updated October 2026.
The exam is 45 scored multiple-choice questions in 90 minutes for USD 200, online proctored or at a test center, with every code snippet in Python. Unscored questions may also appear, unmarked. The detail that changes how you should judge readiness is in the official guide: no test aids, including API documentation. For the full format and the seven sections, see the complete certification guide.
Many questions show four code snippets where three use a real PySpark method with the wrong argument, the wrong order or the wrong mode. Feeling comfortable in a notebook with autocomplete is not the same signal. The honest test is whether you can say, unprompted, which of coalesce and repartition shuffles, or which save mode is the default, and why.
Format, cost, delivery and the no-test-aids rule are from the official Databricks exam guide (30 October 2025 edition) and certification page. The guide publishes no passing score.
The readiness quiz groups the seven official sections into five areas. For each, the signal is a decision you can make without looking anything up. The weights are the summed official section weights.
| Area | Weight | You are ready when you can |
|---|---|---|
| Architecture | 20% | Trace one action to a job, stages split at shuffles and one task per partition; say what spark.sql.shuffle.partitions controls (200 by default) and which storage level cache() uses (MEMORY_AND_DISK). |
| Spark SQL | 20% | Write an overwrite with partitionBy from memory, name all four save modes and the errorifexists default, query a Parquet path directly with SQL, and know a global temporary view lives in global_temp. |
| DataFrame API | 30% | Choose dropDuplicates over distinct for a key subset, use the right join how string (including left_anti), know that union matches by position and keeps duplicates, and say when a broadcast join applies (the 10 MB threshold). |
| Troubleshooting and Tuning | 10% | Pick coalesce to reduce partitions and repartition to rebalance, spot skew as one long task in the Spark UI, explain what AQE changes at runtime, and know whether an OOM belongs in the driver or an executor log. |
| Streaming, Spark Connect and Pandas API on Spark | 20% | Choose append, update or complete output mode, always set a checkpoint location, bound deduplication state with a watermark, place the driver in local, client or cluster mode, name what a Spark Connect client cannot do (RDDs, SparkContext), and prefer a Pandas UDF over a row-at-a-time Python UDF. |
Section weights (20/20/30/10/10/5/5) from the official Databricks certification page; the last area combines Structured Streaming (10%), Spark Connect (5%) and Pandas API on Spark (5%). For a deeper look, see the seven sections and their weights.
registerTempTable, or list Mesos as a cluster manager will teach answers the current exam does not use.withColumnRenamed takes the old name first, the code-reading questions will cost you time and points.The tell: if you cannot name your weakest of the five areas and the chapter behind it, find that out before you book.
The free Spark Developer Associate readiness quiz is 12 questions across the five areas above, with no timer and no signup. At the end it shows how you scored in each area against how much that area weighs on the exam, and which units to study next. Each question targets one common mix-up, such as repartition versus coalesce or append versus complete output mode, so a miss names a specific gap.
It is a diagnostic, not a timed practice exam. Its job is to route you to the right unit of the course, not to simulate exam day.
Because Databricks publishes no cut score, treat 70 percent as our guidance, not an official line. A sequence that works:
If the readiness quiz shows gaps in the core, plan on the longer timelines in how long to prepare for the exam before booking anything.
You are ready when you can write PySpark from memory across all seven sections and score 70 percent or better on timed full-length mocks with no area far behind. Seventy percent is Certify's practice target, since Databricks publishes no passing score.
Databricks does not publish a passing score for this exam. Certify uses 70 percent as its own practice target, and you should confirm current details in the official exam guide when you register.
Book once you pass the section exams at 70 percent or better and are scoring there on timed mocks, then pick a date one to two weeks out. Reread the official exam guide when you book, since Databricks asks candidates to check it for updates about two weeks before the exam.
It is harder than daily work suggests, because the exam allows no test aids, including API documentation. Many questions compare four code snippets that differ by one argument or method name, so recall of exact names and defaults matters.
The DataFrame API section, at 30 percent of the exam, followed by Architecture and Spark SQL at 20 percent each. Do not skip the small sections: Structured Streaming, Spark Connect and Pandas API on Spark add up to 20 percent together.
Twelve questions across all five areas of the Spark Developer Associate exam. No timer, no signup. You get a per-area readout and exactly what to study next.
Take the readiness quiz →