Databricks Databricks-Certified-Data-Analyst-Associate Databricks Certified Data Analyst Associate Exam Exam Practice Test

 Databricks SQL is a service that allows users to query data in the lakehouse using SQL and create visualizations and dashboards1. One of the common use cases for Databricks SQL is to combine data from different sources and formats into a single, comprehensive dataset that can be used for further analysis or reporting2. For example, a data analyst can use Databricks SQL to join data from a CSV file and a Parquet file, or from a Delta table and a JDBC table, and create a new table or view that contains the combined data3. This can help simplify the data management and governance, as well as improve the data quality and consistency. References:

Databricks SQL overview

Databricks SQL use cases

Joining data sources

The approach that cannot be used to share the dashboard and meet the security requirement is D. Generating a Personal Access Token that is good for 1 day and sharing it with the client. This approach would give the client access to the Databricks workspace using the token owner’s identity and permissions, which could expose other information, resources, or artifacts in the database1. The other approaches can be used to share the dashboard and meet the security requirement because:

A. Downloading the Dashboard as a PDF and sharing it with the client would only provide a static snapshot of the dashboard without any interactive features or access to the underlying data2.

B. Setting a refresh schedule for the dashboard and entering the client’s email address in the “Subscribers” box would send the client an email with the latest dashboard results as an attachment or a link to a secure web page3. The client would not be able to access the Databricks workspace or the dashboard itself.

C. Taking a screenshot of the dashboard and sharing it with the client would also only provide a static snapshot of the dashboard without any interactive features or access to the underlying data4.

E. Downloading a PNG file of the visualizations in the dashboard and sharing them with the client would also only provide a static snapshot of the visualizations without any interactive features or access to the underlying data5. References:

1: Personal access tokens

2: Download as PDF

3: Automatically refresh a dashboard

4: Take a screenshot

5: Download a PNG file

Option C is correct because the requirement is to run the SQL query and then operate on the results in PySpark. spark.sql(...) executes SQL text and returns a Spark DataFrame that can be used in PySpark tests. spark.table(...) can load a table directly, but it does not run an arbitrary SQL query. Option A is SQL text by itself, not a PySpark operation. Official Databricks documentation states that a SparkSession can be used to execute SQL over tables, and Databricks notebook documentation specifically references using SQL inside Python with a spark.sql command.

The gold layer of the medallion architecture contains data that is highly refined and aggregated, and powers analytics, machine learning, and production applications. Data analysts typically use the gold layer to access data that has been transformed into knowledge, rather than just information. The gold layer represents the final stage of data quality and optimization in the lakehouse. References: What is the medallion lakehouse architecture?

Option A is correct. Silver tables are cleaned, validated, and enriched versions of data, often still retaining detailed records. Gold tables are typically business-ready, analytics-focused, and more likely to contain aggregations, dimensional models, and reporting-ready metrics. Official Databricks extract: Silver is associated with “Data cleaning and validation,” while Gold is associated with “Dimensional modeling and aggregation.” Databricks also states that Gold data is often highly aggregated and tailored for analytics and reporting.

Option B is correct. PIVOT rotates unique values from rows into separate columns, which is exactly the process of converting long-format data into wide-format data. UNPIVOT does the reverse. SUM is an aggregate function, WHERE filters rows, and TRANSFORM is not the SQL keyword used for this reshaping task. Official Databricks extract: the PIVOT clause “rotat[es] unique values from a column into separate columns.”

Option D is correct. Cluster resizing caused by autoscaling is recorded as a cluster activity event, so the administrator should review the Cluster Event Log. Driver and executor logs are useful for application/runtime troubleshooting, not for the cluster resizing timeline. Ganglia shows metrics but is not the event timeline. Official Databricks documentation identifies cluster activity events as the place to retrieve cluster activity, and Databricks autoscaling documentation describes resize events such as “Cluster resize request started.”

The correct answer is C because the issue is a resource constraint failure, and the analyst needs to identify which operator is consuming excessive memory. In Query Profile, Memory peak shows memory usage at the operator level and helps identify the operator causing resource overuse. Time spent helps identify slow operators, shuffle read size helps analyze data movement, and bytes spilled to disk indicates spill behavior, but the most direct metric for resource overuse due to memory pressure is memory peak.

Official documentation extract used: Databricks Query Profile graph view shows metrics such as “Time spent, Memory peak, and Rows.”

Option A is correct. The model being tuned is a Spark ML model, which is already distributed. SparkTrials is intended to distribute independent single-machine trials across Spark workers. For distributed ML algorithms such as Spark MLlib/Spark ML, Hyperopt should run trials from the driver so each trial can access the full cluster resources. Therefore, SparkTrials() should be changed to Trials(). Official Databricks documentation explains that this setup works for distributed machine learning algorithms including Apache Spark MLlib, and the Databricks notebook guidance states that SparkTrials is incompatible for that distributed-training pattern because each trial must be evaluated on the driver node.

The correct answer is C because Databricks SQL Editor is used to create and save SQL queries. A saved query can later be opened, reused, shared, visualized, and added to a dashboard. A dashboard is not the saved SQL statement itself; it is a reporting object that can contain query results and visualizations. Query History stores past executions, but it is not the correct object for saving a reusable SQL statement.

Official documentation extract used: Databricks states that the SQL editor can be used to “author queries” and “create visualizations.” It also states that the Save button “saves your query,” and query results can be added to a dashboard.

Option D is correct. The scenario describes combining multiple complementary data sources into one comprehensive dataset. That is data blending. Last-mile ETL is usually project-specific final transformation near the end of an analytics workflow, while this question emphasizes combining two source datasets for broader ongoing team use. The current official Databricks exam guide describes the same kind of capability as creating unified datasets by joining data from multiple sources. That aligns with the concept of data blending. Reference: Databricks Certified Data Analyst Associate Exam Guide.

A Serverless SQL endpoint is a type of SQL endpoint that does not require a dedicated cluster to run queries. Instead, it uses a shared pool of resources that can scale up and down automatically based on the demand. This means that a Serverless SQL endpoint can start up much faster than a SQL endpoint that uses a cluster, and it can also save costs by only paying for the resources that are used. A Serverless SQL endpoint is suitable for ad-hoc queries and exploratory analysis, but it may not offer the same level of performance and isolation as a SQL endpoint that uses a cluster. Therefore, a data analyst should consider the trade-offs between speed, cost, and quality when choosing between a Serverless SQL endpoint and a SQL endpoint that uses a cluster. References: Databricks SQL endpoints, Serverless SQL endpoints, SQL endpoint clusters

Option E is correct. A major benefit of open source technologies and open data formats is avoiding vendor lock-in. Databricks supports open formats and interfaces so data can be used across tools and systems rather than being locked into a proprietary platform. Cloud integrations, governance, and workload scalability are Databricks benefits, but the specific benefit of embracing open source is avoiding vendor lock-in. Official Databricks extract: “Using open data formats and interfaces helps to avoid” vendor lock-in, and Databricks states that no proprietary data formats are used because Delta Lake and Iceberg are open source.

External tables are tables that are defined in the Databricks metastore using the information stored in a cloud object storage location. External tables do not manage the data, but provide a schema and a table name to query the data. To create an external table, you can use the CREATE EXTERNAL TABLE statement and specify the object storage path to the LOCATION clause. For example, to create an external table named ext_table on a Parquet file stored in S3, you can use the following statement:

SQL

CREATE EXTERNAL TABLE ext_table (

col1 INT,

col2 STRING

)

STORED AS PARQUET

LOCATION ' s3://bucket/path/file.parquet '

AI-generated code. Review and use carefully. More info on FAQ.

 External tables

Higher-order functions are a simple extension to SQL to manipulate nested data such as arrays. A higher-order function takes an array, implements how the array is processed, and what the result of the computation will be. It delegates to a lambda function how to process each item in the array. This allows you to define functions that manipulate arrays in SQL, without having to unpack and repack them, use UDFs, or rely on limited built-in functions. Higher-order functions provide a performance benefit over user defined functions. References: Higher-order functions | Databricks on AWS, Working with Nested Data Using Higher Order Functions in SQL on Databricks | Databricks Blog, Higher-order functions - Azure Databricks | Microsoft Learn, Optimization recommendations on Databricks | Databricks on AWS

 Option A uses the SELECT DISTINCT statement to remove duplicate rows from the table_bronze and create a new table table_silver with the deduplicated data. This is the correct way to deduplicate data using Spark SQL12. Option B simply inserts all the rows from table_bronze into table_silver, without removing any duplicates. Option C is not a valid syntax for Spark SQL, as there is no MERGE DEDUPLICATE statement. Option D appends all the rows from table_bronze into table_silver, without removing any duplicates. Option E overwrites the existing data in table_silver with the data from table_bronze, without removing any duplicates. References: Delete Duplicate using SPARK SQL, Spark SQL - How to Remove Duplicate Rows

The DROP TABLE command removes a table from the metastore and deletes the associated data files. The syntax for this command is DROP TABLE [IF EXISTS] [database_name.] table_name;. The optional IF EXISTS clause prevents an error if the table does not exist. The optional database_name. prefix specifies the database where the table resides. If not specified, the current database is used. Therefore, the correct command to remove the table table_name from the database database_name and all of the data files associated with it is DROP TABLE database_name.table_name;. The other commands are either invalid syntax or would produce undesired results. References: Databricks - DROP TABLE

Option B is correct after correcting the uploaded typo withcolumn to withColumn. A 10% tax is calculated by multiplying salary by 0.1, not by 10. The age column should be removed using .drop( " age " ). Options A and D multiply salary by 10, which calculates 1000%, not 10%. Options A and C also use dropField, which is not the correct DataFrame method for removing a top-level column. Official Databricks PySpark documentation states that withColumn adds or replaces a column and that drop(*cols) returns a new DataFrame without the specified columns.

The correct answer is C because TRANSFORM applies a lambda function to each element of an array and returns a new array. The expression value - > value / 100 means: for each element named value in dollars_spent, divide that element by 100. The result is then aliased as hundreds_spent. Options A, B, and D do not use the required lambda syntax correctly.

Official documentation extract used: Databricks states that transform returns an array and that the lambda function “produces a new value for each element.”

Databricks SQL allows you to create multiple visualizations from a single query result. These visualizations can be customized independently and each can be added to a dashboard. This feature is explicitly supported and recommended in Databricks’ documentation on dashboards and visualization workflows, enabling flexible reporting without duplicating queries.

 Last-mile ETL is the term used to describe the additional processing of data that is done by data analysts or data scientists after the data has been ingested, transformed, and stored in the lakehouse by data engineers. Last-mile ETL typically involves tasks such as data cleansing, data enrichment, data aggregation, data filtering, or data sampling that are specific to the analysis or machine learning use case. Last-mile ETL can be done using Databricks SQL, Databricks notebooks, or Databricks Machine Learning. References: Databricks - Last-mile ETL, Databricks - Data Analysis with Databricks SQL

 An external table is a table that is defined in the metastore, but its data is stored outside of the Databricks environment, such as in S3, ADLS, or GCS. When an external table is dropped, only the metadata is deleted from the metastore, but the data files are not affected. This is different from a managed table, which is a table whose data is stored in the Databricks environment, and whose data files are deleted when the table is dropped. To delete the data files of an external table, the analyst needs to specify the PURGE option in the DROP TABLE command, or manually delete the files from the storage system. References: DROP TABLE, Drop Delta table features, Best practices for dropping a managed Delta Lake table

Option C is the correct answer because it creates a new aggregated table, store_sales, from the existing cleaned/refined table cleaned_transactions. In a medallion architecture, a table such as cleaned_transactions represents a Silver-layer table because it contains cleaned or validated transaction-level data. The query then creates store_sales by grouping by store_id and calculating SUM(sales), which is an aggregation suitable for analytics and reporting. That is exactly the Silver-to-Gold hop.

Option A creates cleaned_transactions from transactions, so it is moving toward a cleaned Silver table, not creating a Gold table. Option B creates transactions from raw_transactions, which is closer to a raw-to-Bronze or Bronze preparation step. Option D filters rows from transactions into cleaned_transactions, which is also a cleaning/validation step associated with Silver, not Gold.

Exact extract from official Databricks documentation: Databricks describes medallion architecture as improving data through “Bronze ⇒ Silver ⇒ Gold layer tables.” The same official page identifies Silver as “Data cleaning and validation” and Gold as “Dimensional modeling and aggregation.” Databricks also states that the Gold layer “consists of aggregated data tailored for analytics and reporting.”

The reason the alert is not functioning as expected is because Databricks SQL Alerts do not support query parameters. This limitation applies to all types of parameters, including date parameters.

Here’s why:

Alerts require static, deterministic query results so they can compare values consistently during scheduled executions.

When a query includes parameters (e.g., a date range parameter), its results may change based on user input or the default value set in the query editor.

However, Databricks SQL Alerts will always use the default value set for the parameter at the time the alert is created. This means the alert doesn’t dynamically adapt to new date ranges and will not reflect changes unless the query is manually updated.

As a result, if the business logic behind the alert depends on changing date ranges or any user input, the alert will not trigger correctly, or may never trigger at all.

Therefore, the correct explanation contradicts Option B, which is incorrect in saying that alerts cannot work with date-based queries at all. In fact, they can—as long as the query is static (i.e., without parameters).

Option A is correct. The Data Intelligence Engine is the platform-level intelligence layer that understands the semantics and uniqueness of an organization’s data and enables AI-assisted experiences across Databricks. Unity Catalog provides governance and metadata management, Genie Spaces provide a natural-language interface for curated business data, and Databricks Assistant is a user-facing assistant for code/query help. However, the question asks which component is primarily responsible for enabling these capabilities across the platform; that is the Data Intelligence Engine. Databricks describes the platform as powered by a Data Intelligence Engine that understands the uniqueness and semantics of data and helps optimize performance. References: Databricks Data Intelligence Platform documentation and Data Analyst Associate Exam Guide.

the accounts.customers table is an EXTERNAL table, which means that it is stored outside the default warehouse directory and is not managed by Databricks. Therefore, when you run the DROP command on this table, it only removes the metadata information from the metastore, but does not delete the actual data files from the file system. This means that you can still access the data using the location path (dbfs:/stakeholders/customers) or create another table pointing to the same location. However, if you try to query the table using its name (accounts.customers), you will get an error because the table no longer exists in the metastore. References: DROP TABLE | Databricks on AWS, Best practices for dropping a managed Delta Lake table - Databricks

A data analyst can create multiple visualizations from the same query in Databricks SQL by clicking the + button next to the Results tab and selecting Visualization. Each visualization can have a different type, name, and configuration. To add a visualization to a dashboard, the data analyst can click the vertical ellipsis button beneath the visualization, select + Add to Dashboard, and choose an existing or new dashboard. The data analyst can repeat this process for each visualization they want to add to the same dashboard. References: Visualization in Databricks SQL, Visualize queries and create a dashboard in Databricks SQL

Option A is correct. The user is not a member of the auditing group, so is_member( " auditing " ) evaluates to false and the ELSE age > = 18 branch controls the filter. Rows with age > = 18 are returned; rows under 18 are omitted. “Age greater than 17” is equivalent to age > = 18 for integer ages. Official Databricks extract: is_member() “returns TRUE if the current user is a member” of the specified group, and Databricks describes dynamic views as views that can filter rows based on group membership.

Option B is correct. A data analyst commonly uses Databricks SQL for analytics-focused transformations, aggregation, joining, filtering, and preparation of datasets for reporting or dashboarding. Analysis-specific ETL on Gold-layer tables is a “last-mile” analytics task: the Gold layer is already curated for business use, and the analyst performs additional project-specific refinement. The official Databricks exam guide includes creating views, performing aggregate operations, combining tables with joins, filtering, sorting, and using dashboards as Data Analyst Associate skills. It also states that Gold-layer data drives downstream analytics and dashboards. Options A, C, and D are primarily streaming engineering, machine learning, and MLOps activities, not common Databricks SQL analytics applications. References: Databricks Certified Data Analyst Associate Exam Guide and medallion architecture documentation.

Option B is correct. The function to_date converts a timestamp or date-like expression to a date value, which matches the requirement to extract the date from the timestamp field. unix_timestamp converts to a Unix timestamp value, date_format formats a date or timestamp as a string, and from_unixtime converts Unix time into a timestamp/string representation. Official Databricks documentation states that to_date(expr [, fmt]) returns the expression cast to a date.

Option E is correct. The Depends On field is used to define task dependencies in a Databricks job, so a downstream task waits for an upstream task according to the dependency rules. The standard use case is when one task must complete successfully before the next task starts. Official Databricks extract: “You can control the execution order of tasks by specifying dependencies between them,” and Databricks runs upstream tasks before downstream tasks.

The result set provided shows a combination of grouping by two columns (group_1 and group_2) with subtotals for each level of grouping and a grand total. This pattern is typical of a GROUP BY ... WITH ROLLUP operation in SQL, which provides subtotal rows and a grand total row in the result set.

Considering the query options:

A) Option A: GROUP BY group_1, group_2 INCLUDING NULL - This is not a standard SQL clause and would not result in subtotals and a grand total.

B) Option B: GROUP BY group_1, group_2 WITH ROLLUP - This would create subtotals for each unique group_1, each combination of group_1 and group_2, and a grand total, which matches the result set provided.

C) Option C: GROUP BY group_1, group 2 - This is a simple GROUP BY and would not include subtotals or a grand total.

D) Option D: GROUP BY group_1, group_2, (group_1, group_2) - This syntax is not standard and would likely result in an error or be interpreted as a simple GROUP BY, not providing the subtotals and grand total.

E) Option E: GROUP BY group_1, group_2 WITH CUBE - The WITH CUBE operation produces subtotals for all combinations of the selected columns and a grand total, which is more than what is shown in the result set.

The correct answer is Option B, which uses WITH ROLLUP to generate the subtotals for each level of grouping as well as a grand total. This matches the result set where we have subtotals for each group_1, each combination of group_1 and group_2, and the grand total where both group_1 and group_2 are NULL.

To get the number of orders per customer, you need to join the customers and orders tables on the customer_id, count the order_id, and group the results by customer_name. The correct SQL syntax, as outlined in Databricks SQL documentation, is to use GROUP BY on the selected customer field and use COUNT for aggregation. Only option A does this correctly, while the other options contain syntax errors or incorrect field names.

The mean is sensitive to extreme values, often called outliers, which can significantly skew the average away from the true center of the data. The median, however, is a measure of central tendency that is resistant to such outliers because it only considers the middle value(s) when the data is ordered. Therefore, when a variable contains many extreme outliers, there will be a substantial difference between the mean and the median. According to Databricks data analysis materials, this is a fundamental concept when choosing summary statistics for reporting.

Option D is correct. Databricks SQL alerts run a query on a schedule, evaluate a condition, and notify destinations when the condition is met. Since the requirement is notification through a messaging webhook, the alert should use a webhook alert destination. Official Databricks extract: alerts “run queries on a schedule and notify you when a condition that you define is met.” Databricks notification documentation also states that admins can configure alternate notification destinations using webhooks.