Databricks Data Analyst Cheat Sheet

The short version

Use this sheet for a last review, not as your only study source. It covers the nine sections of the Databricks exam guide dated October 30, 2025, in our own words, and it does not cover everything on the exam. Check the Databricks exam page for the current outline.

Exam facts at a glance

Item	Detail
Scored questions	45, multiple choice
Time	90 minutes
Language	English
Delivery	Proctored, online or at a test center
Test aids	None allowed
Passing score	Not published
Retake wait	14 days
Validity	2 years

Weightings: queries 20%, dashboards 16%, query analysis 15%, Genie 12%, platform 11%, managing data 8%, security 8%, importing 5%, data modeling 5%.

Platform and managing data (11% and 8%)

Confused pairs

Delta Lake is the table storage format. Unity Catalog governs access and lineage. Databricks SQL is where you run queries.
Managed table: Databricks controls the files, and dropping the table removes the data. External table: you control the location, and dropping it leaves the files.

Decision rules

Scenario asks who owns, who can see, or where did this come from: think Unity Catalog and lineage.
Scenario asks about cleaning bad rows: think SQL with `DELETE … WHERE`, `IS NULL`, `COALESCE`.

Trap: `NULL = NULL` is not true. A filter on a nullable column quietly drops those rows.

Importing data (5%)

One-off file from your laptop: the upload UI.
New files arriving in cloud storage over time: Auto Loader.
Live data from another organization: Delta Sharing.
A ready-made dataset from outside: Marketplace.

Queries (20%)

Confused pairs

`UNION` removes duplicates; `UNION ALL` keeps them.
View: a saved query, run each time. Materialized view: stored, precomputed results. Streaming table: built for incrementally arriving, append-only sources.
`LEFT JOIN` keeps unmatched left rows; a `WHERE` filter on the right table can turn it into an inner join.

Decision rules

Result is read often and the data changes slowly: think materialized view.
Question mentions a `LOCATION` clause: the table is external.
Need an earlier state of a Delta table: time travel with `VERSION AS OF` or `TIMESTAMP AS OF`.

Trap: where `NULL` values sort by default. Know it, or check in a workspace.

Analyzing queries (15%)

Confused pairs

Query result cache, disk cache and materialized view are three different things. An instant second run of the identical query points to a result cache.
Query Insights finds slow queries across a workspace; the query profile shows steps inside one query.

Decision rules

Large table, queries filter on a few columns: Liquid Clustering with `CLUSTER BY`.
Need to audit changes to a Delta table: `DESCRIBE HISTORY`.
Old version is missing in time travel: `VACUUM` or log retention may have removed it.

Dashboards (16%)

Trend over time: line. Distribution: histogram. Relationship between two measures: scatter. Parts of a whole with few slices: pie.
Parameter: feeds a value into a query, often in `WHERE`.
Scheduled refresh keeps the data current; an alert tests a condition against a threshold and notifies a destination.
Sharing has access levels and can include external viewers; remember the viewer still needs data access as configured.

Genie spaces (12%)

A Genie space needs a SQL warehouse, curated tables or views, instructions and sample questions.
Trusted assets are vetted queries that improve answers.
Two permission layers: access to the space, and access to the underlying data.
Better quality comes from clear instructions and curation, not from adding more tables.

Data modeling and security (5% and 8%)

Star schema: central fact table with denormalized dimensions. Snowflake: normalized dimensions.
Medallion layers: bronze raw, silver cleaned, gold reporting-ready.
Namespace: catalog, then schema, then table.
A user needs `USE CATALOG`, `USE SCHEMA` and `SELECT`. Missing any one blocks access.
Hide columns with a column mask; hide rows with a row filter.

Common traps across sections

Picking a feature because its name sounds right instead of what it does.
Forgetting a privilege at one level of the namespace.
Mixing up cache types.
Confusing Delta Sharing (live sharing) with Auto Loader (ingesting files).
Not counting answers when a question says to pick two.

Last-day checklist

Re-read the nine section titles and weightings.
Run your system check for online proctoring.
Review the confused pairs above and say each one out loud.
Skim your own notes on the weakest section only.
Sleep. No test aids are allowed, so there is nothing to bring.

More guides for DATABRICKS-DATA-ANALYST-ASSOCIATE

Next step

Open the exam overview for context, then the domains explained page for the detail behind each line above. To test recall under timing, try our Databricks Data Analyst Associate practice tests.

Frequently asked questions

Is this enough to pass?

No. It is a review aid. Hands-on work and the Databricks exam guide come first.

Does it contain real exam questions?

No. Everything here is our own summary.

How often does it change?

When Databricks updates the exam guide, this sheet needs a recheck.

Try these practice exams