Searching for Tableau Certified Data Analyst exam questions and answers usually turns up two useless things: leaked dumps that are years out of date, and vague "study the docs" advice. What's genuinely useful is knowing the shapes of question you'll face, because this exam has a small number of recurring patterns and each one has a reliable way to attack it. Here are the nine that matter, with what the exam is really asking each time.
The format, briefly
Tableau Certified Data Analyst is 60 questions in 120 minutes, $250 USD, and you need 750 out of 1000 scaled — roughly 75%. It's split between knowledge-based multiple choice and hands-on lab questions where you drive an actual Tableau environment. That lab component is the thing that makes this exam different from most analytics certs, and the thing most candidates under-prepare for. It's valid for two years.
Domains: Connecting to & Transforming Data (~26%), Exploring & Analyzing Data (~35%), Creating Content (~18%), Publishing & Managing Content (~11%), Understanding Tableau Concepts (~10%).
The nine question types
1. "Which join/relationship produces this row count?"
You get two tables and a result and must pick the join type — or, more often in modern Tableau, decide between a relationship (the noodle) and a physical join. The trap is that people default to joins because that's what they learned in SQL.
Attack: relationships keep tables at their native granularity and only join at query time based on the viz. Joins duplicate rows immediately. If a question mentions row duplication or inflated measures, the answer is usually "use a relationship instead."
2. LOD expression selection
FIXED, INCLUDE, EXCLUDE. Guaranteed to appear, usually more than once.
Attack: memorise this and you'll get all of them:
-
FIXEDignores the viz dimensions entirely — use it when the calculation must stay constant regardless of what's on the shelves. -
INCLUDEadds a dimension finer than the viz. -
EXCLUDEremoves a dimension the viz has.
Then know the order of operations: FIXED LODs compute before dimension filters, which is why a FIXED result "ignores" your filter. If a question describes a number that stubbornly refuses to respond to a filter, you're looking at a FIXED LOD, and the fix is to add the filter to context.
3. Context filter questions
Related to the above and reliably examined. A context filter runs before FIXED LODs and before top-N filters.
Attack: whenever a question involves a Top N filter combined with another filter and the result "looks wrong," the answer is "add the other filter to context."
4. Table calculation scope and direction
WINDOW_SUM, RUNNING_SUM, INDEX(), RANK(), and the compute-using / addressing-partitioning distinction.
Attack: the exam's favourite is a running total that resets at the wrong point. Remember that "Compute Using → Table (Across)" vs "Pane (Down)" changes the partition. Anything that resets per group is partitioned by that group.
5. Hands-on: build a specific viz
The lab drops you in Tableau and says "create a dual-axis chart showing X vs Y with synchronised axes" or "add a reference line at the average."
Attack: there is no substitute for having done it. Practise these specific builds until they're muscle memory: dual axis + synchronise, combined axis, reference line/band, sets, groups, hierarchies, parameters wired to a calculated field, and a dashboard action (filter, highlight, URL). If you can build all of those without thinking, the lab is easy marks. If you can't, you'll burn 10 minutes on one question.
6. Data prep and cleaning
Pivot, split, data interpreter, union vs join, changing data types and geographic roles.
Attack: know when to use union (same columns, more rows — e.g. monthly files) vs join (different columns, same entities). Know that Data Interpreter handles Excel files with headers and footers. Know pivot turns wide to tall.
7. Calculated field debugging
They show you a formula and ask why it errors or returns nulls.
Attack: the usual culprits are aggregate/non-aggregate mixing (SUM([Sales]) - [Profit] is invalid — you can't mix), division by zero needing ZN() or a IIF([x]=0, NULL, ...) guard, and type mismatches. The aggregate-mixing one appears constantly.
8. Publishing, permissions, and extracts
Tableau Server/Cloud content: projects, permissions, extract refresh schedules, live vs extract, data source certification, and who can see what.
Attack: this domain is only ~11% and it's the one people skip because it's dull. Learn the permission model (project → workbook → view inheritance, and that Deny beats Allow) and the difference between a published data source and an embedded one. It's a few hours for a few guaranteed marks.
9. "Which chart for this question?"
Given a business question, pick the viz type.
Attack: Tableau has opinions. Trend over time → line. Part-to-whole → treemap or stacked bar, rarely pie. Distribution → histogram or box plot. Correlation → scatter. Comparison across categories → bar. Geographic → filled or symbol map. When a question stresses "compare precise values," bar beats pie every time.
How to prepare for the lab half
This is where I see people lose. The knowledge questions you can study for by reading. The lab questions you can only prepare for by building.
Get Tableau Desktop (the 14-day trial, or Tableau Public for the free-but-limited version) and rebuild the same handful of vizzes until it's automatic. Use Superstore — the exam environment uses similar sample data and the familiarity genuinely helps.
Time yourself. 60 questions in 120 minutes is two minutes each, but a lab question takes four or five. That means your multiple-choice answers need to average well under two minutes to leave room. Practise pace, not just accuracy.
Where to get realistic questions
Since the real exam's questions aren't public, what you want is a bank that mirrors the patterns above rather than a dump claiming to be the real thing. Dumps are both a ToS violation that can void your cert and usually wrong — Tableau updates this exam regularly.
I use ExamCert's free Tableau practice tests to baseline, and the full Tableau Certified Data Analyst question bank for volume — it's a couple of dollars and the explanations are the actual value, since on LOD and table-calc questions the reasoning is the thing you need to absorb. For the ones where I genuinely can't see why my answer was wrong, the AI simulator at ai.examcert.app will walk through each option, which beats scrolling forum threads.
The one-line summary
Two-thirds of this exam is LODs, table calculations, filter order of operations, and being fast in the actual product. Get those four right and the rest is reading comprehension.
Top comments (0)