In this article
Data fields and connections
What is the difference between a dimension and a measure in Tableau?
A dimension groups or segments data and usually holds qualitative values such as names, dates or locations, while a measure holds numbers that Tableau aggregates with a function such as sum or average (Tableau Help, field types and roles). Dimensions set the level of detail of the view. A numeric identifier that should not be added up belongs in a dimension, and a postal code is a dimension.
Why it is asked: wrong roles cause wrong totals. Related reading: chart and dashboard questions in the data analyst guide.
What do blue and green pills mean?
A pill is the field token placed on a shelf in the view (Tableau's term). Blue pills are discrete fields, which are individually separate and distinct and create headers; green pills are continuous fields, which form an unbroken whole and create an axis. Role and setting are separate: a dimension or a measure can be either. String and Boolean dimensions cannot be continuous, and dates are the most common continuous dimension (Tableau Help).
Why it is asked: it tests whether you know why a chart shows headers or an axis.
When should you use an extract instead of a live connection?
An extract is a subset of data saved separately from the original dataset. Tableau lists benefits for very large datasets, better performance, and features the source may lack, and a full or incremental refresh keeps it current (Tableau Help, extracts). Choose a live connection when results must reflect the source as it is now, or when the source is a virtual connection whose data policy uses a user function such as USERNAME(): Tableau says not to extract it, because the extract keeps only the rows the policy allowed when it was created.
Why it is asked: it separates speed from freshness.
Combining tables
If the query language behind the workbook is the weak spot, start with the Structured Query Language (SQL) guide and conditional sums in SQL.
What do inner, left, right and full outer joins return?
An inner join returns only values with matches in both tables; a left join keeps every row of the left table and shows nulls where the right has no match; a right join does the reverse; a full outer join keeps all rows from both (Tableau Help, joins). With key sets {East, West, North} and {East, West, South}, the four joins return 2, 3, 3 and 4 keys.1
Why it is asked: it checks the basics before any data-model follow-up. Related reading: how joins work.
What is the difference between a join and a relationship?
A join merges physical tables into one before analysis and you pick the type, while a relationship links logical tables by matching fields and Tableau builds the joins at analysis time from the fields in each view. Relationships keep each table at its own level of detail and query only the tables a view uses (Tableau Help, data model). The page recommends joins only when necessary.
Why it is asked: it shows whether you know the modern data model. Related reading: the grain of a fact table.
Why can a join inflate a total?
A join repeats a row of the "one" side once for every matching row of the "many" side, so summing that field overstates it. In an illustrative example the true target total is 250 for two regions, and the target summed after the join to five order rows is 600, which is 2.4 times the true total.1 Relationships avoid this because tables keep their own level of detail, and a FIXED expression can deduplicate: SUM({ FIXED [Region] : MIN([Target]) }) returns 250 on the joined rows.1
Text description of the diagram
- The true target total is 250 for two regions.
- The target summed after the join to five order rows is 600.
Why it is asked: it silently produces wrong totals. Read the full explanation: why a join returns one row per matching pair.
How does data blending differ from a join?
Blending queries each source separately, aggregates it to the view's level, then combines the results visually, while a join combines the rows first and aggregates after. The first source used is primary and the secondary shows only where it matches the primary, which behaves like a left join (Tableau Help, blending). Linking fields must share values, and blended sources cannot be published as one unit.
Why it is asked: it separates sources that cannot be joined from ones that can. Related reading: how a left join keeps unmatched rows.
What is a union, and how is it different from a join?
A union is not a join: it stacks rows from one table onto another, ideally with the same number of fields and matching names and data types. Tableau describes it as appending rows of data from one table to another (Tableau Help, joins). A join adds columns by matching keys.
Why it is asked: it checks that you pick rows versus columns correctly.
Knowing it is half the job. Saying it is the other half.
Interview Clozo is an AI interview copilot for real interview rounds: it hears the question and puts a strong answer on your screen, and you say it in your own words. See the features or how it works.
Start Free Trial →Calculations
What is a level of detail expression, and how do the three forms differ?
A level of detail (LOD) expression is a calculation that computes an aggregate at a level you name instead of the view's level. The FIXED form uses only the named dimensions and ignores most view filters, INCLUDE adds dimensions to those in the view, and EXCLUDE removes them (Tableau Help, LOD expressions). Syntax: { FIXED [Region] : SUM([Sales]) }. In our made-up data, the FIXED region sales put 600 on each East order row, so order shares were 17, 33 and 50 percent.1
Why it is asked: it tests whether you can change the level of detail.
What is a table calculation, and how do addressing and partitioning work?
A table calculation is a transformation applied to the values in a visualization, computed on the data in the view, so values filtered out of the view are ignored. Partitioning fields split the view into sub-tables and the calculation restarts in each; addressing fields set the direction it moves (Tableau Help, table calculations). Every dimension in the view is one or the other.
Why it is asked: it shows whether you can predict a running total's direction.
When would you use a LOD expression instead of a table calculation?
Use a FIXED LOD expression when the value must not depend on what is in or filtered out of the view, and a table calculation when you need to move across the marks that are shown (the points, bars or cells drawn in the view). The Tableau page says FIXED expressions are computed before dimension filters, while a dimension filter is applied before a table calculation (order of operations).
Why it is asked: it tests whether you understand the order the filters run.
Why would you add a context filter?
A context filter runs before any other filter you create in a worksheet, so it narrows the data the later filters see. Tableau notes that Top N and general dimension filters apply at the same level, which is why it recommends adding the other dimension filter to context when a Top N must follow it (order of operations). Context filters also still apply to FIXED expressions (LOD page).
Why it is asked: it separates people who debug filter order from those who guess.
Sets and parameters
What is the difference between a fixed set and a dynamic set?
The members of a fixed set do not change when the data changes, and it can be based on one or more dimensions; the members of a dynamic set change with the data, and it can only be based on a single dimension. A set can split the view into In and Out (Tableau Help, sets).
Why it is asked: it tests whether you know when membership is frozen.
What is a parameter, and how is it different from a filter?
A parameter is a workbook variable that holds a value, such as a number, date or string, and can stand in for a constant in a calculation, filter or reference line. A filter restricts which data appears; a parameter filters nothing until it is tied to an element in the view (Tableau Help, parameters). Parameters are global across the workbook.
Why it is asked: it shows whether you can build user-controlled thresholds.
Dashboards and performance
What types of dashboard action does Tableau offer?
Tableau offers Filter, Highlight, Go to URL, Go to Sheet, Change Parameter and Change Set Values actions. Filter uses the data from one view to filter another, Highlight dims all other marks, and the last two let viewers change a parameter or a set by interacting with marks (Tableau Help, actions).
Why it is asked: it tests whether you design dashboards for exploration. Related reading: dashboard layout in the analytics guide.
How do you speed up a slow workbook?
Connect to an extract when possible, bring in only the data you need, filter on summary-level values, and reduce the number of marks in the view. Tableau calls filtering one of the most common causes of poorly performing workbooks and recommends the performance recorder to find where to focus (Tableau Help, performance checklist).
Why it is asked: it separates builders from users.
Which calculations and dashboard settings are slow?
COUNTD is one of the slowest functions, and Booleans and integers are faster than strings and dates. Fixed-size dashboards do not need re-rendering for different window sizes, and custom geocoding can greatly increase workbook size (Tableau Help, performance checklist). Publishing to Tableau Server does not improve performance, so tune in Tableau Desktop first.
Why it is asked: it tests concrete tuning habits over vague advice.
Why does a range filter often beat a long list of values?
Tableau states that filtering on a range of values is often faster than filtering on large itemized lists of discrete values. Its checklist also suggests the Show Apply Button option so a user's selections run as one query instead of one per click (Tableau Help, performance checklist).
Why it is asked: it checks that you design filters with query cost in mind.
- Dimensions set the level of detail of a view and measures are aggregated; blue pills make headers and green pills make axes.
- A join repeats rows: in our illustrative data a target of 250 summed to 600 after the join.
FIXEDlevel of detail expressions run before dimension filters; table calculations run after them on the data in the view.- Fixed sets freeze their members, and dynamic sets follow the data on one dimension.
- Speed comes from extracts, fewer marks, summary-level filters and avoiding slow functions such as
COUNTD.
Sources
- Data Field Types and Roles (Tableau Help) — Tableau (page shows no date; read 2026-10-24), 24 Oct 2026
- Extract Your Data (Tableau Help) — Tableau (page shows no date; read 2026-10-24), 24 Oct 2026
- Join Your Data (Tableau Help) — Tableau (page shows no date; read 2026-10-24), 24 Oct 2026
- Relationships, Joins and Blends: the data model (Tableau Help) — Tableau (page shows no date; read 2026-10-24), 24 Oct 2026
- Blend Your Data (Tableau Help) — Tableau (page shows no date; read 2026-10-24), 24 Oct 2026
- Level of Detail Expressions (Tableau Help) — Tableau (page shows no date; read 2026-10-24), 24 Oct 2026
- Transform Values with Table Calculations (Tableau Help) — Tableau (page shows no date; read 2026-10-24), 24 Oct 2026
- Tableau's Order of Operations (Tableau Help) — Tableau (page shows no date; read 2026-10-24), 24 Oct 2026
- Create Sets (Tableau Help) — Tableau (page shows no date; read 2026-10-24), 24 Oct 2026
- Create Parameters (Tableau Help) — Tableau (page shows no date; read 2026-10-24), 24 Oct 2026
- Actions and Dashboards (Tableau Help) — Tableau (page shows no date; read 2026-10-24), 24 Oct 2026
- Designing Efficient Workbooks: performance checklist (Tableau Help) — Tableau (page shows no date; read 2026-10-24), 24 Oct 2026
