skip to content

In Tableau, what is the difference between a fixed set and a computed set?

level: middleimportance: nice to knowfreq 38%

answer

  1. one stores names, one stores a rule
  2. binary membership on one dimension
  3. new data changes only one of them
  4. it is a peer of dimension filters
  5. IN and OUT, not many groups

basics

~20 s

A fixed set stores a hand-picked list of members in the workbook and never changes as data arrives. A computed set stores a condition or a top-N rule and re-evaluates it each time the workbook queries, so its membership tracks the data.

solid answer

~50 s

Both are sets: a binary IN/OUT field built from one dimension in Tableau, which you can put on shelves, use as a filter, or colour by. The difference is what is stored. A **fixed set** stores the chosen member names, so it keeps the same members forever — useful for a curated list such as strategic accounts, and stale the moment the business changes. A **computed set** stores a rule instead — a condition like `SUM(Sales) > 100000`, or a top-N by a field or parameter — and Tableau re-evaluates it on every query, so a customer joins or leaves the set as the data moves. Because that rule is evaluated on the same rung as dimension filters, a computed set is blind to other dimension filters unless they are added to context. Set actions let a viewer's selection rewrite a set's membership.

code

text · 7 lines
text
-- Fixed set: stored membership, never re-evaluated
Top Accounts (fixed) = { Acme, Globex, Initech }

-- Computed set: stored rule, evaluated per query
High Value (computed) = members where SUM([Sales]) > 100000
-- new customer crossing 100k joins automatically; Acme leaving does not
-- remove it from the fixed set

go deeper

for a junior

Recall that a Tableau set splits one dimension's members into IN and OUT, and that you choose members by hand or by a rule when you create it.

for a middle

Explain the storage difference — a member list versus an evaluated condition or top-N — and note that combining sets requires them to be built from the same dimension.

for a senior

Point out the failure modes: fixed sets go stale silently, and computed sets are peers of dimension filters, so 'top ten within the selection' needs context or a late table calculation.

for a principal

Own whether the membership rule belongs in the workbook at all. A definition like key account repeated as a set in ten workbooks is an ungoverned metric; consider pushing it into the model instead.

## What a set is A set in Tableau is a field derived from a single dimension whose value is binary: every member of that dimension is either **IN** the set or **OUT** of it. You create it by right-clicking a dimension and choosing *Create > Set*. Once created it behaves like a dimension you can drop on Color, Rows, or the Filters shelf, and you can build a calculation that tests it. That binary shape is the point. A set answers "is this member in my group?" — which makes an IN/OUT comparison (my top accounts versus everyone else) a two-mark chart rather than a hundred-mark chart. ## Fixed sets A fixed set is defined by ticking members in the **General** tab of the set dialog. Tableau stores those member names in the workbook. Nothing re-evaluates: if a new customer appears in the data next quarter, they are OUT, forever, until somebody opens the workbook and ticks them. That is exactly right for a curated list — a set of pilot stores, an internal-test accounts exclusion, the four brands the executive sponsor cares about. It is exactly wrong for anything that ought to track the business, and "the set is stale" is a real production failure that nobody notices, because a stale set produces plausible numbers rather than an error. ## Computed sets A computed set stores a rule rather than a list. Two flavours: - **Condition**: a formula per member, for example `SUM([Sales]) > 100000`, evaluated per member of the dimension. - **Top**: top or bottom N by a field, where N can itself be a parameter so a viewer can change it. Tableau evaluates the rule at query time, so membership follows the data. A customer who crosses the threshold this month is IN this month. ## Where a computed set sits in the order of operations This is the part that catches people. A computed set's rule is evaluated on the same rung as dimension filters. So a computed set is blind to the other dimension filters on the view, for exactly the same reason a Top N filter is blind to a region filter: peers on the same rung are evaluated independently. If you want "top ten within the selected region", the region filter must be added to **context** so it runs above the set. ## Combining sets Two sets built from the **same dimension** can be combined into a third: members in both, members in either, or members in one but not the other. That is how you express "customers who were in the top ten last year but are not this year" — a churn question that is otherwise awkward. Sets on different dimensions cannot be combined this way, because IN/OUT is defined per member of one dimension. ## Set actions A set action lets a selection in the view rewrite a set's membership: a viewer clicks marks, and those members become the set. That turns a set into an interactive grouping control, and it is the mechanism behind proportional-brushing and "compare my selection to the rest" dashboards. It is worth knowing that this rewrites membership rather than filtering — the unselected marks are still in the view, still aggregated, just OUT. ## Sets versus groups A **group** combines several members into a new named member — three spellings of the same city become one city. It is n-way, it changes the labels in the view, and it is static. A **set** is binary and can be dynamic. A group answers "treat these as one thing"; a set answers "is this one of my things?". Interviewers ask for this distinction often, and "a group is many-to-few relabeling, a set is IN/OUT membership" is the crisp version. ## Sets versus a filter Using a set as a filter keeping only IN is equivalent to filtering, and people often do exactly that. The reason to prefer a set is that you get to *keep* the OUT members and compare against them, and that the membership rule lives in one reusable object rather than being retyped in every worksheet's filter dialog. ## What interviewers listen for They want the storage difference stated plainly — a list versus a rule — and then a consequence: staleness for the fixed one, order-of-operations blindness for the computed one. A candidate who adds set actions or the same-dimension combination rule is showing real usage rather than a definition memorised from documentation.

  • How does a set differ from a group in Tableau?
    A group combines several members of a dimension into one new named member, so it relabels many values as few and is static. A set is binary — every member is IN or OUT — and can be computed from a rule, so it can track the data. Group means treat these as one; set means is this one of mine.
  • Why might a computed set ignore the filter card on the view?
    A computed set's condition or top-N rule is evaluated on the same rung of the order of operations as dimension filters, so it cannot see its peers. Membership is derived from the contexted data, not from the filtered view. Add the filter to context if the set must be computed inside it.
  • What does a set action let a dashboard do that a filter action does not?
    A set action rewrites which members are IN the set rather than removing marks, so the unselected members stay in the view and can be aggregated as an OUT bucket. That is how proportional brushing and selection-versus-remainder comparisons are built; a filter action would simply delete the comparison group.

saying these in an interview costs you the question

  • Says a fixed set updates when new data arrives
  • Uses set and group interchangeably
  • Thinks sets can hold members from two different dimensions
  • Expects a computed set to respect other filter cards automatically
  • Believes a set always filters the view rather than labelling it

context

resolving links…