In Tableau, how do FIXED, INCLUDE and EXCLUDE LOD expressions differ in granularity?

## Level of detail, in one sentence Every Tableau worksheet has a *view level of detail*: the set of dimensions on Rows, Columns and the Marks card, including anything on Detail or Tooltip. Ordinary aggregations inherit that grain — `SUM([Sales])` produces one number per mark and changes the moment you add a dimension. A level of detail (LOD) expression is an aggregate that declares its own grain instead of inheriting the view's. ## The syntax ``` { FIXED [Dim1], [Dim2] : AGG([Measure]) } ``` The braces make it an LOD. The keyword is FIXED, INCLUDE or EXCLUDE. Between the keyword and the colon is a dimensionality declaration; after the colon is an aggregate expression. The whole thing returns an aggregate value, which the view will then aggregate again — so what you wrap it in matters. ## FIXED — absolute grain FIXED ignores the view's dimensions entirely and computes at exactly the dimensions you list. `{FIXED [Customer ID] : SUM([Sales])}` is each customer's total sales. Drop Region into the view, swap it for Category, remove it altogether — the per-customer number does not move. With no dimensions at all, `{FIXED : SUM([Sales])}` is one number for the entire table, which is the standard building block for a percent-of-total that survives view changes. Because FIXED refuses to look at the view, it also refuses to look at ordinary dimension filters: those are applied after FIXED expressions are computed. Context filters, which are applied before, do affect it. That asymmetry is the single most-asked follow-up on this topic. ## INCLUDE — the view plus something finer INCLUDE computes at the view's dimensions *plus* the ones you name, then hands the result back to the view to be re-aggregated. The canonical use is an average over a grain finer than the view. `AVG({INCLUDE [Customer ID] : SUM([Sales])})` computes each customer's sales inside whatever the mark covers, then averages those per-customer totals. Put Region on Rows and it reads "average sales per customer in this region"; swap Region for Category and it re-reads itself as "average sales per customer in this category" with no edit. That relativity is the point: INCLUDE follows the view. INCLUDE is only meaningful when the dimension you name is finer than, or absent from, the view. Including a dimension already in the view adds nothing. ## EXCLUDE — the view minus something EXCLUDE computes at the view's dimensions *minus* the ones you name. In a view broken out by Category and Sub-Category, `{EXCLUDE [Sub-Category] : SUM([Sales])}` returns the category total on every sub-category row. Divide the mark's own sales by it and you have percent of category — one that keeps working when someone reorders the shelves. Excluding a dimension that is not in the view does nothing at all, which is a frequent silent no-op: the calculation saves, the number looks plausible, and it is simply the ordinary aggregate. ## A worked comparison View: Region on Rows, Category on Columns, `SUM([Sales])` as the measure. ``` Region Category SUM(Sales) {FIXED [Region]} {EXCLUDE [Category]} East Furniture 40 100 100 East Tech 60 100 100 West Furniture 30 80 80 West Tech 50 80 80 ``` Here FIXED and EXCLUDE agree, because removing Category from this particular view leaves exactly Region. Now move Category off the view: the FIXED column is unchanged at 100 and 80, while the EXCLUDE column becomes the same as `SUM([Sales])`, because there is no Category left to remove. Same numbers today, different behaviour tomorrow — that is the difference candidates are being tested on. ## The result is aggregated again An LOD returns an aggregate, and the view still needs one number per mark, so Tableau applies an outer aggregation — SUM by default when you drop it as a measure. `AVG({INCLUDE [Customer ID] : SUM([Sales])})` and `SUM({INCLUDE [Customer ID] : SUM([Sales])})` are both valid and answer completely different questions. Choosing that wrapper deliberately is part of writing an LOD, not an afterthought. ## Nesting and using an LOD as a dimension LOD expressions nest: `{FIXED [Region] : AVG({INCLUDE [Customer ID] : SUM([Sales])})}` is legal and computes the inner expression first. A FIXED expression whose result is a stable per-entity attribute — the classic being `{FIXED [Customer ID] : MIN([Order Date])}`, a customer's acquisition date — can be converted to a dimension and used to group marks, which is how cohort analysis is built in Tableau without touching the source. ## What interviewers listen for They want the one-line contrast (absolute versus relative to the view), a concrete use for each, and the awareness that FIXED sidesteps dimension filters. A strong answer volunteers that the LOD's result is re-aggregated by the view, and that EXCLUDE of a dimension not in the view quietly does nothing.