What does FIXED do in a Tableau LOD expression?

FIXED computes the aggregate at exactly the dimensions you list, regardless of what else is in the view. With no dimensions listed, as in { FIXED : SUM([Sales]) }, it computes one value across the whole data source.

What is the difference between FIXED, INCLUDE and EXCLUDE?

FIXED ignores the view's dimensions and uses only the ones you name. INCLUDE adds a dimension to the view's level. EXCLUDE removes one from it. INCLUDE and EXCLUDE depend on what's in the view; FIXED doesn't.

When do I need an LOD expression?

When you need a value at a different level from the view, such as percent of total, each customer's first order date, or an average of per-customer totals shown by region.

Tableau

LOD Expression

Also called: Level of detail expression, LOD calculation, FIXED calculation

A level of detail (LOD) expression is a Tableau calculation that computes an aggregate at a level you choose, using FIXED, INCLUDE or EXCLUDE, instead of at whatever level the view is split by.

The expression { FIXED : SUM([Sales]) } returns the 6.0 million dollar total on every region row. Dividing each region's sales by it gives percent of total: APAC 1.2M is 20%, EMEA 1.8M is 30%, LATAM 0.6M is 10% and North America 2.4M is 40%.

What it does

A normal Tableau aggregate is calculated at the level of detail of the view: put Region in the view and SUM([Sales]) gives one total per region. A level of detail (LOD) expression lets you calculate at a different level, then bring the result back into the view.

The example above is { FIXED : SUM([Sales]) }. It lists no dimensions, so it totals everything and returns the same $6.0M on every row. Dividing each region’s SUM([Sales]) by it gives percent of total: 1.2 / 6.0 = 20%, 1.8 / 6.0 = 30%, 0.6 / 6.0 = 10% and 2.4 / 6.0 = 40%, which adds up to 100%.

LOD expressions are a kind of calculated field, written inside curly braces:

{ FIXED [Customer] : MIN([Order Date]) }

FIXED, INCLUDE and EXCLUDE

The three LOD keywords: FIXED computes at exactly the dimensions you list, INCLUDE adds a dimension to the view's level, and EXCLUDE removes one

FIXED computes at exactly the dimensions you list, whatever the view contains. Use it for values that shouldn’t move when the view changes, like a customer’s first order date.
INCLUDE adds a dimension to the view’s level. Use it to work at a finer grain, such as sales per customer, then aggregate that back up, for example the average customer spend per region.
EXCLUDE removes a dimension from the view’s level. Use it to keep a broader total, such as the all-region total, on a view split by region.

Common uses

Percent of total. As above, a fixed total as the denominator.
First or latest event per customer. The first purchase date, from which cohorts are built.
Average of per-customer totals. Total per customer first, then average, rather than averaging individual orders.
Comparing to a benchmark. Show each region next to a company-wide figure.

In Power BI vs. Tableau

Power BI has no LOD keywords. The nearest equivalents are DAX measures that change the filter context, for instance with CALCULATE, which is a different mechanism that reaches many of the same results.

Best practices

Start from the question. Say “what level should this number be at?” before writing the expression.
Prefer FIXED when the value shouldn’t move. It’s easier to reason about because it doesn’t depend on the view.
Remember filters. A regular dimension filter doesn’t change a FIXED result. If you need it to, make the filter a context filter.
Name them by level. “Sales per customer (fixed)” tells the next author what they’re looking at.
Check the grain of your data. Many LOD problems are really a fact-and-dimension table question: what does one row represent?

Mocking up LOD-based numbers

A percent of total or a customer-level average looks simple on a mockup. Note beside each such number what it’s a percentage of, or an average over, so whoever builds it picks the right level. The mockup guide covers what to put on one.