Variables in DAX: more readable and faster measures

Introduced in DAX in 2015, variables (VAR/RETURN) are perhaps the most impactful addition to the language. They let you name sub-expressions, making code more readable and potentially faster.

Syntax

Measure =
VAR vSales  = SUM(fSales[Revenue])
VAR vCost   = SUM(fSales[Cost])
VAR vMargin = vSales - vCost
RETURN
    DIVIDE(vMargin, vSales)

Readability benefits

Compare with the version without variables:

-- Without variables (hard to read and maintain)
Margin % = DIVIDE(
    SUM(fSales[Revenue]) - SUM(fSales[Cost]),
    SUM(fSales[Revenue])
)

With variables, each part has a semantic name. Other team members (or you in 6 months) immediately understand the intent.

Performance benefits

A variable is evaluated only once in the filter context where it is declared. If the same sub-expression appears three times in the measure, without variables it is evaluated three times; with variables, it is evaluated once and the result reused.

Variables for debugging

Debug Measure =
VAR vBase     = CALCULATE(SUM(fSales[Revenue]), REMOVEFILTERS())
VAR vFiltered = SUM(fSales[Revenue])
RETURN
    "Base: " & vBase & " | Filtered: " & vFiltered

Change the RETURN to reveal intermediate values during development, without altering the final logic.

Important limitation

A variable captures the filter context at the moment it is declared. If the context changes afterwards (e.g. inside a CALCULATE), the variable is not re-evaluated. This behavior is intentional and consistent, but worth knowing.

Conclusion

Use VAR/RETURN in every measure with more than one sub-expression. The cost is zero and the gains in readability and maintenance are enormous.