← All Kits · Tableau Kit

Why Your Tableau Filter Did Not Do What You Expected: The Order of Operations

Michael Nocito · Updated August 2026 · Every number on this page was worked before it was published

You put a Top 10 Customers filter on the view and a Region filter beside it. Choose North and eight rows appear. Nothing is broken: Tableau computed the top 10 before it applied the region filter, then threw two of them away.

What you do: learn the sequence, then use context filters to move a filter earlier when the order is wrong for your question.

The short version. Filters do not all run at the same time, and the earlier ones decide what the later ones can see.

The sequence

Two rows of this table explain most confusion. Stage 4 is why a FIXED expression seems to ignore the filter you just set: it does ignore it, unless that filter is in context. And stage 8 is why a table calculation filter hides rows while leaving a running total looking like it still counts them, which is usually exactly what you want.

The context filter, worked

Back to the Top 10 problem. Region is a dimension filter at stage 5, and so is the Top 10, so they compete inside the same stage with the top N computed against everything the earlier stages allowed through.

Right click the Region filter and choose Add to Context. It turns blue and moves to stage 3. Now the top 10 is computed within North only, and ten rows appear.

The cost: a context filter creates a temporary table behind the scenes, so it is not free. Use it where it is needed rather than as a habit, and remember that changing a context filter forces more work than changing an ordinary one.

Where a filter can be applied

Rows that should never appear in any view belong in the extract or data source filter, not on a shelf. It is the same argument as filtering in Power Query rather than in a report: a filter one click away from being removed will eventually be removed.

The three symptoms and their causes

1. Top N returns fewer rows than N. A dimension filter is competing with it. Add that filter to context.
2. A percent of total does not change when you filter. The denominator is a FIXED expression running before the filter. Add the filter to context, or use INCLUDE, or write the total as a table calculation.
3. Hiding a row changes the running total. You used a dimension filter where you wanted a table calculation filter. Filtering with a table calc, or right click and Hide, keeps the calculation intact.

Hide is not filter

Right clicking a header and choosing Hide removes the mark from the view without removing the row from the calculation. That is genuinely useful for a running total or a percent of total, and it is a trap when you meant to exclude the data. Hidden members do not appear in the filter shelf, so nothing on screen says they exist, and a colleague opening the workbook will not find them.

How to apply this to your own work

1. Open your most complicated sheet and read the filter shelf colours. Note which are in context.
2. For every top N filter, check whether another dimension filter should be in context beside it.
3. Search your calculated fields for FIXED and ask which filters it is meant to respect.
4. Move permanent exclusions into a data source filter and remove them from the shelves.
5. Check for hidden members before you trust any total, using Analysis, Reveal Hidden Data.

The one habit to keep

When a view surprises you, ask which stage each part of it runs in. That question converts a mysterious result into a sequence you can reason about, and it works for every version of Tableau.

Is the top N on your dashboard computed before or after the region filter?