Tableau Aliases: Rename What Readers See Without Touching the Data

1. What an alias actually is

Before the explanation: you alias the value "E" to "East". What does a filter on that field now show, and what does the underlying data hold?

The filter shows East. The data still holds E. Both are true at the same time, and that's the whole idea.

Tableau's own description is that aliases are "alternate names for members in a dimension so that their labels appear differently in the view." Two words in that sentence are doing the work.

Members. A member is one of the distinct values a dimension takes. Region has the members E, W, N and S. Aliases rename members, one at a time.

Appear. The change is to how the value appears. Nothing is written back to the data source, no calculated field is created, and no row is modified.

Where the alias shows up is broader than most people expect. Axis and row headers, the legend, filter cards, tooltips and labels all use it, because they're all displaying the member. So one edit fixes the codes everywhere in the workbook at once rather than in one chart.

2. Why measures can't have one

Before the answer: you right-click Sales, a measure, looking for Aliases. It isn't there. Is that a bug, a permissions problem, or intended?

Intended, and the reason is worth understanding because it stops you hunting for a workaround. Tableau's documentation is exact: "Aliases can be created for the members of discrete dimensions only. They can't be created for continuous dimensions, dates, or measures."

Say out loud why a measure couldn't sensibly have an alias, before reading on. This is the hinge of the guide.

Here's the reason. A measure doesn't have members, it has values, and its values are unbounded. Sales might hold 4,182 distinct numbers today and 4,200 tomorrow. There's no fixed list to attach names to. More importantly, renaming the number 4,182 to something else would be changing what the number is, not what it's called. Tableau's own wording says it does not permit re-aliasing measures "because it involves modifying data values themselves."

The same logic covers the other two exclusions. A date and a continuous dimension both hold values on a scale rather than a fixed list of members, so there's nothing to name.

So what do you do when you genuinely need a measure to read differently? Two answers, depending on what you meant. If you want the number displayed differently, that's number formatting, which changes appearance and leaves the value alone. If you want the numbers sorted into named ranges, such as Low, Medium and High, that's a bin or a calculated field, because you're creating members that didn't exist.

3. An alias against renaming the field

Before the distinction: your chart shows a column headed "cust_seg" containing "A", "B" and "C". Which of those four things does an alias fix?

The A, B and C. Not the heading. Both operations get called renaming in conversation, and they're different edits on different objects.

Most exported or system-generated data needs both. The field is named for a database column and the values are stored as short codes, so you rename the field once and alias its members once, and the chart stops needing a translator.

Picture the last chart you built from an export. How many codes on it would a reader outside your team have had to ask about? Each one is an alias you haven't set yet.

4. Setting them, and putting them back

Before the steps: you alias four members and then decide the original codes were better. How much work is undoing it?

One click. Tableau keeps the originals, so nothing you do here is one-way.

To set them:

1. In the Data pane, right-click a dimension and select Aliases.
2. In the Edit Aliases dialog, under Value (Alias), select a member and type the new name.
3. In Tableau Desktop, click OK. On Tableau Server or Tableau Cloud, click the X in the top-right corner of the dialog.

To undo them, open the same dialog and click Clear Aliases. Tableau's description of what that does: "To reset the member names back to their original names, click Clear Aliases." All of them, at once, back to the stored values.

There's a second route worth knowing for one-off edits. Double-click a header in the view itself and type. That edits a single member's alias without opening the dialog, which is faster when you're fixing one label and slower when you're fixing eight.

5. Details people miss

Before the list: the Aliases option is greyed out on a discrete dimension, which section 2 says should work. What's going on?

You're almost certainly on a published data source. Tableau states the limit plainly: "When using a published data source, you can't create or edit aliases. You can only change aliases on fields that you create in your workbook." So the fix is to create a field in your own workbook, or ask whoever owns the published source to set the alias there, where it will benefit everyone using it.

Five more.

An alias is not a filter and not a group. Aliasing two members to the same name makes them look identical on screen while remaining two separate members underneath. They will not add together. If you want them combined, that's a group.

Aliases live with the data source, so they travel. Every worksheet in the workbook using that field gets them. This is the same arrangement as sets and hierarchies, and it's the reason all three are edited from the Data pane rather than from a sheet.

Sorting still uses the underlying value in some places. Renaming E to East does not by itself reorder anything, and an alphabetic sort may not land where the new names suggest. Set the order you want explicitly with a manual or computed sort rather than assuming the labels drive it.

A calculation still refers to the original value. If you write a calculated field testing [Region] = "East" and the stored value is "E", it returns false. The alias never reached the calculation, because the alias is a display layer.

Aliases do not fix bad data. If a field holds "E", "East" and "east" as three separate members, aliasing all three to East makes them look like one thing while remaining three. That's an entity resolution problem, and it needs cleaning rather than labeling.