In Tableau, what changes when a field on Rows is discrete (blue) versus continuous (green)?
answer
- one draws labels, the other draws a ruler
- headers and panes versus a single axis
- ask what happens to a month with no rows
- reference lines need something to attach to
- dates come in two flavours from one field
basics
~10 sA discrete (blue) Tableau field draws headers and separate panes; a continuous (green) field draws an axis with a measurable range. The choice is independent of dimension versus measure, and dates can be either.
solid answer
~50 sBlue pills are **discrete**: Tableau treats the values as a finite set of members and draws a **header** for each one, in sort order, with one pane per member. Green pills are **continuous**: Tableau treats the values as a range and draws an **axis**. That single difference cascades. A continuous axis spans the whole range, so a month with no rows shows as a visible gap; discrete headers only appear for values actually present, so the missing month silently disappears. Reference lines and bands attach to a continuous axis, not to headers. A continuous field on Color gives a gradient; a discrete one gives a categorical palette. Filters differ too — a range slider versus a member list. Crucially this is a separate axis from dimension/measure: a date dimension can be continuous, and an aggregated measure can be made discrete.
code
text · 9 lines-- Discrete date part: 12 headers, every January pooled together
Columns: MONTH(Order Date) [blue]
Rows: SUM(Sales) [green]
Jan Feb Mar ... Dec
-- Continuous date value: one axis position per month per year
Columns: MONTH(Order Date) [green]
Rows: SUM(Sales) [green]
2024-01 ... 2024-12 2025-01 ...go deeper
Recall that blue draws headers and green draws an axis, and that the colour is not the same thing as dimension versus measure. Be able to point at a pill and say what the view will look like.
Explain the mechanics and the knock-on effects: panes versus axis, categorical palette versus gradient, member-list versus range filters, and the date part versus date value distinction with an example of each.
Demonstrate the diagnosis: a trend chart that hides an empty period, a reference line that cannot be placed, a colour legend that came out as a gradient. Explain the root cause in pill terms and pick the fix that keeps the chart honest.
Own the convention: which default (discrete or continuous, date part or date value) is set on shared date fields in a published data source, so that dozens of downstream workbooks do not each rediscover the seasonality-versus-trend trap.
## The distinction in one line Discrete means "individually separate": a finite set of members, each drawn as a **header**. Continuous means "forming an unbroken whole": a range of values, drawn as an **axis**. Tableau colours the pill blue for discrete and green for continuous, and that colour is the fastest way to predict what a view will look like. ## What each one does to the view Put a **discrete** field on Rows and you get a stack of row headers, one per distinct value, in the field's sort order. Each header defines a **pane** — its own little region of the view. Panes matter later for things like table calculations and per-pane formatting, and they are why discrete fields nest: `Region` then `Category` on Rows gives categories nested inside regions. Put a **continuous** field on Rows and you get a single axis with tick marks. There are no headers and no panes; every mark is positioned by value along that axis. The axis has editable range, tick intervals and titles, and it can carry reference lines, reference bands and trend lines — features that have nowhere to attach on a discrete header. ## Dates: the case that decides most interviews Dates are where the distinction bites, because Tableau offers both flavours from the same field. - A **date part** is discrete: `YEAR(Order Date)`, `MONTH(Order Date)` as a blue pill. `MONTH` as a date part gives twelve members — January through December — pooling every January from every year into one header. That is exactly right for a seasonality comparison and exactly wrong for a trend line. - A **date value** (truncation) is continuous: a green `MONTH(Order Date)` truncates each date to the first of its month and lays them along a time axis, so January 2024 and January 2025 are two separate positions. Right-click-dragging a date onto a shelf pops the menu listing both sets, which is the reliable way to pick the one you meant. The second date consequence is **missing periods**. On a continuous axis, a month with no rows occupies its position on the axis and shows as a gap or a dip — the absence is visible. With discrete date headers, only values present in the data get a header, so a month with no sales simply is not drawn and the line joins the surrounding months as if nothing were missing. "Our sales never dropped to zero, the chart just closed the gap" is a real reporting bug born from this. ## Discrete measures and continuous dimensions Both off-diagonal combinations are legal and useful, which is why merging the two concepts is the classic error. - **Continuous dimension:** a numeric dimension set to continuous draws an axis — useful for a numeric bin, an age, or a date value trend. - **Discrete measure:** convert `SUM(Sales)` to discrete and its values become headers or text rather than an axis position. This is how crosstabs and some table layouts get built, and it is also how you sort or nest by an aggregated value. ## Other places the pill colour changes behaviour **Color.** A discrete field on Color gets a categorical palette — one distinct hue per member. A continuous field gets a sequential or diverging gradient with a numeric legend. Dropping a continuous measure where you expected categories, or a high-cardinality discrete field on Color, are two of the most common ugly-chart causes. **Filters.** Filtering a discrete field opens a member list with checkboxes; filtering a continuous field opens a range control (at least / at most / between). The same underlying column offers different filter dialogs depending on how you filter it. **Sorting.** Sorting applies to discrete headers. A continuous axis is ordered by value by definition, so there is nothing to sort. **Size and shape.** Size takes a continuous measure naturally as a magnitude; Shape is a discrete channel, since shapes are categorical. ## Converting between them Use the field's or pill's menu to switch between Discrete and Continuous. For dates use the two-part menu (date part versus date value) rather than converting blindly, because the truncation semantics differ, not just the rendering. Changing a field's default in the Data pane affects future placements; changing the pill affects that one placement. ## What an interviewer is testing Three things: that you never say "blue means dimension"; that you can predict the *rendering* consequence (headers and panes versus an axis) before you drop the pill; and that you know the date-part-versus-date-value distinction well enough to pick the right one for a trend, a seasonality view, or a chart that must show missing periods honestly.
- A monthly sales line silently skips a month with no orders. Which pill setting causes that, and what fixes it?Discrete date headers only render values present in the data, so an empty month gets no header and the line closes over the gap. Switch the date to a continuous date value so the axis covers the full range and the missing month occupies its position. Tableau's Show Missing Values option on the date field also restores the empty period when you need headers.
- Why can't you add a reference line to a discrete field in Tableau?Reference lines, bands and distributions attach to an axis, and only continuous fields produce one. Discrete fields produce headers and panes, which have no numeric scale to place a line against. If you need a line across a categorical chart, make the relevant field continuous, or use a constant-value reference line on the measure axis instead.
- What changes when you convert SUM(Sales) from continuous to discrete?The aggregated values stop positioning marks along an axis and become headers or text instead. Practically that turns a bar chart into a crosstab-style layout and lets the measure participate in nesting and sorting like a dimension would. The aggregation itself is unchanged — it is still SUM at the view's level of detail.
- Does a discrete MONTH pill mean the same thing as a continuous MONTH pill?No. Discrete MONTH is a date part: it extracts the month number, so every January across all years collapses into one member — right for seasonality. Continuous MONTH is a date value: it truncates each date down to the first of its month, so each year's January is its own point — right for a trend. Same source field, different semantics.
Discrete is a row of labelled bins on a shelf — one slot per label, with nothing between them. Continuous is a ruler: every position exists whether or not any data lands there.
saying these in an interview costs you the question
- Says blue means dimension and green means measure
- Uses a discrete date part for a multi-year trend line
- Assumes a chart with no gap proves no period was missing
- Thinks continuous only applies to measures
- Expects to add a reference line to a discrete header