Tableau Dashboard Rules: Fixed Sizing, Containers, Fast Extracts
Tableau specifics, not generic design advice: why fixed sizing beats automatic, how layout containers actually nest, and the extract habits that keep it fast.
Gabriel ThieryTableau will let you put anything anywhere, at any size, on any part of the canvas. That freedom is the reason to use it, and it is also why so many Tableau dashboards arrive cluttered, slow, or broken on the first tablet that opens them. Nothing in the tool spends the freedom for you.
Work in Power BI instead? See the companion guide: Power BI dashboard rules.
1. Set the Dashboard Size First
Automatic is the default that costs the most. You build at your own resolution, everything lines up, and then the layout drifts on every screen that isn't yours.
Tableau gives you three sizing modes.
| Mode | What Tableau does | Use it when |
|---|---|---|
| Fixed | Holds the dashboard at one resolution, say 1000×800, on every screen that opens it. | Almost always. Dimensions that never change let Tableau Server and Cloud cache the rendered image, so the dashboard loads faster. |
| Automatic | Resizes the dashboard to fill whatever screen opens it. | You are still prototyping and nobody has published it. |
| Range | Scales between a minimum and a maximum size, then stops. | You know the two or three resolutions your audience runs. |
Canvas size is a decision, and almost nobody makes it. Find out what resolution your audience has, set Fixed to that, and do every bit of layout work at that size. Tableau's own dashboard best practices (opens in a new tab) say to set the size before you start arranging, because layout done at the wrong size gets done twice.
2. Build the Structure with Layout Containers
Layout containers are the feature people skip that changes the most. A horizontal or vertical container (opens in a new tab) groups items so they move and resize together, which is what keeps alignment intact after you change something.
Horizontal containers control the width of the views and objects inside them. Vertical containers control the height. Tiled objects snap into the container grid, and floating objects sit on top at fixed coordinates.
Build the skeleton out of nested tiled containers first, then drop views into it. Save floating for the few things that belong on top of a layout rather than in it: a navigation button, an info icon, an overlay. A dashboard assembled by dragging views onto a blank canvas has no skeleton, and you find that out the first time you resize anything.
3. Limit Views to Two or Three per Dashboard
Tableau's guidance is two or three views per dashboard (opens in a new tab). Every view fires its own query when someone opens the page, so the count is a performance budget you spend before you make a single design decision. Legibility comes along with it.
Framed as a clutter rule, this one is easy to ignore, because five views rarely feel cluttered to the person who arranged them.
Stephen Few defines a dashboard in Information Dashboard Design (opens in a new tab) as a display for at-a-glance monitoring. An analytical workspace is a different product with different rules. When you need more than three views, the answers Tableau gives you are additional dashboards, drill-down, and dashboard actions.
For how many metrics belong on one screen, see how many KPIs should be on a dashboard.
4. Use Dashboard Actions Instead of More Charts
A stakeholder asks for more detail and the reflex is to add a view. Tableau's answer is a dashboard action, which puts the detail one click away instead of on the page.
- Filter actions: click a region on a map, and every other view filters to that region.
- Highlight actions: hover a category, and it lights up across the dashboard.
- URL actions: jump out to the source record or a related report.
- Parameter and set actions: let someone change the metric or the grouping without leaving the view.
One filter action can replace three charts that were showing the same measure cut three ways, which keeps you inside the query budget from the rule above.
5. Move Detail into Tooltips (Viz in Tooltip)
Viz in Tooltip embeds a whole secondary chart inside the tooltip of a mark. Hover a bar, see a mini trend line for that one category, and spend no canvas space on it.
Use it for anything that earns its place only while someone is investigating a specific point. That covers most of what gets added to a dashboard in its second week. The main view stays scannable, and the detail is still there for whoever goes looking for it.
6. Design for Devices with the Device Designer
Automatic sizing will not rescue a phone layout. Tableau's Device Designer lets you build separate Desktop, Tablet, and Phone layouts of the same dashboard, and the published view serves whichever one matches the device that opened it.
On a phone, stack everything into one column and put the most important number at the top. A desktop grid squeezed to phone width produces labels nobody can read and marks nobody can tap. This is the mobile-first card stack pattern, built in Tableau's own layout tab.
7. Optimize Performance Deliberately
A slow dashboard gets abandoned whatever it looks like, and its speed was mostly settled at the data source before anyone opened the dashboard tab. The levers, roughly in the order they pay:
- Use an extract rather than a live connection whenever real-time data isn't the point. Extracts (
.hyper) are columnar and built for the aggregate queries dashboards run, and they usually beat querying the source live by a wide margin. - Run the Performance Recorder (opens in a new tab) (Help → Settings and Performance → Start Performance Recording). It names the slow queries and the slow views, so you fix the bottleneck you have instead of the one you assumed.
- Reduce marks. A view drawing tens of thousands of marks is slow to render before any query runs. Aggregate, filter, or roll up first.
- Reach for context filters and extract filters to cut the data early, rather than filtering a large result set at render time.
- Drop unused fields from the extract. Every column costs size and refresh time.
If a dashboard over a modest dataset is slow, look at the view count, the mark count, and whether the connection is live. Row count is rarely the thing.
8. Format Once, at the Workbook Level
Tableau lets you format per worksheet, which is how one dashboard ends up carrying three fonts and two number formats. Set a colour palette and a number format once, at the workbook level, then apply it everywhere.
The same colour has to mean the same thing in every view: actuals one colour, targets another, red and amber and green reserved for the handful of things somebody has to act on. Keep one accent colour for the element you want noticed. When six views each run their own bright palette, they cancel out and nothing reads as important.
9. Strip the Chart Junk
Tableau's defaults are cleaner than most, and there is still ink to take out. On a typical view you can delete gridlines, zero lines, and heavy borders without losing any information, which is Edward Tufte's data-ink principle at the level you can act on it. Separate views with container padding, inner and outer, instead of borders. White space does the same job without adding a line the eye has to process.
This is the cheapest work on the list and the least valuable. Gridlines were never what stopped a reader finding the answer. Arrangement was. Do rule 2 before you do this one.
10. Whether to Wireframe First
Tableau's build cycle is long: data source, calculated fields, worksheets, containers, actions, device layouts, testing. Rearranging the layout after all of that means re-nesting containers and re-wiring actions, which is why late layout changes get argued about instead of made.
Sketching first makes those decisions cheap while they still are. Plenty of people skip it and ship fine dashboards, usually because they already know the stakeholder and the data. If you don't know either, a rough sketch has one specific advantage over a polished mockup: people keep arguing with a layout that looks unfinished, and stop arguing with one that looks done. Polish reads as a decision somebody already made.
Figma and Miro hand you rectangles, and a dashboard sketch wants KPI cards, charts, and filter placeholders as things you place. I built datawirefra.me (opens in a new tab) for that: drag the pieces onto a canvas, share the URL, and get sign-off before you write a calculated field. Paper works too. The step-by-step wireframing guide has the longer version.
The Short Version
- Set Fixed size first, at the resolution your audience actually has.
- Build the structure from nested tiled containers, and keep floating for overlays.
- Hold it to two or three views, because each one is a query the server runs.
- Reach for dashboard actions and Viz in Tooltip before another chart.
- Build the phone layout in Device Designer rather than hoping.
- Extract first, then measure with the Performance Recorder.
Tableau's flexibility is the reason to use it and the reason its dashboards sprawl. Fixed sizing and layout containers are how you keep the flexibility and stop paying for it every time someone opens the thing on a tablet.
Start with the size. It is the only decision here that gets more expensive with every view you add.
Frequently asked questions
What is the best dashboard size setting in Tableau?
Use Fixed size for most published dashboards. Because the dimensions never change, Tableau Server or Cloud can cache the rendered image, so it loads faster. Set it to your audience's most common resolution and do all your layout work at that size. Automatic sizing is convenient but reflows unpredictably and loses the caching benefit.
How many views should a Tableau dashboard have?
Tableau's official guidance is to limit a dashboard to two or three views. Beyond that, the big picture gets lost in detail and published performance suffers, because every view is a separate query the server runs. If you need more, use additional dashboards, drill-down, or dashboard actions rather than crowding one canvas.
How do I make a Tableau dashboard load faster?
Use an extract instead of a live connection when you don't need real-time data, run the Performance Recorder to find the actual slow queries and views, reduce the number of marks a view renders, use context and extract filters to trim data early, and remove unused fields from the extract. On a modest dataset, slowness almost always comes from too many views, too many marks, or a live connection — not raw row count.
What are dashboard actions in Tableau used for?
Dashboard actions add interactivity that reveals detail on demand instead of showing everything at once. Filter actions filter every view from one click, highlight actions emphasize a category across views, URL actions jump to external records, and parameter or set actions let users change the metric or grouping. A single filter action often replaces several redundant charts.
Gabriel Thiery
Builder of datawirefra.me. I help BI teams plan dashboards people actually use — before they write a single DAX formula.
Connect on LinkedIn